Enterprise CX & AI lead generation

A multinational can be the right account and still be the wrong opportunity. Its UK service desk, German returns queue and US renewals team may have different budgets, platforms, owners and sourcing rules. A meeting with the group CIO does not automatically open any of them.

This guide is for commercial teams selling customer experience (CX), AI implementation, business process outsourcing (BPO) and managed operations. It covers firms that advise, build technology, run processes and improve them over time. It is not a guide to generating consumer enquiries for those firms' clients.

Define the unit you are pursuing

Parent account → buying entity → region → journey or process → workstream → owner → next decision.

An actionable target is: “The UK retail operation's refund-status workflow, owned by customer operations, with an agreed review before peak trading.” “A large retailer with lots of customers” is only a research lead.

Start with a decision you can help the buyer make. Should they renew a service contract, add overflow capacity, put an AI use case into production, fix a digital journey, improve renewal coverage or clear a controlled backlog? Pick one. Do not open with the whole catalogue.

The first conversation should produce a bounded next step: a baseline review, technical fit check, approved sourcing route or a decision not to proceed.

Use this guide in three passes

The main constraint to test is access to a funded, feasible workstream, not the size of the contact database. If meetings repeatedly die in discovery, repair the brief before increasing outreach.

All schedules, thresholds, scorecards and examples are proposed operating tools. They are not sector benchmarks. All worked numbers and example accounts are fictional. Replace assumptions with approved buyer or provider evidence before making a live decision.

An assessment, a platform build, an outsourced service and an AI product can sit inside one account. They are not the same sale. Each needs its own acceptance event, owner, cost basis and route to revenue.

Build the sequence around the decision

A sensible route might be workstream review → paid design → controlled implementation → managed operation. It is not mandatory. A buyer with an approved request for proposal may be past discovery. Another may buy a stand-alone assessment and keep its incumbent.

State what is free, what is paid and who owns the output. Do not manufacture a pilot to evade procurement or label a sales call a workshop when it produces no usable deliverable.

Know what happens when automation succeeds

In a people-priced contract, fewer staffed hours may reduce provider revenue. With transaction pricing, retries may inflate billing unless excluded. With outcome pricing, the provider may carry risks it cannot control.

Before pitching AI, answer: What changes in the buyer's cost? What changes in the provider's revenue? Who pays for technology and monitoring? How will gains be verified and shared?

Consider a base fee for continuing responsibilities and a clear variable unit where justified. It is an option to evaluate, not a universal price structure. Specify volume bands, minimum commitments, service credits and change control.

Pipeline rule: record the first purchasable scope. A small assessment is not a full outsourcing win. A hoped-for global rollout is not contracted value.

Use company size to find candidates, not to prove demand. A large workforce does not reveal service-contact volume, outsourcing willingness, available budget or the cost of a specific process.

Exclude early without pretending to know too much

Hold when geography, ownership or the supposed renewal window is unclear. Reject an explicit no-contact instruction, binding conflict, delivery prohibition or procurement exclusion.

“No budget allocated” need not kill early discovery. “No owner, no decision and no willingness to explore one” should stop expensive solution work. Ask for the minimum missing evidence that could reverse the choice.

Start from work that was good for both sides

Review completed engagements by workstream, not only vertical. Compare contribution after transition and rework, retention, time to a useful result, pursuit cost and referenceability. Look for shared traits among engagements that were valuable to buyers and viable to deliver.

Create a narrow campaign cell: “Regional telecom operators reviewing billing-contact operations after a confirmed platform change, with permission to consider hybrid delivery.” Treat this as a hypothesis, not a claim that every firm in the cell needs help.

Write exclusions into the campaign brief. An exclusion protects sellers, brand and delivery more than an impressive but vague addressable-market number.

Action: approve the unit, scope, evidence and exclusions before buying more contact data.

A signal gives you a reason to ask. It does not establish pain, budget, outsourcing intent or dissatisfaction with an incumbent. Record the original source, event date, affected unit and a recheck date.

Write an evidence sentence

“Source X reports event Y on date Z. We infer that workstream W may need review. Pain, funding and supplier change are not confirmed.”

Prefer first-party announcements, public procurement notices, annual reports, product notices and relevant job descriptions. A job advert can suggest a role or technology. It does not prove a funded project or installed system.

Review-site complaints can suggest questions, not incidence rates or causes. Do not turn a sensitive incident into fear-based outreach or suggest access to private facts.

Set freshness by the event. A launch window may expire quickly; an architecture constraint may persist. Recheck when the next action depends on it rather than applying an arbitrary universal expiry.

Action: every active record needs a fact, a labelled inference and one unanswered buying question.

A senior-title list is not a buying group. Map what each person can approve. One person can hold several roles; several people can share a role across regions.

Run two connected conversations

The operational conversation asks whether the work should change and what success means. The approval conversation asks whether that change can be bought, implemented and governed. Bring procurement in through the buyer-approved route. Do not bypass it.

One informed owner may be enough for an initial discussion. That person alone cannot approve a production pilot when data, architecture or risk approvals remain unresolved.

Record “role inferred” separately from “authority confirmed.” A positive reply is not proof of budget ownership. A referral is useful when it clarifies the next decision, not merely another email address.

Make the internal forward useful

Give a champion one page: the workstream, baseline, proposed test, people needed and decision requested. Include the strongest reason not to proceed. Ask them to correct it before it travels.

Coordinate through one account owner. Do not send identical pitches from several service lines to multiple executives in the same week. The buyer should receive one coherent conversation.

Action: fill the Buying Group tab with decision rights and evidence, not just contact details.

A large potential contract does not override a blocked delivery model. Separate hard gates, research priority and commercial judgment. Do not compress them into one impressive lead score.

Pass hard gates first

Confirm an allowed contact route, no conflicting account ownership, plausible service fit and no known delivery prohibition. Unknown means investigate. Explicitly prohibited means stop. A high score overrides neither state.

Then compare five questions: Is there a specific decision? Is the work meaningful? Is the next owner reachable through an allowed route? Can you show relevant proof? Can you staff the next step without hurting live delivery?

A proposed research score can use 0 for absent evidence, 1 for a plausible hypothesis and 2 for support. Keep evidence beside it. The score is not a probability of buying.

Protect scarce specialist time

Set a weekly limit for solution-engineer hours used before qualification. Choose it from actual capacity, not a universal benchmark. Review how many hours produced a buyer-agreed next decision.

Before a full proposal, estimate remaining pursuit cost and contribution if won. Use a range for uncertain win assumptions. Large total contract value does not make weak economics attractive.

Start where the important assumptions can be verified at the lowest cost. Existing-account expansion may beat a net-new bid, but only when the service is healthy and its owner approves.

Action: classify each pursuit as pursue, research, nurture or stop. Name the owner, missing evidence, next action and review date. Fund only the next commitment that the evidence supports.

Sell the decision: whether a defined queue should remain internal, stay with its incumbent, move to another provider or use a hybrid model. A carve-out is a bounded portion of work, not a promise to replace the whole operation.

