Guide · Investor or project evaluator

How to evaluate an AI project: evidence, risks and milestones

A demo shows possible behaviour. Evaluation must determine whether that behaviour solves a real problem, holds under representative conditions and can become a sustainable system.

Model, interface and narrative are only part of the evidence. Data, human intervention, infrastructure, rights, cost and responsibility determine what the project actually is.

Editorial noteDraft generated by the system on 14 August 2026. Not yet reviewed.

From context to usable evidenceFive relations to preserve
  1. ThesisWhat relationship between user, problem and outcome is claimed?
  2. EvidenceWhich data or observable behaviour supports the claim?
  3. DependenciesWhat must remain available, authorised and sustainable?
Open the mapPrepare a test

Starting question

The project thesis must survive a chain of evidence.

Who uses the system, which activity changes and what observable outcome follows? If these elements are not distinct, model accuracy, user counts or commercial interest do not establish project value.

Technical and business due diligence meet in dependencies: the data, suppliers, people, permissions and costs that make performance possible.

Operational relations

Five steps that must remain connected.

01

Thesis

What relationship between user, problem and outcome is claimed?

02

Evidence

Which data or observable behaviour supports the claim?

03

Dependencies

What must remain available, authorised and sustainable?

04

Risk

Which failure changes value, safety or the right to operate?

05

Milestone

Which uncertainty will the next proof remove?

Boundaries and responsibility

Accuracy, benchmarks and ROI only have meaning in context.

A score may depend on an unrepresentative dataset; savings may omit human checking and operating costs; growth may be supported by free trials or manual work. Every number needs its method and boundary.

Evaluation must also distinguish owned, licensed, API-accessed and hypothetical components. Current availability cannot be inferred from a roadmap.

Practical object

AI Project Evidence Canvas

Complete the AI Project Evidence Canvas. Each field exposes a relationship to verify before extending the system.

01

User

Who adopts, uses, pays or experiences the effect?

02

Problem

Which work or decision matters enough?

03

Outcome

What changes in an observable way?

04

Evidence

Which data, tests or real use supports the thesis?

05

System

Which parts are owned, external or still manual?

06

Dependencies

Which data, suppliers, people and permissions are critical?

07

Risks

Which errors, constraints or concentrations change value?

08

Milestone

Which next proof reduces the decisive uncertainty?

First test

A short test should produce knowledge, not merely an output.

  1. 01
    Reconstruct the thesis

    Write user, problem, activity and outcome without using the product name.

  2. 02
    Request primary evidence

    Inspect data, examples, protocol and failure cases rather than the summary alone.

  3. 03
    Decompose the system

    Separate model, data, software, people, services and external components.

  4. 04
    Stress one dependency

    Simulate higher cost, supplier change, poorer data or loss of a key person.

  5. 05
    Define the milestone

    Connect resource and time to removing a verifiable uncertainty.

Public sources

References for verification and further work.

These sources support initial design. Legal, professional, ethical and organisational requirements depend on the case and the responsible functions.

Frequently asked questions

Investor or project evaluator: questions to clarify before the project.

What evidence should an AI project provide?

Representative examples, evaluation method, baseline, failures, human intervention, dependencies and behaviour in context. A demo is one piece of evidence, not the full case.

How should an agentic system be evaluated?

Observe actions, tools, state, permissions, stopping and recovery—not only the quality of generated text.

Is a benchmark sufficient?

No. Its dataset, metric, comparison, representativeness and relationship to real work must be known.

How can economic benefit be verified?

State the baseline, period, included costs, human checking, infrastructure and assumptions. Without them, value remains a claim to test.

What makes a milestone useful?

It removes a decisive uncertainty through an observable result rather than merely recording activity.