All insights
EN · AI Strategy & Transformation

How to evaluate whether an AI project timeline is realistic before hiring

Test critical path, capacity, data, dependencies and contingency with a 32-point scorecard before accepting an AI services firm's schedule.

AI delivery plan with blue milestones, linked dependencies and an amber bottleneck before the operational gate.
AI Delivery Plan Reality Test-8 tests critical path, capacity, dependencies and contingency before award. · Generated with OpenAI

A credible AI timeline ties every milestone to testable deliverables, owned dependencies, named-team capacity, data and environment access, evaluation, remediation and contingency. Before award, require a baseline schedule, dependency register, capacity plan and assumption log—then have the proposed delivery team defend the critical path. A date without that evidence is a sales target, not an executable plan.

Do not look for a universal duration. An enterprise agent connected to operational systems, an AI platform and a regulated product carry different uncertainty. The review should reveal whether mandatory work is inside the calendar and whether new evidence triggers a clear replan, pause or stop decision.

AI Delivery Plan Reality Test-8: a 32-point scorecard

Score each proof from zero to four: zero means absent; one, assertion; two, partial evidence; three, consistent evidence; four, evidence validated with the people who will deliver. As a screening rule, require at least 24 of 32, no zero and all blockers cleared. Adjust the threshold to risk rather than treating the total as an automatic award rule.

  • Outcome and acceptance — each milestone ends in a demonstrable capability, metric, environment, approver and acceptance rule.
  • Work breakdown — discovery, data, integration, evaluation, security, adoption, operations and transfer are scheduled activities.
  • Dependencies and critical path — logic is traceable, owners are named and float is visible.
  • Data and integration readiness — access, quality, permissions, APIs, environments and third parties have dates and entry criteria.
  • Team capacity — named people, allocation, concurrency, scarce skills, calendars and substitutions are compatible.
  • Evaluation and risk gates — testing, human review, security, remediation and regression fit inside the schedule.
  • Operational transition — observability, runbooks, support, training, rollback and handover are not deferred past launch.
  • Uncertainty and contingency — assumptions, reserves, scenarios, reforecasting and go/no-go decisions are explicit.

Six conditions that should block approval

  • A fixed date with no assumption or exclusion log.
  • No critical path, calculated float or dependency owner.
  • Parallel work that relies on the same specialists, data or environments.
  • Data, security, legal or integration work marked TBD outside the schedule.
  • Only a happy path, with no time for evaluation, remediation, regression and retest.
  • No reforecast cadence, change control or stop rule.

Require four comparable artifacts

Ask every finalist for a baseline schedule with activities and milestones, a dependency register with owners and need-by dates, a capacity plan by role and an assumption log. Standardize enough detail to compare bids without prescribing the solution. The team should identify the critical path and show what happens when a dependency arrives late.

The GAO Schedule Assessment Guide describes ten practices for reliable schedules: capture activities, sequence logic, assign resources, estimate durations, verify horizontal and vertical traceability, identify critical path and float, analyze schedule risk and maintain updates. Use them as diligence questions, scaled to the engagement.

Run a 90-minute plan-defense session

  • 15 minutes — the team explains outcomes, decomposition and critical path.
  • 20 minutes — the buyer traces three milestones through data, integration, evaluation and acceptance.
  • 20 minutes — test loading and concurrency across architecture, engineering, data, security and client SMEs.
  • 20 minutes — inject three events: late data, an unavailable integration and a failed evaluation gate.
  • 15 minutes — request the new forecast, cost impact, go/no-go decision and executive communication.

The people proposed for delivery—not only sales—should lead the session. Judge causal reasoning, trade-offs, assumptions, recovery options and willingness to protect quality and safety, not how quickly the team produces a new date.

Validate readiness before compressing the calendar

The GAO Technology Readiness Assessment Guide emphasizes evidence of maturity at key decisions. If data, integrations, evaluation or controls remain hypotheses, buy a short, capped discovery or paid slice first. Do not convert low readiness into an aggressive production promise.

Use assurance gates proportional to risk

GovS 002 makes assurance proportionate to risk and value. NIST AI RMF organizes risk work into Govern, Map, Measure and Manage. Turn that discipline into gates: context and accountability before architecture; baselines and data before build; evaluation and risk disposition before release; operations and ownership before handover.

Negotiate the update system, not just the end date

  • Reforecast cadence and executive reporting format.
  • Variance limits that trigger a decision or formal change.
  • Treatment of buyer and third-party dependencies.
  • Schedule reserve and authority to consume it.
  • Rules for descoping without weakening guardrails.
  • Rights to pause, remediate, retest, transition and exit.

A sound forecast changes when evidence changes; control comes from making that change explainable and governable. The UK Digital, Data and Technology Playbook helps connect delivery planning, commercial model and supplier management across the lifecycle.

Connect timeline, proposal, acceptance and strategy

Use https://makinai.co/insights/en/how-much-ai-consulting-services-cost to test cost and timing together, https://makinai.co/insights/en/how-to-evaluate-ai-consulting-proposals-scorecard to normalize finalists and https://makinai.co/insights/en/ai-project-acceptance-criteria-payment-milestones to tie milestones to evidence and payment. Explore https://makinai.co/services/en/ai-strategy-transformation-consulting.

When to involve MAKINAI

MAKINAI can challenge the delivery plan, identify the critical path, test capacity and dependencies, and turn the schedule into executive and contractual gates. The goal is not the shortest promise; it is the most decision-useful forecast that preserves value, safety and options.

Sources and references

  1. GAO — Schedule Assessment Guide · U.S. Government Accountability Office

    Defines ten practices for a reliable schedule, including activities, logic, resources, duration, critical path, risk and updates.

    2026-09-03
  2. UK Government — GovS 002 Project Delivery · UK Government Project Delivery

    Sets risk- and value-proportionate governance and assurance across project and program life cycles.

    2026-09-03
  3. UK Government — Digital, Data and Technology Playbook · UK Cabinet Office

    Guides commercial and digital delivery decisions across the lifecycle, including planning, capability and supplier management.

    2026-09-03
  4. NIST — AI RMF Core · National Institute of Standards and Technology

    Organizes AI risk management into Govern, Map, Measure and Manage functions that connect context, evaluation and response.

    2026-09-03
  5. GAO — Technology Readiness Assessment Guide · U.S. Government Accountability Office

    Assesses technology maturity using consistent evidence at key decision points.

    2026-09-03
Making connections

Continue exploring

AI Strategy & Transformation

How to define AI provider governance and performance management before hiring

Read insight
AI Strategy & Transformation

How to evaluate an AI consulting ROI business case before hiring

Read insight
AI Strategy & Transformation

Boutique AI firm, global consultancy, or systems integrator: how to choose

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