Start from a buyer-confirmed review, new language, acquisition integration, service expansion or approved overflow need. Hiring is not proof of failure. Check whether headcount represents growth, replacement or work that must stay internal.

Prepare a queue and transition brief

Research the customer-operations owner and vendor-management route together. Bring workforce planning into a capacity discussion. Reserve the full solution team until the workstream is clear.

Request a decision-shaped meeting

“Would a short review of the [queue] scope help before you decide whether to extend, supplement or re-tender it? The output would be a transition-risk brief, not a generic capability presentation.”

Begin with aggregate information. Do not request customer recordings in the first email. Ask which delivery alternatives the buyer is genuinely allowed to consider.

Advance: the scope owner confirms meaningful work and an allowed buying route exists. Nurture: timing is real but distant. Stop: the requested cutover requires unapproved access, unsupported language skills or unsafe overlap.

Show proof from a comparable queue, coverage model and transition. State what your team controlled and what it did not. A list of client logos cannot answer how the service was mobilised.

Expansion path: stabilise the first queue, then review languages, channels, knowledge, quality or automation. Each expansion needs authority and a fresh scope. One contract does not automatically cover every operating country.

Sell the decision: whether one customer task can be completed safely through automation and what human service remains. An AI agent here means software permitted to take defined actions in systems, not merely generate text.

Choose a task: checking a refund, retrieving an invoice, changing a delivery instruction within policy or completing an allowed account-access step. Define identity checks, source of truth, action permissions and failure route.

Prepare a resolution-boundary map

A public AI announcement is a reason to ask which use case has a production owner. It does not prove a stalled pilot or a funded external requirement.

NIST's generative-AI profile addresses risks across the lifecycle, including evaluation and third-party components. It is risk context, not proof that an individual system is safe. [S2]

Request the first meeting

“Before adding automated journeys, would it help to map one task from request to verified completion, including the cases that should go straight to a person?”

Do not claim that an avoided contact removes its full accounting cost. Separate workload released, capacity redeployed, actual spend removed and new technology costs. Chapters 30–31 provide the model.

Production gate: agree the task outcome, evaluation set, permissions, human fallback, data treatment and accountable operator before any launch. A successful demonstration passes none of those gates by itself.

Expansion path: another verified intent, then language or channel expansion after re-evaluation. A working English text flow does not prove that a voice or multilingual version is ready.

Sell the decision: which part of an employee's workflow can improve without moving work or risk somewhere else. This is a different sale from replacing human service.

Separate knowledge search, response drafting, summaries, after-contact work, quality review and coaching. One “productivity” promise hides different owners and acceptance tests.

Research a confirmed knowledge consolidation, platform change, product launch or training need. Job adverts can help identify ownership; they do not prove poor performance. Start with operations or quality, then add knowledge and platform owners.

Offer an agent-workflow friction review

Map a representative case from acceptance to completion. Record system changes, search time, customer holds, after-contact work, corrections and escalation. Separate waiting time from active employee effort.

Request the first meeting

“Could we compare one agent workflow before and after assistance, including corrections and after-contact work? That would show whether time is being saved or moved.”

Compare cohorts by case mix, tenure, channel and demand. A lower average handling time may reflect easier work. Check which time components the platform records rather than assuming a shared definition. [S10]

Do not sell automatic employee discipline based on model scores. Keep coaching and monitoring within approved workforce policies and applicable requirements.

Proof: show workflow evidence, adoption, quality, rework and a fair comparison. Tool usage is not proof of improvement. Avoid a case study that counts saved after-contact time but ignores extra corrections.

Stop: no reliable knowledge source, no accountable owner or no method to review quality. Expansion: knowledge operations, languages, quality and managed monitoring. Autonomous actions require a separate outcome and risk review.

Sell the decision: whether a specific journey gap needs redesign, integration, migration or better operation. Do not assume the answer is a replacement platform.

Entry points include CRM or contact-platform migration, acquisition integration, self-service failure, disconnected commerce and support, or a new digital product that generates service work.

A web team may own the journey while operations handles its failures. IT may own the platform while nobody owns knowledge after go-live. Treat this as a discovery question, not an outside diagnosis.

Prepare a journey-to-operation dependency brief

Map the experience owner, architect, business process owner and future service owner. Ask who runs the workflow after the project team leaves.

Request the first meeting

“Is the [journey/platform] project also covering what happens when customers cannot finish the task, and who owns that work after launch?”

Compare real alternatives: fix the current platform, use native features, engage another integrator, build internally or change the process without new software. A credible “keep what works” option makes the discussion useful.

Start with a high-level architecture conversation. Never ask cold prospects for credentials, restricted diagrams or production logs. Arrange authorised access only through the security process.

Advance: there is a journey, sponsor, architecture owner and credible route to scope funding. A technical curiosity call without a business decision remains technical discovery.

Proof: show a comparable workflow, deployment boundary, integration responsibility, adoption result and post- launch operating model. A partnership badge is not proof of every delivery capability.

Expansion: scoped build, explicitly priced run service, then evidence-based improvement. Count each only at its own accepted stage, not as one assumed transformation contract.

Sell the decision: which customer segment needs a different coverage model and what the provider can own. Demand generation, inside sales, customer success, renewals and channel operations have different controls.

Keep two funnels separate

Provider acquisition: your team wins an enterprise contract to run a workstream.

Client revenue delivery: the contracted team generates, progresses, supports or renews the client's customers.

A meeting in the second funnel is not a new client in the first. Never mix counts, revenue or acquisition costs.

Offer a coverage-and-attribution brief

Define eligible accounts, territories, ownership, existing coverage, commercial goal, handoff, attribution window and exclusions. Use the client's CRM definitions.

“Would it help to isolate one under-covered segment and agree what an outsourced team would own, what your sellers retain and how incremental results would be measured?”

Ask who controls pricing, product fit, closing, channel conflict and approval. Do not guarantee closed revenue when these remain outside the provider's control.

Advance: a defined segment, approved data route and a named person to accept or reject handoffs. If internal follow-up capacity is absent, repair it before adding demand.

Proof: cohort-level output with attribution and the work actually performed. Separate pass-through advertising spend from provider service revenue.

Expansion: adjacent segments, customer success or partner operations. A good result in one cohort does not give permission to reuse client data or contact customers outside the agreed scope.

Sell the decision: how to handle a defined policy or risk queue correctly, consistently and within its response window. Moderation, identity checks, fraud review and regulated casework are not interchangeable analyst seats.

For content, ask about policy, language, priority and appeals. For identity or fraud, ask what evidence reviewers may access and which actions they may approve. For regulated work, establish where professional judgment must remain.

A new category, market or confirmed backlog can justify a question. A public incident does not prove that the prospect is unsafe or non-compliant.

Prepare a queue-control map

Request the first meeting

“Could we map one queue from intake through decision and appeal, including the cases that must remain with your internal team?”

Start with the policy or queue owner. Add risk, legal, security and operations as required. A customer-service title does not establish authority over trust-and-safety policy.

