All insights
EN · AI Strategy & Transformation

How to evaluate an AI consultancy’s knowledge-transfer plan before hiring

Test whether your internal team will be able to direct, operate, govern, and improve the AI system without permanent vendor dependence.

Knowledge, documentation, and operating practice cross a transparent bridge into an independent internal AI capability.
The AI Capability Transfer Proof-8 turns training into evidence that the team can direct, operate, govern, and improve the solution. · Generated with OpenAI

A strong AI consultancy should leave more than a working solution. It should leave the buyer able to make decisions, operate controls, diagnose failures, and improve the system. Before hiring, require a transfer plan tied to real tasks, named owners, usable artifacts, and proof that internal staff can perform without the consultancy leading every step.

Training, documentation, and autonomy are different outcomes. Slides and recordings can help, but transfer is proven when internal staff can explain a decision, make a safe configuration change, run evaluations, respond to an incident, and own the backlog. Put that target state in the RFP, statement of work, and acceptance criteria.

AI Capability Transfer Proof-8: a 32-point scorecard

Score each dimension from zero to four: zero is absent; one is a promise; two is partial material; three is current evidence used by internal staff; four adds independent execution in a new scenario. As an illustrative gate, proceed at 24 of 32, with no zero and at least three for decisions, operations, risk, and continuity.

  • Decisions — business, product, model, and architecture context, options, assumptions, and trade-offs are reproducible.
  • Data — internal owners control access, quality, lineage, preparation, and permitted use.
  • Build — code, prompts, configurations, integrations, tests, and environments can be changed safely.
  • Evaluation — the team can run benchmarks, acceptance tests, red teaming, and regression.
  • Operations — runbooks cover observability, cost, quality, fallback, escalation, and incidents.
  • Governance — roles, approvals, risks, exceptions, audit evidence, and model changes have owners.
  • Evolution — backlog, debt, experiments, updates, and roadmap can continue without hidden tacit knowledge.
  • Continuity — exit, replacement, handover, and access to artifacts have been rehearsed.

Apply six kill criteria

  • The plan promises training but names no tasks internal staff must perform.
  • Critical knowledge sits with one individual at the supplier or buyer.
  • Documentation has no owner, version, validated environment, or update trigger.
  • Internal staff lack adequate access to code, prompts, data, evaluations, logs, or configurations.
  • Acceptance and payment do not depend on teach-back, operate-back, or solving a new scenario.
  • Exit work begins near contract end with no capacity reserved for transition.

A failed criterion may show that the buyer is not ready to absorb the work. Make a deliberate choice: add internal capacity, buy managed services, or limit the first engagement to discovery and a pilot. The dangerous outcome is buying autonomy in the pitch and dependency in the delivery model.

Require eight reusable artifacts

  • Architecture and product decision record with alternatives and reassessment triggers.
  • Data map covering permissions, lineage, quality, and accountable owners.
  • Repositories, component inventory, build instructions, and environment configuration.
  • Evaluation suite with baselines, results, limits, and regression routine.
  • Operations runbooks for observability, cost, fallback, and incidents.
  • Risk and control register with approvals, exceptions, and evidence.
  • Prioritized backlog, known debt, roadmap, and model-or-supplier replacement criteria.
  • Transition plan covering people, access, assets, milestones, dependencies, and acceptance.

Run a 90-minute operate-back test before best and final offer

Ask the supplier to teach an internal owner to investigate a new case: quality drops after a model update, spend jumps, or an integration fails. The internal owner must find the decision record, inspect logs and evaluations, propose a fix, explain risk, and update the runbook. Score clarity, access, traceability, and how often the supplier had to take control.

Decide what must become internal

Internalize decision rights, business context, data ownership, quality and risk criteria, access to assets, and the ability to replace the provider. External specialists may continue to accelerate engineering, evaluation, or operations. The goal is not to duplicate the consultancy; it is to keep irreversible decisions and essential operating knowledge under buyer control.

Put transfer into the RFP, SOW, and payments

Name target capabilities by role, internal availability, pairing cadence, artifacts, update rules, and autonomy tests. Reserve time on both sides. Tie milestones to buyer-led demonstrations and hold final acceptance until access, documentation, and transition are complete. Require refreshes after material system changes.

United States context

Clarify employee and contractor roles, privileged access, data restrictions, subcontractors, and sector obligations. Keep repositories, cloud accounts, evaluation assets, and operating records in buyer-controlled systems where feasible. Counsel should adapt intellectual property, confidentiality, security, employment, and transition terms to the contract and regulated environment.

Connect transfer to adjacent diligence

Use https://makinai.co/insights/en/internal-team-hire-govern-ai-services-company to assign owners, https://makinai.co/insights/en/ai-provider-governance-performance-management-before-hiring to set cadences, https://makinai.co/insights/en/how-to-assess-ai-vendor-lock-in-exit-plan to test exit, and https://makinai.co/insights/en/how-to-choose-managed-ai-services-provider when operations will remain external.

When to involve MAKINAI

MAKINAI can define the target capability state, turn transfer into testable deliverables, and structure governance, architecture, and transition around verifiable evidence. Explore https://makinai.co/services/en/ai-strategy-transformation-consulting.

Sources and references

  1. UK Government — Digital, Data and Technology Playbook · UK Government Commercial Function

    Calls for building in-house capability, embedding knowledge transfer at every level, and maintaining current documentation for transition and operations.

    2026-09-11
  2. UK Government — The Sourcing Playbook · UK Government Commercial Function

    Calls for early exit and transition planning with roles, milestones, dependencies, assets, data, and sufficient time for knowledge transfer.

    2026-09-11
  3. NIST AI RMF Playbook — Govern · National Institute of Standards and Technology

    Recommends clear responsibilities, role-specific training, proficiency, documentation, and continuing oversight of AI and third-party risk.

    2026-09-11
Making connections

Continue exploring

AI Strategy & Transformation

How to assess whether an AI consultancy is vendor-neutral or platform-aligned

Read insight
AI Strategy & Transformation

How to assess an AI services provider’s subcontractors and fourth-party dependencies

Read insight
AI Strategy & Transformation

How to evaluate an AI solution architecture proposal before hiring a provider

Read insight
Related capability

AI strategy & transformation

An AI transformation consultancy should answer four questions before recommending technology: where business value exists, which capabilities and data are required, how risk will be controlled, and who will operate the change. MAKINAI connects those answers in an executable plan with priorities, owners, metrics and scale decisions.

Explore this capability