AI consulting services that end in a plan you can build

For leadership teams deciding where AI is worth investing. Our AI consultants are engineers: they assess your ideas against your real processes, data and systems, and hand over a prioritised plan with a first project defined well enough to start.

Book a first conversationSee the deliverables
format
Timeboxed assessment
evidence
Tests on your data
result
A plan you can build
A use-case map from an AI assessment: candidates placed by business value and feasibility. Invoice intake and support answers sit in 'start here'; contract review needs more planning; meeting notes is a quick improvement; a voice bot is 'not now'. The roadmap below sequences a pilot, a rollout and later candidates.
Example · fictional data

01 / When to start here

When AI consulting is the right first step

If you already know the process and the outcome you want, you may not need consulting - go straight to a scoped project such as AI workflow automation or an AI agent. Consulting earns its cost when the questions below are still open.

Definition · AI consulting

A scoped advisory engagement that identifies where AI can improve your business, tests feasibility against your data and systems, compares build, buy and extend options, and ends in a prioritised implementation plan.

  1. 01

    You have a long list of AI ideas and no ranking

    Ideas from teams, vendors and the board compete for attention, with no shared way to compare them.

  2. 02

    Pilots happened, but nothing reached production

    Demos looked promising and then stalled on data, integration, risk or ownership.

  3. 03

    You are unsure your data is ready

    Information lives in documents, inboxes and several systems of varying quality.

  4. 04

    Vendors are pitching products

    You need an independent view on whether to buy, extend what you have, or build.

  5. 05

    Leadership wants a plan with numbers attached

    Effort, risks, dependencies and expected benefits need to be written down before a budget is approved.

02 / Scope

Questions an AI consulting engagement answers

Our artificial intelligence consulting services are organised around your decisions, not around a technology catalogue.

  1. Q1

    Where does AI pay off?

    Which processes and products would benefit, and by how much effort or time, based on how the work is done today.

  2. Q2

    Is our data ready?

    What data each idea needs, where it lives, its quality, and what must happen before a build.

  3. Q3

    Build, buy or extend?

    Whether a product you can buy, a feature of a system you own, or custom software is the better path.

  4. Q4

    Which models and deployment?

    Hosted, in your cloud or self-hosted - judged against your data rules, cost and quality needs.

  5. Q5

    What governance do we need?

    Rules for data use, human review, staff use of AI tools, and how decisions are recorded.

  6. Q6

    How will we know it works?

    Measures and baselines for each candidate, so a pilot ends in evidence rather than opinion.

03 / How opportunities are assessed

Every idea scored on the same five criteria

Scores come from interviews, sample data and small feasibility tests - not from enthusiasm in the room.

  • Plan carefully

    High value, harder to deliver: needs data work, integration or risk controls first.

    • Contract clause review
    • Demand forecasting

  • Start here

    High value and feasible with current data and systems: candidates for the first pilot.

    • Supplier invoice intake
    • Support answers from help content

  • Not now

    Low value or low feasibility today. Revisit when conditions change.

    • Voice ordering bot

  • Quick improvements

    Feasible but modest value; often solved with an existing tool rather than a build.

    • Meeting notes
    • Email tone check
Illustrative example with fictional use cases: a two-by-two matrix of business value against feasibility, used to rank AI use cases. Placement comes from the scoring criteria in the table below.
Example · fictional data
CriterionWhat we look atEvidence used
ValueTime, errors, revenue or service quality affected; how often the task happensVolumes, interviews, current costs
FeasibilityWhether current models handle the task reliably enoughSmall tests on sample cases
Data readinessAvailability, quality and access to the data neededSamples, system review
RiskConsequences of errors, regulatory exposure, data sensitivityProcess review with owners
EffortIntegrations, change management, ongoing operationArchitecture sketch, estimate ranges

04 / Deliverables

What you have at the end

Written so that any competent team - yours, ours or another vendor's - can act on them.

  • Opportunity map

    All candidate use cases, scored and placed by value and feasibility, with the reasoning.

  • Data readiness notes

    What each priority use case needs, where it is, its quality and the work required.

  • Architecture sketch

    How the first projects would fit your systems, including deployment and model options.

  • Pilot scope & evaluation plan

    The first project defined: users, data, success measures, test set and limits.

  • Implementation roadmap

    Sequenced projects with dependencies, effort ranges and decision points.

  • Governance checklist

    Data use, human review, staff AI tool rules and record keeping, sized to your organisation.

05 / Self-check

AI readiness checklist

A quick self-check. Each "no" is not a blocker - it is something the assessment will plan for.