Where UK financial-services outsourcing rules apply, the regulated firm retains responsibility. Do not position outsourcing as a transfer of regulatory accountability. [S4]

Proof: comparable policy complexity, calibration, review turnaround and escalation handling. Volume processed alone is insufficient.

Stop: undefined policy, unavailable escalation cover, unsupported regulated work or a request to maximise approvals, removals or recoveries regardless of correctness.

Expansion: adjacent queues, languages, quality review or policy operations only after fresh calibration, access and staffing checks. This is a controlled operating service, not an unlimited pool of interchangeable labour.

Sell the decision: what evidence the buyer needs before trusting a model or workflow, and who maintains that evidence. This is not the same as implementing a customer-facing bot.

Work can include permitted data preparation, labelling, human evaluation, multilingual review, task-specific testing, feedback and production-quality monitoring.

A funded product release, language launch, policy change or model update may create a need. Ask which deployment decision depends on the work. “We need more data” is incomplete without intended use and acceptance.

Prepare an evaluation-coverage brief

Research the AI product owner, evaluation lead, data-governance owner and operating lead. Procurement may buy units, but the technical owner must define usable output.

NIST discusses downstream quality risks from data and third-party components. Include evaluation-data quality rather than treating labels as unquestionable ground truth. [S2]

Request the first meeting

“Which release decision is the evaluation meant to support, and where is your current test set least representative of real use?”

Do not price every example as equal. A simple label, specialist review and multilingual adjudication have different costs. Define rejected work, rework and acceptance before unit pricing.

Avoid evaluation leakage. Do not tune on the same examples used to claim independent performance. Record versions of models, prompts, policies, retrieval sources and tests where relevant.

Advance: lawful access, task owner, agreed rubric and a release or operating decision. Stop: no data rights or no valid evaluation route. Expansion: recurring quality operations, then adjacent tasks or languages with separate validation.

Sell the decision: how to resolve a transaction or exception queue without creating errors, customer harm or accounting confusion. Billing enquiries, invoice disputes, cash-application exceptions, order-to-cash handoffs and permitted collections support are distinct scopes.

Follow one exception

A customer questions a charge. Service forwards it to finance. Finance lacks order context. The customer calls again. Selling a cheaper service agent leaves the loop intact.

Research the owner across service, billing and finance. Establish who changes policy, corrects the transaction and sends final confirmation.

Offer an exception-flow brief

“Could we trace one billing exception from customer contact to ledger update and final confirmation? That would show whether the fix sits in service, finance or the handoff.”

For collections, confirm jurisdiction, permissions, customer protections and authorised escalation before designing live work. Do not use aggressive scripts, unrelated personal data or promises of recovery as a shortcut to qualification.

Keep the economics honest

Collected customer money is not automatically provider revenue. Faster collection improves cash timing; it does not automatically create revenue or profit. Separate recovery fees, transaction charges and managed- service fees.

A change in days sales outstanding may reflect customer mix, payment terms or seasonality. Define the eligible portfolio, comparison, disputed amounts and exclusions before attributing the change to the service.

Advance: one queue, authorised process owner, reconciliation method and explicit approval limits. Expansion: adjacent exceptions, documents, communication or analytics after the first loop is controlled.

Add AI only where information quality, allowed actions and accountable review make it appropriate. Do not automate a broken cross-team process just to shorten the demo.

These are campaign hypotheses, not claims about any prospect. Choose one regulated entity, product and process. Confirm local requirements with the buyer's specialists.

Banking and payments: disputes and servicing

Entry signal: a new product, acquisition, servicing-platform change or confirmed review of a dispute queue.

First owner: Head of Servicing, Card Operations or Disputes. Map fraud operations, compliance, customer support and the economic sponsor separately.

Opening question: “Is the review focused on the first customer contact, the investigation queue or keeping the customer informed while the case is open?”

First asset: a contact-to-case map covering authentication, intake, evidence, escalation, status updates and closure. Mark decisions that must remain with the institution.

Qualify: case types, jurisdiction, service rules, sensitive data, authority, incumbent scope, complaints, approved delivery locations and change window. Ask how unresolved cases are owned during handoffs.

Measure: case age, correct routing, rework, completed outcomes and repeat contacts. Faster closure is not better if the decision is wrong.

Stop: unapproved autonomous decisions, concealed subcontracting or an expectation that outsourcing transfers accountability. FCA guidance makes retained responsibility explicit where applicable. [S4]

Insurance: claims intake and status updates

Entry signal: a portfolio acquisition, distribution launch, claims-system change or planned surge review. A catastrophe is not permission for pressure-based selling.

First owner: Claims Operations or policy servicing. Separate intake from licensed assessment, coverage decisions, settlement authority and broker service.

Opening question: “Which part of the claims journey generates follow-up work that could be reduced without changing who makes the coverage decision?”

First asset: a claims-status brief covering required evidence, customer messages, missing-document loops and urgent escalation.

Qualify: product line, policy and claim systems, authority limits, data, surge assumptions, partner roles, vulnerable-customer needs and capacity decision date.

Measure: complete intake, appropriate escalation, time to the next valid step and customer comprehension. Settlement value alone cannot establish service quality.

Expansion: prove one process before adding languages, agent assist or broader administration. A successful intake queue does not validate an entire claims portfolio.

Message discipline: do not infer non-compliance, understaffing or buying intent from an annual report. Ask a precise question that an informed owner can correct.

Administrative support must not drift into promises about clinical judgment, eligibility, vulnerable customers or safety-critical work. Define the service boundary before the list.

Healthcare: access, scheduling and member service

Entry signal: a new region, payer programme, provider acquisition or approved access-centre review.

First owner: Patient Access, Member Services or Contact Centre Operations. Add clinical governance when the boundary touches care, and privacy/security before sensitive information is exchanged.

Opening question: “Is the priority completing an administrative task, or reducing handoffs when it needs a specialist?”

First asset: an administrative journey map covering identity checks, scheduling or status, exclusions, approved information and specialist escalation.

Qualify: payer versus provider; administrative versus clinical scope; populations; languages; accessibility; protected data; urgency; systems; approved access and specialist cover.

Measure: task completion, correct routing, abandoned attempts, repeat contacts and appropriate escalation. Shorter calls must not substitute for safe access.

In relevant US arrangements, HHS guidance describes business-associate status and written safeguards for protected health information. Determine applicability before sharing that information. [S5]

Stop: a request for diagnosis, clinical triage or restricted work outside authorised capability. Use aggregate or synthetic information during early sales discussions.

Energy and utilities: billing, moves and exceptions

Entry signal: a tariff or billing-platform change, market expansion or confirmed service review.

First owner: Customer Operations or Billing. Map collections, vulnerability support, digital and emergency- response ownership separately.

Opening question: “When customers cannot complete a move or billing task online, where does the unresolved work go?”

First asset: a move-in/move-out or bill-query map with account validation, meter or billing data, exceptions, handoff and confirmation.

Qualify: market, customer type, permitted advice, emergency exclusion, accessibility, dispute routes, systems, seasonality, hours and approved delivery locations.

Measure: completed transactions, repeat contacts, corrections and suitable escalation. Routine account work must remain distinct from emergency or essential-service decisions.

