Pricing
Getting a price out of an AI consultancy usually takes three calls and a discovery workshop.
Here is how the work is priced, what changes a number, and what costs nothing. You get a firm figure before you commit to anything, and nothing here is billed by the hour.
The work
Seven things you can engage us for
Grouped by how each one is priced, because that is the part that decides whether it fits your budget cycle.
AI opportunity assessment
Two to three weeks inside your company, ending in a ranked list of what is worth building and a longer list of what is not.
$15,000 to $25,000, fixed fee
What this actually involvesWhat you get
- A 12 to 20 page opportunity note, written to be read by your CFO as well as your engineers
- A baseline number and a cost model for every surviving candidate
- The second list: what not to build, with the reason for each
- A walkthrough session with whoever needs to sign off
What moves the number
- How many teams we need to sit with. One process is not three.
- Whether the processes are instrumented. If nothing is measured, measuring it comes first and is billed as its own piece.
- How fast we get access. Two weeks of waiting for a systems login is two weeks.
Technical due diligence, first memo
A technical read on one company you are about to back, inside 48 hours, with the memo stating its own gaps on the first page.
No fee, no obligation, one per fund
What this actually involvesWhat you get
- Three to six pages on whether the codebase matches the story
- Where the team carries risk it has not priced
- What the first six months of engineering will cost
- The questions worth asking the founder directly
What moves the number
- Nothing. It is free because it is the cheapest way for a fund to find out whether the memo is any good.
- Memos after the first are priced per company, and repeat funds usually move to a per-quarter arrangement.
Builds
AI features in your product, software and platform work, and the integrations around both. Scoped once the assessment says what is worth building.
From $40,000. Most land between $40,000 and $120,000
What this actually involvesWhat you get
- Working software in your repository, running in your stack
- An evaluation set of real inputs your engineers can rerun
- A runbook written for whoever inherits it
- The failure cases handled, not listed as future work
What moves the number
- Whether it is one feature or a programme. The first is weeks and the second is quarters.
- The state of the systems it has to reach. A documented API is a different job from a database nobody owns.
- Whether an evaluation set exists. Building one from scratch is real work and it is not optional.
AI transformation
The multi-team version: mapping the processes, building into systems already running, and training the people who inherit them.
From $80,000, scoped per programme
What this actually involvesWhat you get
- A process map with numbers on it
- The builds that came out of it
- A before-and-after note per process
- Training for the teams that own the result
What moves the number
- The number of processes in scope, which is the whole of it.
- Whether the honest finding is a data quality problem rather than an AI one, which happens often enough to plan for.
AI training for your team
Onsite or remote, half a day through to a four-week course, built on your tasks and your documents.
$4,000 for a half day. $30,000 for a four-week programme
What this actually involvesWhat you get
- The session, live, in your office or over video
- A recording, the exercises and the prompts your team wrote
- The same tasks tested before and after, so you get a number rather than a satisfaction score
- The material stays with you
What moves the number
- Length, and how many people. A half day for eight engineers is not a four-week course for a department.
- Onsite travel, which is billed at cost and named before you commit.
- How much of the material has to be built from your own documents rather than adapted.
Fractional CTO
A technical partner on retainer for a founder who has raised and has no CTO. Ten to twenty hours a month, starting with a written assessment inside thirty days.
$6,000 to $15,000 a month, one month notice
What this actually involvesWhat you get
- A written technical assessment inside the first thirty days
- A ninety-day plan with milestones somebody outside the company could check
- Interviews for engineering hires, with you in the room rather than instead of you
- A read on any code an agency or contractor delivers, before you sign it off
- The technical answers your next investor will ask for, prepared rather than improvised
What moves the number
- Hours. Ten a month is oversight. Twenty is oversight plus doing some of the work.
- Whether there is an engineering team yet. Managing three contractors is more hours than advising one founder.
- Whether a raise is imminent. Diligence preparation is a burst of work rather than a steady rate.
Fractional AI advisory
A technical partner for a team that needs AI judgement without hiring an AI team to get it. A recurring session plus availability between them.
$4,000 to $8,000 a month, one month notice
What this actually involvesWhat you get
- A standing session at an agreed cadence
- Availability between sessions for the decisions that will not wait
- A written record of each decision and why it was made
- One month notice either way, no annual lock-in
What moves the number
- Cadence. Fortnightly is not weekly.
- Whether you want review of work in progress or only direction.
Onboarding, retention and measurement
Product and growth engagements rather than AI ones. Where new users stall, why customers leave, and making a process measurable so anything built on it can be proved.
$20,000 to $60,000 per project
What this actually involvesWhat you get
- The metric measured before and after, against a definition you agree first
- The changes that moved it, built and shipped
- A churn breakdown with the addressable share separated out, where that is the work
What moves the number
- Whether the analytics already exist. Instrumentation first is common and it is usually the cheap part.
- How much of the fix is product work rather than a model, which is most of it more often than not.
No fee
Three things that cost nothing, and are not a trial of something that does.
The first due diligence memo
One per fund, inside 48 hours, no obligation.
GoThe unit economics tool
Whether an AI feature pays for itself at your pricing. Email for the full read, nothing else.
GoThe first call
Thirty minutes. If the honest answer is that you do not need us, you get that on the call rather than after a proposal.
Go
The limit
Where a published band stops being useful
A band tells you whether to have the conversation. It does not tell you what your project costs, and anyone who quotes you a firm number before looking at your systems is either guessing or has decided in advance what to build.
The place a number moves most is the one nobody puts on a pricing page: whether the process you want improved has ever been measured. Where it has not, the first piece of work is instrumentation, it is usually cheap, and without it nothing built afterwards can be proved to have worked. We would rather say that here than discover it in week two and send a change order.
If a band on this page does not fit your budget, say so on the call. Sometimes the answer is a smaller first piece, and sometimes it is that we are the wrong firm for this and we will tell you who is not.
FAQ
Questions about money, answered before the call
Next step
Thirty minutes, and you leave with a number and a scope rather than a proposal to wait for.