  • We can name the processes where people spend the most time on repetitive reading or typing.
  • Someone owns each of those processes and can describe the exceptions.
  • The data involved can be accessed through an API, export or shared storage.
  • We know which data must not leave our environment.
  • We have rules, or intend to write them, for staff use of AI tools.
  • We can measure today's time, cost or error rate for the candidate processes.
  • There is a budget holder for a first pilot.
  • Someone will own the system after launch.

06 / When generative AI is in scope

Generative AI consulting

Generative AI raises its own questions, and we cover them explicitly when it is in scope:

  • Which tasks suit generative models, and which need deterministic software
  • How outputs will be grounded in your content and checked
  • Which models and deployments meet your data and cost constraints
  • How staff use of general AI assistants should be governed
  • How to measure quality before and after launch

For building generative applications, see generative AI development.

07 / Engagement format

A timeboxed assessment with a clear finish

The scope and duration are agreed before starting, and the engagement ends with a decision meeting - not an open-ended retainer.

  1. stage

    Kick-off & interviews

    Goals, constraints and the people who run the candidate processes.

    out →Candidate list

  2. stage

    Assessment

    Data samples, system review and small feasibility tests on sample cases.

    out →Scored opportunities

  3. stage

    Options & priorities

    Build, buy or extend compared for the top candidates; risks and effort ranges.

    out →Ranked shortlist

  4. stage

    Plan & handover

    Roadmap, pilot scope and governance checklist presented to decision makers.

    out →Implementation plan

  5. stage

    Optional: first build

    If you choose, the pilot starts with the evaluation plan already written.

    out →Your decision

Where AI consulting ends and other help begins

09 / Choosing a partner

How to choose an AI consulting company

Artificial intelligence consulting firms range from strategy houses to tool resellers. These questions help you find an AI consulting company whose plan will survive contact with your systems.

Have the consultants built AI systems themselves?
An AI strategy consultant who has never shipped a model to production tends to underestimate data work, integration and running costs.
What exactly will we receive, and when?
Ask for the list of deliverables and the end date before signing. Open-ended advisory hours rarely end in a decision.
Will the plan be usable by another team?
Deliverables should stand on their own, so you are free to build with your own engineers or another vendor.
Do they test ideas on our data?
An AI implementation consultant should run small feasibility tests on real examples rather than score ideas from slides.
Do they sell a product they will recommend?
Independent advice compares build, buy and extend. Generative AI consulting tied to one vendor's platform is a sales process.

10 / Questions

Frequently asked questions

What does an AI consultant actually do?

An AI consultant helps you decide where AI is worth using and how to do it. In practice that means interviewing the people who run the processes, looking at the data and systems involved, testing ideas on real examples, comparing build, buy and extend options, and turning the result into a prioritised plan with a scoped first project. Our consultants are engineers who build AI systems, so the plan is grounded in what implementation will really take.

How much do AI consulting services cost?

An assessment is a fixed, scoped engagement agreed in advance, sized by the number of processes and teams involved. We agree the scope and the deliverables before starting, so there are no open-ended consulting hours. We do not publish a price list because the scope differs widely between a single department and a company-wide review.

How long does an AI assessment take?

It is timeboxed and agreed at kick-off, and short enough to keep momentum - it is meant to lead to a decision, not to run for months. The duration depends mainly on how many processes are assessed and how quickly people and sample data are available.

Do we need clean data before talking to an AI consultant?

No. Understanding what data exists, where it lives and how usable it is, is part of the assessment. Many useful AI applications work with documents, emails and records you already have; the assessment tells you which ideas depend on data work first.

Will you recommend building something with you?

Only if that is the right answer. The plan compares building, buying a product and extending systems you already own, and the deliverables are written so any competent team can carry them out. Some recommendations will be "not now" or "use an existing tool".

Do you offer generative AI consulting?

Yes. Generative AI consulting covers which tasks suit generative models, how to ground them in your content, which model and deployment fit your data rules, how to measure quality, and what governance staff use of AI tools needs.

What happens after the consulting engagement?

You keep the deliverables and decide. Options include carrying out the first project with Comtau, with your own team, or with another vendor. For ongoing technical leadership rather than a one-off plan, Comtau also offers fractional CTO support.

Start node · your brief

Have a list of AI ideas and no clear first step?

Tell us about the ideas on the table, the processes they touch and the decision you need to make. We'll propose a scope for the assessment - or tell you if you can skip it.

Book a first conversation