Expansion: one administrative journey to knowledge, coverage or automation under a new approval gate.

Message discipline: a public disruption is not proof that outsourcing would solve it. Ask about an approved planning process rather than promising to fix an incident you have not diagnosed.

Distinguish temporary demand, avoidable work and structural process failure. Each leads to a different offer and contract.

Retail and ecommerce: post-purchase exceptions

Entry signal: a new market, fulfilment route, returns-policy change or confirmed peak-planning window.

First owner: Customer Care or Ecommerce Operations. Add logistics, payments, digital product and finance when the exception crosses their systems.

Opening question: “Does the peak plan focus on extra capacity, or on completing order and refund tasks without another contact?”

First asset: a post-purchase exception map. Pick order status, late delivery, returns or refund status. Identify the source of truth and the action service staff can take.

Qualify: store versus online scope, brands, countries, languages, order and fulfilment systems, refund authority, fraud checks, stock rules, contact patterns and change freeze.

Measure: verified completion, repeat contact, refund accuracy, exception age and total cost. A contact avoided is not proof of a retained order.

Expansion: solve one loop, then test languages, channels or product categories with fresh volume and policy checks.

Travel and transportation: disruption and after-sales service

Entry signal: route expansion, reservation-system change or a planned continuity review.

First owner: Customer Operations, Guest Services or Disruption Operations. Map authority to change, refund or rebook separately from the service queue.

Opening question: “Where does the flow break when a request falls outside normal rebooking or refund rules?”

First asset: a normal-versus-disruption service brief. Define allowed actions, partner dependencies, priority cases, agent authority and surge activation.

Qualify: direct versus intermediary bookings, reservation and payment access, supplier rules, languages, time zones, refunds, compensation decisions and escalation cover.

Measure: correct completion, repeat contacts, time to a valid next step, error cost and customer effort. Do not promise one handling time across routine and disrupted journeys.

Stop: urgent coverage requires training or access that cannot be completed safely. Offer a future continuity review instead of an impossible mobilisation date.

Message discipline: avoid “we will cut costs during your busiest week.” Lead with the decision, exclusions and safe implementation window. A larger queue is not automatically an outsourcing opportunity.

Choose the economic job before the persona. Technical support, renewals, advertiser growth and subscription retention are different campaigns.

Technology and consumer electronics: support or renewals

Entry signal: a product launch, channel expansion, acquisition or confirmed coverage change.

First owner: Support Operations for technical cases; Customer Success or Renewals for recurring accounts; channel leadership for partner work.

Support question: “Which cases can the first-line team complete, and which lose context on the way to engineering?”

Renewals question: “Which cohort needs coverage your team cannot give it without duplicating named- account ownership?”

First asset: a support-tier boundary map or an eligible-renewal cohort brief. Pick one.

Qualify: complexity, entitlement, tiers, authorised diagnostics, warranty rules, systems, languages, territories, reliable renewal dates and commercial authority.

Measure: resolved issues, proper escalation, age and quality for support. For renewals, use the eligible cohort, retained revenue, margin and exclusions. Product usage is not a renewal.

Expansion: support to knowledge/quality, or one renewal segment to adjacent segments. Do not combine the two cases to inflate the business case.

Telecom and media: activation, billing or retention

Entry signal: a subscriber-platform change, new bundle, service launch or customer-journey review.

First owner: Customer Operations, Retention or subscriber operations. For an advertising platform, target the advertiser-lifecycle owner instead.

Opening question: “Is the priority correct first activation, fewer billing follow-ups or a better cancellation conversation?”

First asset: one journey map with policy limits on credits, retention offers, cancellation and technical escalations.

Qualify: customer type, bundle, sales versus service authority, approved offers, billing access, fault boundaries, subcontracting, hours, quality and complaint rules.

Measure: successful activation, correct first bill, repeat contact and customer-understood outcomes. Do not reward agents solely for preventing cancellation. Friction or an unsuitable offer is not sound retention.

Expansion: repair the handoff before expanding headcount. Add AI only with a defined task, knowledge owner, permissions and fallback.

Message discipline: churn and complaints may frame a question. They do not establish which workstream caused the result or whether an outside provider is wanted.

A brand, dealer, finance arm, agency and programme office can share a customer journey without sharing buying authority.

Automotive and mobility: ownership and service

Entry signal: a connected-service launch, vehicle launch, dealer-system change or regional expansion.

First owner: Customer Experience, Aftersales, Connected Services or Mobility Operations. Confirm whether the manufacturer, dealer group, finance business or platform owns the contract.

Opening question: “When an owner starts a request digitally, who owns completion if the next step sits with a dealer or another unit?”

First asset: an owner-to-service handoff map: identity, entitlement, appointment, dealer confirmation, status and exception.

Qualify: market, product, dealer responsibilities, warranty authority, technical scope, safety exclusions, customer/vehicle data, scheduling access, language and hours.

Measure: completed handoffs, correct appointments, repeat contacts, exception age and appropriate escalation. Do not claim that a contact-centre change improves vehicle safety or repair quality without evidence.

Expansion: one ownership journey before dealer support, finance servicing or connected-service operations. Each requires its own authority.

Public services: information and administrative support

Entry signal: a public procurement notice, programme launch or authorised market-engagement event.

First owner: programme owner and the procurement contact named in the process. Respect the prescribed communication route.

Opening question: “Which parts of the service can a supplier operate, and which decisions must remain with the public body?”

First asset: an eligibility and scope checklist based on public information. Do not offer private help to shape evaluation criteria or bypass competition.

Qualify: supplier eligibility, accessibility, language, records, security, location, continuity, background requirements, contract terms, mobilisation and clarification rules.

Measure: access, correct information, task completion, equitable service and suitable escalation. Contact volume is not a programme outcome.

Stop: missing mandatory eligibility, improper influence as the only route or a required capability that cannot be delivered.

Expansion: only through a valid contracting route. A relationship with one programme does not authorise another.

Message discipline: never imply an endorsement, certification or procurement status that has not been granted. A compliant early market discussion is not a contract award.

Assign one researcher and one account owner. Research structure before buying contact records. The output is a usable workstream brief, not a dossier.

Pass 1: decide whether the pursuit is real

Read the official site and the most relevant first-party announcement or notice. Identify what the company sells, the affected unit and the process your service could address. Check CRM ownership, open opportunities, existing contracts, previous refusals and suppression.

Stop if the signal is irrelevant, delivery is outside scope or another team owns the conversation. Do not buy contacts for a failed account.

Pass 2: make one workstream testable

Record the actual URL and date. Keep the fact and interpretation separate. A job description can suggest a role or system, not a funded programme.

Enrich only the next useful contacts

Find the operational owner and one relevant approval role through legitimate sources. Verify role, employer, geography and details. Technical deliverability is not permission to market. Apply the regional review in Chapter 36.

Use no leaked customer records, credentials or unrelated personal details. Public means observable, not automatically reusable for every purpose.

Release the record through a quality check

The account owner checks workstream, evidence, delivery fit, contact route and existing activity. The researcher records a next action and review date.

