AI audit for your business: where AI actually pays off
I examine your processes and your data to find where AI genuinely saves you time or money, and where it isn't worth it. For SME and mid-market leaders who want to decide on numbers, not hype.
Free first exchange, no commitment, reply within 24 hours.
AI Audit
4days
One square = one day
- Who it's for
- You're unsure where to start, or you have doubts about a project already underway.
- What you get
- Your processes put under the microscope
- Use cases prioritized by ROI
- Pitfalls and costs flagged early
- Deliverable
- A clear report and a costed roadmap.
What an AI audit is, and what it isn't
An AI audit, the way I run it, isn't a trends report or a catalogue of fashionable tools. It's a concrete look at your business to answer one question: where would AI save you time or money, and where isn't it worth the trouble.
I look at your real processes, not an idealized version. I talk to the people who do the work, spot the repetitive, low-judgment tasks, and put every idea through three filters: does the data exist, is the gain measurable, is the risk manageable. Whatever doesn't pass, I tell you.
This page covers scoping: where to start, what to build, in what order. If your AI is already running and the real issue is the bill, LLM cost optimization has its own page.
What you get
At the end you get a clear, actionable document, not a deck that gathers dust. Concretely, it contains:
- Quantified use cases: for each one, the expected gain, the effort to build it, and a ballpark return on investment.
- Traps flagged early: the places things usually go wrong (incomplete data, runaway costs, compliance, team adoption), caught before you invest.
- A prioritized roadmap: what to launch first, what to wait on, what to drop. Ordered by payoff versus effort, not by hype.
- An honest budget and timeline estimate for the cases worth keeping.
What that report looks like, in full
Describing a deliverable is worth little: you can't judge a sentence. So I publish one in full, on a case fabricated from end to end, with the five candidate uses examined, the verdicts argued, the arithmetic shown and the technical annexes.
Read it before you call me and decide whether that level of detail suits you: example AI audit report. On that case two of the five are dropped and a third deferred, which also answers the question of whether I know how to say no.
How it works, and how long it takes
The first exchange is free, with no commitment, and I reply within 24 hours. We use it to scope the perimeter: your pain points, your tools, what you expect.
Then the audit itself is short. I give you a first readout within 48 hours: the two or three most promising leads, so you have something to work with right away. The full report follows within a few days to two weeks, depending on the size of the perimeter and the state of your data.
I work solo, so you have one contact from the first exchange to the last line of the report. No handoffs, no junior discovering your case halfway through.
Your data: step 0 before any code
Before writing a line of code, I look at your data where it lives. It's the step most projects skip, and the one that sinks them.
Data scattered across ten tools, incomplete or never updated, kills the best use case. I check what's usable right away, what needs cleaning up or gathering first, and the shortest path to get there, without a pointless data overhaul.
You don't build on sand. And at every step, your data stays with you.
What if the audit finds AI is useless for you
It happens, and I'll tell you. Sometimes the best advice is to automate nothing: a well-built spreadsheet, a reworked process or a tool you already own does the job for less and with no risk.
An honest no after a few days of audit beats a letdown after months of building. I'd rather lose a project than sell you something that won't hold. It's also why you can trust the recommendations I do keep.
No commitment to what comes next: the report is yours
The audit stands on its own. The report and the roadmap are yours: you can build with me, in-house, or with someone else. Nothing ties you to what comes next.
Everything is documented so any team can pick it up: the target architecture, model choices, cost assumptions. No dependency on me, no black box, no lock-in.
What I carry
You are handing a budget to an independent you don't know. That is a risk, and it isn't reasonable to ask you to carry it alone. Here is the share I take on, written into the proposal rather than promised on request.
- The price is firm on the proposal. On a fixed price, the overrun is on me, not on you.
- A scope change goes through an amendment signed before the spend, never a surprise invoice at the end.
- NDA signed before the audit, from the first conversation if you want one.
- The report, the roadmap, the code and the documentation are yours.
- No dependency on me: standard architecture, open standards, handover included in every build.
- If the audit concludes AI brings you nothing, I say so, and we stop there.
Public funding and schemes
Depending on your profile, an audit or a data/AI diagnostic may fit within a French public support scheme, such as the Diagnostic Data IA or Bpifrance grants. I don't promise any eligibility: the criteria change and depend on your situation.
I can point you to what exists and scope the audit so it qualifies where possible. You check your own eligibility with the relevant body.
Straight answers
What people ask me about this.
How long does an AI audit take?
A first readout within 48 hours, with the most promising leads. The full report follows within a few days to two weeks, depending on the size of the perimeter and the state of your data. The scoping call itself takes thirty minutes and costs nothing.
How much does an audit cost?
It depends on scope: an audit focused on one process is nothing like a full company-wide review. I give you a clear number from the first exchange, before any commitment. I don't post a blind price tag, because it would be wrong for your case.
What's in the deliverable exactly?
An inventory of your processes, use cases quantified by return on investment, traps flagged early (data, costs, compliance, adoption) and a prioritized roadmap with a budget and timeline estimate. A clear document, readable by leadership and engineers alike, and it's yours.
Does my data need to be clean already?
No, and it rarely is. The state of your data is part of the audit: I tell you what's usable now, what needs gathering or cleaning first, and where to start. Imperfect data isn't an obstacle to the audit, it's one of its subjects.
What happens after the audit?
You decide. The report is yours and commits you to nothing: you build with me, in-house, or elsewhere. If we continue together, the audit doubles as a spec; if not, any team can still use it.
How is this different from LLM cost optimization?
The scoping audit answers "where do I start and what do I build". Cost optimization is for an AI that's already running and whose bill is spiraling: that's a different subject, covered on its own page. The two can combine, but you rarely start with both at once.
Contact
Ready to go from demo to production?
Reply within 24 hours · first conversation free, no strings attached.