Start with a batch small enough to inspect every record. Increase capacity only when sales uses the briefs and source quality holds. If sellers repeatedly discard records, fix the definition of fit before asking researchers to produce more.

Output: one brief that makes the first conversation intelligible to both the buyer and the receiving seller.

Choose distribution from the buying conditions. Cold email is not automatically the primary route to every enterprise contract.

A partner brief that can be used

State the unit, workflow, decision, platform, your role, partner role, evidence, commercial route and one introduction requested. Add what you will not sell into the account.

Microsoft describes co-selling as collaboration with separate programme statuses and eligibility. A partnership does not automatically qualify every offer for funding or marketplace benefits. Verify the programme before claiming either. [S11]

Events should create context, not a blast list

Choose relevant sessions and public business topics. After a real conversation, send the promised asset and refer to the decision actually discussed. Do not write “great meeting you” when you only saw a name on an attendee list.

Keep account activity coherent

One owner should see direct, partner, event and service-line activity. Log the introduction source and current conversation before the next message.

A partner introduction and a cold reply from the same account are not two opportunities unless they concern separately qualified workstreams. Several relevant contacts can support one buying decision.

Measure accepted opportunities and buyer-agreed next steps by route. A large stack of signed partnerships is not a distribution channel if none produces useful customer conversations.

Action: choose the closest credible route for the first cohort. Add another route to solve a specific access problem, not to create more activity.

These are original starting points, not performance-tested scripts. Replace brackets with verified information. Offer an asset only after it exists. Use a lawful route and your approved sender block: real identity, company, contact details, required address, commercial disclosure where applicable and an easy opt-out. B2B email is not exempt from US CAN-SPAM requirements. [S7]

1. Confirmed service review

Use a buyer-confirmed review, not a renewal date inferred from contract age.

If they engage, confirm the sourcing route and next decision before requesting detailed volumes.

2. New region or language

Use a verified expansion that your delivery team can support.

Do not imply the current team cannot cope. The work may already be staffed.

3. Capacity versus avoidable work

Use while there is still a safe planning window.

When the change window is closed, offer a later review. Do not invent a rapid transition promise. A useful email leaves room for correction and offers one bounded next step.

Use a production decision, employee workflow or post-launch responsibility as the reason for the conversation. “AI transformation” is too broad by itself.

4. Pilot to operating service

A public AI initiative is relevant; production ownership may still be unknown.

Ask who owns the outcome after launch. Do not claim the pilot has stalled.

5. Agent-assist workflow

Use when knowledge, agent tools or quality operations are under review.

Only use proof that compares similar work. Do not disclose another client's restricted screenshots.

6. Platform deployment to run ownership

Use a confirmed project with a plausible operational handoff.

This asks about ownership rather than offering to replace a platform already selected.

Match the next step: send the task boundary, workflow comparison or run checklist. Resource interest is not permission to request logs or schedule a long workshop.

These scopes belong to different owners. Route them separately even inside the same account. Use accurate identity and the approved contact controls on every message.

7. Revenue coverage

Use a defined segment, region, product or renewal cohort.

Ask whether the gap is finding, progressing, closing or renewing demand. Not every gap needs more appointments.

8. Trust-and-safety queue

Use a new category, policy or market with plausible relevance.

Do not imply the platform is unsafe, unlawful or failing its users. None of those facts has been established.

9. Evaluation coverage

Use a task, model or language release with an evaluation owner.

Ask what release or operating decision the evidence supports. Cheap labels without a task definition are not yet an evaluation opportunity.

Send test: name a real workstream, invite correction and offer a useful asset. Without a verified event, remove the event claim and ask an honest role-relevant question.

Use these when the opportunity is a handoff rather than a whole department. Keep restricted decisions outside scope unless capability and buyer approval have been established.

10. Billing exceptions

Research the likely process and role first. Do not state that a backlog exists unless confirmed. Ask who can reconcile a changed balance and who must authorise a customer commitment.

11. Administrative healthcare journey

This concerns an administrative workflow. It is not clinical triage, medical advice or an attempt to access protected records. Confirm population, urgency and specialist cover before any live scope is discussed.

12. Digital journey into assisted service

The buyer can clarify whether this belongs to digital, operations, IT or nobody's present priority. A clear “not a priority” is useful information.

Keep the boundary clear: a free checklist is not a custom integration or production-ready design. Prepare the promised asset first. Apply the approved sender and contact controls.

The first exchange confirms relevance and ownership. Identify yourself, use an approved route and respect a refusal.

Phone and LinkedIn openings

“Hi [Name], it's [Name] from [Company]. This is a business-development call about [workstream]. I saw [verified event]. Have I reached the person who owns that decision, or does it sit elsewhere?”

When relevant: “Is the review about capacity, how the process works, or both?” Listen before proposing a specialist meeting.

A LinkedIn note can say: “Your post on [actual topic] raised the handoff between [team A] and [team B]. We work on that boundary in [service]. Happy to share a short checklist.”

Use permitted platform features. A connection acceptance is not permission for unrestricted email or phone marketing. Voicemail should not imply an existing relationship or a false operational emergency.

Handle the reply, not your preferred pitch

When a meeting is welcome

Confirm topic, participant, intended output and time zone. Offer suitable times or a scheduling link according to preference. Do not send an invitation before agreement.

Record the buyer's actual meaning. Polite curiosity is not project intent. A refusal is not an objection to overcome through another channel.

The reply owner should leave the record in a clear state: meeting proposed, asset sent, referred, nurture at an agreed point, no fit or suppressed.

A sequence is a ceiling on attempts, not an instruction to keep contacting someone who has refused. Local law, channel rules and recipient preferences take priority.

A starting sequence to test

Intervals should reflect the buyer, event and account owner's plan. This is not a universal cadence. Avoid daily pursuit of senior buyers across channels. Active conversations leave automation immediately.

Staff the reply desk

Assign a primary owner and cover. Respond during the buyer's business hours as promptly as actual staffing allows. Record receipt, first meaningful response and next action. Do not promise response speeds the team cannot deliver.

Separate workstream interest, information requests, wrong role, timing, procurement, no fit, opt-out and automated replies. Courtesy messages and automatic responses are not positive intent.

Technical controls

Use approved sending identity and authenticated infrastructure. Verify data quality, monitor failures and complaints, and follow provider requirements. Google's sender guidance concerns authentication and behaviour; meeting it does not confer marketing permission. [S9]

Do not use fabricated warm-up activity, rotating identities or domains to evade spam controls. Protect the company's security and brand as well as deliverability.

Learn where work stops

Relevant replies without bookings can signal a poor meeting offer. Held calls rejected by sales can signal qualification failure. Opportunities that stall in solution design can signal weak scope, economics or approval access.

Inspect actual conversations before changing the campaign. More sending cannot repair a missing buyer- owned decision.

Scale gate: increase volume only when research, reply handling, sellers and specialists can absorb the next step. Measure the useful output, not the activity that is easiest to count.

Separate ready to book, held and sales accepted. The workbook and web checker assess readiness only. A person confirms attendance and acceptance later.

A No on the allowed route or capability fit stops this path. A No, Unknown or blank elsewhere means hold and resolve it. All Yes means ready to book, not a guaranteed opportunity.

Use the lightest useful meeting

An ownership call finds the right participant. Scoped discovery tests a need. Technical discovery checks a known dependency. Procurement discussion clarifies the route.

Do not put all four into one invitation. A senior attendee does not turn an introductory discussion into qualified project discovery.

A fully approved budget or final architecture need not exist before early discovery. But deep engineering work needs a decision, sponsor and credible approval path. Do not confuse a useful early discussion with a funded buying project.

Review sales acceptance after the call

Record whether it happened, who attended, whether the promised topic was discussed and whether the receiving team accepts it under agreed criteria.

Use specific rejection reasons: wrong workstream, no buyer agreement, unsupported scope, duplicate opportunity, unsuitable participant or misleading invitation. Do not reject a well-scoped first call solely because the sale did not close.

Where payment or incentives depend on meeting events, use written definitions. Do not reward a setter for a meeting the buyer never knowingly accepted.

Action: agree the gate before launch, then review both seller rejection and buyer experience. Changing the criteria after results arrive makes measurement unreliable.

Send the agenda before the call. Ask only what is needed for its purpose, then follow the buyer's answers rather than reading a long checklist.

Ask the question that fits the workstream

Managed service: “What can change: provider, scope, location, technology or only price?”

AI: “What task can the system complete, and who handles what it cannot?” Ask about knowledge, evaluation and operating ownership.

Platform build: “Who accepts the technical work, and who owns the business outcome after deployment?” They may be different people.

Revenue service: “Which accounts and actions would the outsourced team own, and how will incremental results be identified?”

Controlled casework: “Who sets policy and resolves disagreement when an analyst or system cannot decide?”

End with an explicit branch

Arrange a baseline review when missing evidence is obtainable. Bring a specialist when a specific dependency must be tested. Follow procurement when a formal route exists. Nurture a genuine future decision. Stop when the premise was wrong.

Do not promise a complete solution before the relevant specialists inspect the scope.

Give the receiving team a usable packet

Record workstream and region; the buyer's description; facts versus hypotheses; current model; workload and source; exclusions; systems; stakeholders; timing; procurement; risks; material unknowns; promised asset; and next decision.

Attach authorised notes or the relevant thread as appropriate. Recording is not automatically permissible. Apply consent, security and retention rules.

The receiving specialist should be able to answer: Why this meeting? Why this work? What must be decided next? If not, return the brief for clarification instead of booking another vague workshop.

A good handoff protects both sides from repeating the same discovery and paying for avoidable design work.

Agree definitions before pricing or launching a pilot. A platform metric is not automatically a contract outcome. Amazon Connect defines contact and timing measures at particular stages; check the actual source system instead of assuming universal denominators. [S10]

Keep three benefits separate

Workload released: fewer hours needed for the measured work. This may exist without lower staffing cost.

Capacity redeployed: those hours are assigned to useful work elsewhere. Document that work. Do not also claim the same hours as removed payroll.

Cash saving: a documented reduction in spend, such as avoidable overtime or an external invoice. Name the action, owner and date that turns capacity into cash.

Include failed automation attempts, knowledge maintenance, quality review, security, integration, run support and human fallback. Residual cases may be harder, increasing their handling time.

The bounded model in this kit

The buyer model covers requests that each previously required one assisted contact. Each eligible request gets one automation attempt. A failed attempt or counted recontact returns to one assisted contact. It does not model repeated retries, several later contacts or total enterprise demand.

Extend the model when those assumptions fail. Keep baseline and future cohorts comparable. Negative released hours mean more work, not zero saving. The model charges this added effort at the full entered hourly cost.

Action: get a data owner to approve each measure, cohort and observation window before using a calculated benefit in outreach or a proposal.

Fictional scenario. USD. Not a quote, benchmark or forecast. Each request previously needed one assisted contact. The model assumes at most one automation attempt and one later assisted contact.

Calculate in the right order

Attempts: 120,000 × 45% = 54,000.

Durable automated completions: 54,000 × 65% × 90% = 31,590. This is not 54,000 contacts removed.

Remaining assisted contacts: 120,000 − 31,590 = 88,410.

Hours released: (120,000 × 8 − 88,410 × 8.8) ÷ 60 = 3,033.2 hours per month.

Theoretical labour value: 3,033.2 × $30 = $90,996. Only half is assumed cash-realisable: $45,498.

Added monthly cost: 54,000 × $0.18 + $12,000 = $21,720.

Net monthly cash benefit: $45,498 − $21,720 = $23,778.

Simple steady-state payback is about 6.7 months after steady state is reached, before ramp effects and excluded costs. It is not launch-to-payback.

What can break the case

If no staffing, supplier or overtime spend changes, set cash realisation to zero. The model then shows a $21,720 monthly cash cost, even though capacity is released. That capacity needs a separate use case.

Harder residual work and later recontacts weaken savings. Several teams claiming the same avoided contact overstate them. The workbook includes downside and upside assumptions to expose these dependencies, not to forecast the future.

Ask the buyer to name each input's evidence owner and the actual spending change. A spreadsheet assumption is not a saving guarantee.

Buyer value and provider profitability are separate cases. Mobilisation, training, payment delays and scope changes can consume cash before an attractive contract pays.

A fictional 12-month managed service

USD. All figures invented. The full monthly fee is $180,000. Billing ramps to 50% in month one, 75% in month two and 100% thereafter. An accepted setup deliverable earns and invoices $120,000 in month one.

Direct labour, technology and other delivery costs total $147,000 monthly from the start. Mobilisation costs $200,000 and pursuit costs $60,000, paid in month one. All invoices are collected one month later. Opening cash is zero.

Change qualification, not just the spreadsheet

Ask for the actual acceptance event, invoice basis, payment terms, payroll start, technology commitments, transition overlap and minimum volume. “Large annual contract” is not an answer.

The workbook changes fee, revenue ramp, cost ramp, setup revenue, mobilisation, pursuit and collection lag. It assumes full invoices are collected after an integer number of months, without tax, bad debt or deposits. Real contracts may need a finer model.

Reconcile AI with the contract model

If automation reduces billable seats, test the revised revenue before assuming the old margin survives. For outcome billing, define completions, retries, reopenings, exclusions and reconciliation. Separate software, implementation, service fees and pass-through charges.

Do not fund open-ended transition work. Consider paid design, earned mobilisation fees, deposits or staged milestones where appropriate. Prepayment improves liquidity; it does not automatically create earned revenue or profit.

Bid gate: finance approves contribution, funding and downside before a delivery commitment. Delivery approves the capability and start conditions. Neither a sales forecast nor a large TCV replaces those approvals.

Use account measures for account access, opportunity measures for distinct buying decisions and contact measures for communication. Never mix populations to create an attractive conversion rate.

Research-cleared accounts → reached accounts → relevant conversations → booked meetings → held meetings → sales-accepted meetings → accepted opportunities → proposals → wins → collected cash.

One account can have several workstreams. Several meetings can support one opportunity. Preserve parent, buying entity and opportunity identifiers.

Price the next commitment

Expected incremental contribution = assumed win probability × contribution if won − remaining pursuit cost.

Contribution if won should exclude pursuit cost so it is not subtracted twice. Use remaining cost for the continue/ stop decision; separately track total acquisition cost for final economics.

Fictional example: $400,000 contribution if won, a 25% win assumption and $40,000 remaining cost produce $60,000 expected incremental contribution. Break-even probability is 10%. Neither percentage estimates an actual deal without evidence.

Use a range. At 5%, the result is negative. Optimism cannot override eligibility, delivery or cash constraints. An expected positive value can still be unsuitable if the downside is unaffordable.

Report progress without pretending it is revenue

Track validated workstreams, buyer-agreed decisions, accepted opportunities, stalled reasons, specialist hours and review dates. Keep a first-touch cohort view separate from current-period activity.

Do not celebrate reply growth while proposal quality falls. Do not treat undecided enterprise opportunities after 90 days as a zero-win decided cohort. Do not call uncommitted pipeline earned revenue.

Action: keep the funnel and incremental pursuit decision in separate worksheet sections. A large historic campaign spend is neither a reason to continue nor a cost to subtract twice.

A pilot is a bounded test of a decision. It needs a buyer, comparison, acceptance rule and route to stop. Do not launch one just to show what AI can do.

NIST's AI RMF is voluntary guidance, not a product certification. Treat evaluation and governance as work to perform, not a badge in a pitch. [S1]

Use hard gates, not one blended score

Productivity does not compensate for unauthorised action, unsafe outcomes or a data breach. Agree guardrails before exposure. A high “pilot score” must not hide a hard failure.

For AI, test identity, access, incorrect information, missing systems, policy exceptions, language, context transfer and escalation. For people-led operations, test training, calibration, routing and continuity.

Compare fairly

Do not put easy cases in treatment and difficult cases in the control. Record case mix, employee experience, channel and demand conditions. Allow the recontact or downstream outcome window to mature.

A small test may expose failures without proving average uplift. Label that limit. There is no universal minimum sample size: design depends on the decision, baseline, effect size, risk and acceptable uncertainty.

Close the test

Choose scale, extend within a new limit, revise or stop. Record the next evidence and spend required. Technical success without an operating owner or buying route is not a managed-service win.

Action: agree both the technical test and the buying decision before committing specialist time. Do not let a free pilot become an indefinite, unpaid production service.

An RFI can be a genuine buying step, a market scan or a price check. Determine which without assuming bad intent. Do not spend weeks on a bespoke proposal before asking.

Unknown means clarify before related spending. An explicit hard failure means no-bid unless an authorised change resolves it. A score cannot create eligibility.

Ask through the permitted channel

What can change? What is mandatory? How will proposals be evaluated? Is a pilot allowed? What information is available equally? Which date is fixed? Which dependencies remain with the buyer? How will transition, data, subcontractors and exit be handled?

Do not request confidential competitor information or ask a sponsor to bypass procurement. Honour tender clarification and access rules.

Structure the proposal around the decision

Start with the workstream and outcome. State scope and exclusions, assumptions, alternatives, solution, transition, operating model, measures, responsibilities, price, billing units, risks, evidence and decision requested.

Separate committed work from optional expansion. Distinguish implementation, recurring service, usage and pass-through costs. Explain volume changes and the handling of failed or reopened transactions.

Avoid unlimited free consulting

Give enough detail to establish fit. Price deeper analysis when it needs substantial expert work and the buyer needs a reusable design. Agree ownership and permitted use.

A no-bid can preserve the relationship. Explain the requirement you cannot meet and the condition under which you could reconsider. Do not promise an unsafe or unprofitable scope merely to join a supplier list.

Action: the commercial owner records proceed, reduce scope, paid phase, hold or stop, together with a resource limit and review date.

This is a review checklist, not legal advice or a complete account of any jurisdiction's law. Use qualified reviewers for the actual countries, channels, data and service. Public guidance was checked on 13 October 2026; recheck before launch.

Keep evidence of the approved outreach basis, source, notices, suppression and retention. Honour refusals across relevant systems and teams. Do not move channels to evade a stop request.

Separate sales research from production data

Use public, aggregate or synthetic information early. Customer recordings, health information, financial records, credentials and restricted architecture do not belong in prospecting tools or a public AI assistant.

Before access, agree purpose, minimum data, owners, location, retention, subcontractors and controls. In relevant US healthcare arrangements, review business-associate requirements with the buyer's privacy team. [S5]

Constrain AI actions

OWASP describes prompt injection as a risk from untrusted input changing model behaviour. A written instruction alone is not sufficient application protection. [S3]

Our proposed review includes restricted permissions, validated actions, separation of untrusted content from instructions, human approval for sensitive actions, logging and a stop mechanism. Have specialists validate the design.

Be precise about assurance

State which entity, service, site, system and dates an assurance covers. Do not imply that a logo, partnership or certification applies to every delivery location. Check scope before adding a claim to outreach.

No checklist here grants permission, certifies compliance or replaces buyer approval. A restriction is a hard boundary, not an objection to overcome with a better sales script.

An award is not a working service. Confirm the dependencies that determine readiness. A sales date without a transition plan passes risk to delivery.

Average volume is not a staffing promise

Handling workload = contacts × handling minutes ÷ 60.

Indicative handling capacity per paid employee = paid hours × (1 − shrinkage) × planned handling occupancy.

Shrinkage means paid time unavailable for the modelled handling work, such as training or leave. Occupancy is the share of remaining available handling time spent on that work. Align definitions so time is not subtracted twice.

This is a capacity approximation, not a contact-centre schedule. Interval arrivals, service goals, concurrency, case mix, peaks, skills and absences require workforce planning. Do not use the arithmetic to promise a service level.

Model the ramp before the award

When do recruiting, payroll, technology and facilities begin relative to billing? Include parallel running, training and buyer approval delays. Test later approval, lower initial productivity and slower collection.

Avoid transferring unfamiliar high-risk work immediately before a peak without an approved and tested plan. Use a limited cutover where appropriate, with explicit stop and return conditions.

Sales owns an accurate handoff. Delivery owns readiness. Finance owns the approved funding plan. A calendar date cannot override an unresolved hard dependency.

Action: obtain the acceptance evidence for each work package before confirming the start. Record client-owned dependencies as clearly as provider commitments.

A live service is evidence and a responsibility, not a licence to sell the catalogue. Resolve serious service issues before asking for expansion.

Map region, business unit and journey. Mark current scope, healthy adjacent need, unconfirmed hypothesis, active opportunity, blocked or out of scope. Record the relationship owner and authorised source of insight.

Make the service review useful

Show the agreed baseline, actual results, unresolved issues and one adjacent hypothesis. Ask whether that issue belongs on the buyer's agenda. Do not turn every operating review into a sales presentation.

“We see [permitted evidence] in the work we run. Does the adjacent team see the same issue, and would you be comfortable introducing us to compare it?”

Do not share information outside its approved purpose or across units without authority. Access to a record does not make it sales material.

Keep expansion accounting clean

An extension is not a new logo. A distinct workstream can be a new opportunity within an existing account. Separate retained revenue, price change, scope expansion and new customers.

Do not count the entire existing contract again when an amendment adds a queue. Forecast only incremental scope under the relevant stage.

Use a second qualification gate

A known client still needs a new scope, owner, price, delivery plan and control review when the work changes. Existing proof can ease the conversation; it does not validate every language, system or risk boundary.

Action: bring one verified adjacent problem to the account owner. Earn permission for a small decision before involving more service lines. Expansion should follow evidence, not the size of the provider's offering.

Illustrative situations, not client stories. Both accounts are invented. The sequence shows what to verify and when to stop, not a promised conversion result.

Pursuit A: a retailer's post-purchase service

A retailer announces a new market and revised returns proposition. This supports an expansion question. It does not prove rising contacts, incumbent failure or outsourcing intent.

The account lead selects one country, language and refund-status journey. They record the announcement, date and question: “Who owns the handoff from refund portal to assisted support?”

First, check an authorised technology partner for a useful joint introduction. If unavailable, use an approved direct route to the operations owner. Do not contact six executives with identical messages.

The asset is a one-page automation boundary map. It covers authentication, the refund-status source of truth, permitted answers, human cases and evidence of completion. No customer records are needed.

Fictional buyer reply: “We have a bot already.” The next question is whether the task completes without repeat contact. If the buyer says it is solved and not under review, close the thread.

If a review is useful, invite the operations and digital journey owners. The call decides whether a baseline review is worthwhile. It does not promise bot replacement or headcount reduction.

Move to data review only after the buyer identifies a cohort, approved access and a findings owner. Use Buyer Case to test recontact, harder residual work, new costs and the specific spending change.

Expansion to another language waits for the first task's agreed quality and economics. A different refund policy or identity flow requires a new test.

A software firm launches in another region and advertises customer-success roles. Hiring suggests a coverage question, not poor retention or intent to outsource.

The initial unit is a product family, region and customer tier. The hypothesis is an uncovered renewal queue. Check account ownership and partner agreements before contact.

Offer a coverage worksheet separating onboarding, adoption, renewal preparation, negotiation and expansion. Include account assignment, permitted contact, attribution, escalation and exclusions.

Fictional buyer reply: “Our internal team handles strategic accounts.” Agree. Ask whether a permitted non- strategic segment lacks clear coverage. If all segments are covered and no change is planned, do not invent a gap.

A useful meeting pairs the customer-success or revenue leader with revenue operations. It decides whether a scoped coverage pilot fits, not whether the provider can generate impressive appointment counts.

Define the eligible cohort, baseline, exclusions, discounts and credit for prior work. Do not count a renewal already secured internally as incremental provider performance.

Expansion requires accepted attribution and customer-experience evidence. A strong renewal pilot does not prove capability to run technical support, regulated advice or every sales channel.

The shared pattern: use a verified event, isolate one workstream, invite correction, agree a next decision and earn expansion through evidence.

This is an implementation sequence, not a sales-cycle forecast. Enterprise contracts can need longer approval, security and sourcing processes. Judge progress by reliable evidence and useful decisions, including a decision to stop.

Give each role a decision

Commercial sponsor sets the outcome and loss limit. Account owner controls coverage and duplicates. Researcher records evidence and unknowns. Sales development owns approved contact, replies and scheduling.

Solution lead accepts technical discovery and limits unpaid expert time. Delivery confirms capability and mobilisation. Risk, legal and procurement approve within their authority. Finance validates contribution, cash and exposure.

One person may hold several roles. A shared inbox cannot hold accountability.

Five decisions each week

1. Which workstreams earned a buyer-agreed next step or were rejected? 2. Where does work wait: replies, scheduling, specialists, approval or buyer action? 3. Which assumption failed, and what is the smallest useful change? 4. Are cash, expert hours, client obligations or contact rules at risk? 5. What will each owner finish before the next review?

Keep background notes outside the meeting. Review recordings only through an approved consent and data- handling process.

Replies without interest may indicate a fit, message or offer problem. Held meetings without accepted work may indicate qualification or handoff failure. Opportunities without a buyer-owned next step may lack a real decision or route.

These are hypotheses. Check the actual conversations. The smallest conversion percentage is not automatically the constraint.

Scale one variable

Increase account volume, add a language, enter a vertical or add a play separately, only after permission, buyer value and receiving capacity pass review. Otherwise, the source of any change remains unclear.

Copy this brief into the account record. Keep it short enough that sales and solution teams will read it. Link to evidence rather than filling the page with background.

Start without breaking the examples

Save a working copy. Financial sheets open with fictional assumptions to make the logic visible. Replace them and their source notes with approved inputs. Blank means unknown; zero means an entered zero. A zero denominator is undefined.

Formula cells calculate; they do not validate evidence. A score, badge or positive reply is not proof of budget or permission.

The web edition includes a brief builder, meeting gate, buyer case and pursuit calculator. It processes entries in the browser without submission or automatic storage. Save work before closing. Do not enter customer records, credentials or sensitive information.

The additional workstream templates give a blank, service-specific brief for each of the eight plays.

Before the first message, ask a colleague to inspect one pursuit from evidence to meeting. They should be able to explain why the buyer should care and what happens next. Repair that record before expanding the list.

A strong resource removes one source of uncertainty. It does not need to promise a global transformation to be worth reading.

Workstream: a bounded operational process or problem with scope and an owner.

Buying entity: the organisation that can contract for the work. It may differ from the parent brand.

CX: customer experience across the journey, not just a contact-centre interaction.

BPO: contracted performance of a business process. It does not transfer every client responsibility.

Managed operations: ongoing responsibility for agreed work, measures and controls. The contract defines the boundary.

AI implementation: work required to put an AI-enabled process into an operating environment, including access, data, integration, evaluation and support.

RFI / RFP: a request for information or proposal. Neither proves a funded, winnable opportunity.

Accepted opportunity: a buyer-relevant workstream accepted by the commercial team against agreed rules. It is not every positive reply.

TCV: total contract value over a stated term. It is not current earned revenue, contribution or collected cash.

Contribution: revenue less the explicitly included costs. State whether delivery, mobilisation and pursuit are included. It is not net profit unless all relevant costs are included.

FTE: full-time equivalent capacity. A monthly average is not a service-level staffing plan.

Shrinkage: paid time unavailable for the handling work represented in the model, such as leave or training. Avoid double counting.

Occupancy: the share of available handling time used on measured workload. It is distinct from shrinkage, not a universal target.

Durable completion: a task completed under the agreed definition without counted same-issue recontact in the observation window.

Mobilisation: the work and spend needed to make a contract ready to deliver. Signing is not readiness.

The principle to keep

Make the first conversation small enough to accept and specific enough to matter.

Win permission to examine one workstream. Prove what changes. Protect the transition. Earn the right to expand.