Guide
An enterprise AI assessment shows where AI will save your teams time, with numbers: a map of the work as it is really done, quantified use cases ranked by value, and a roadmap. Two weeks is enough for a focused scope, as long as the figures come from the field and not from a workshop.
Contents
What an enterprise AI assessment is
An AI assessment answers three questions: what work is really done in the organisation, how much of it AI can take over or speed up, and where to start. It lets you decide with numbers, before you invest in a rollout.
In 2025, 18% of businesses with 10 or more employees in France (mainly market sectors) said they used at least one AI technology, and 58% of those with 250 or more employees (Insee, July 2026). More than half of the businesses using AI say a lack of expertise limits them: better to scope before buying tools.
What a good assessment should deliver
A good assessment delivers three things, not a single report. The table shows what each contains and how to test it before you sign.
On our reference engagement, an accounting firm in France (client anonymised), we count 6 agents in production, about 1,337 hours a year that can be automated, an estimated value of about €60k a year, a build cost of about €14k and an estimated payback of 4 months. All of these figures are estimates.
| Deliverable | What it contains | The test |
|---|---|---|
| A map of the real work | Tasks as they are actually done, by team, with time spent and volumes, plus how many people were interviewed out of how many are concerned. | Can you trace a line back to the people who confirmed it? |
| Quantified use cases | Per case: volume, current time, annual hours saved, implementation cost, business owner. Scored on impact, feasibility and cost; cases set aside appear too, with the reason. | Can you redo the calculation from the stated assumptions? |
| A roadmap | Ordered steps, prerequisites (data, skills), expected gains per step, support for the teams, a first use case ready to launch. | Does every line have an owner and a decision date? |
How it runs in two weeks
The SHIFTLABS Pilot runs in two weeks on a focused scope (a function or a department, for instance).
AI agents handle the volume: interviews, grouping the answers, first calculations. Domain experts handle the judgement: they correct, add context and validate every conclusion. Nothing is delivered without their sign-off. As an option, usage logs from your tools (with your IT department) cross-check what teams report.
| When | What happens | What you receive |
|---|---|---|
| Scoping | A 30-minute call to set the scope. From day 1, the teams know why they are being interviewed and how their answers are used. | An agreed scope |
| Days 1 to 3 | Interviews run by an AI agent that introduces itself as one: real tasks, time spent, volumes. About 30 to 45 minutes per person. | The map of the real work |
| Days 4 and 5 | AI maturity measured across six dimensions, against a documented scale. | A score from 1 to 4, ready for the executive team |
| Days 6 to 8 | Use cases scored on impact, feasibility and cost, then ranked. | Quantified, ranked use cases |
| Week 2 | Expert review and sign-off, scoping of the first use case, readout to the executive team. | The roadmap and the business case |
A working assessment or a frozen report: how to tell
A consulting report can be excellent. It becomes a problem when nobody can check it or update it.
What happens after delivery matters as much as the rest: volumes, tools and teams change, and the figures have to follow.
| What to check | Frozen report | Usable assessment |
|---|---|---|
| Where findings come from | A few workshops with management | The teams, task by task, plus measurements where they exist |
| The figures | Percentage gains, no stated assumptions | Volume × time × cost, visible assumptions, a calculation you can redo |
| The ranking | A list of ideas with “high potential” | Scored cases, including the ones set aside |
| Validation | Read by its own author | Read and signed off by a domain expert |
| After delivery | A PDF dated the day it was handed over | A living document, with tracking of gains and adoption |
Questions to ask a provider
Before you sign, ask these nine questions. A serious provider answers in writing and shows an anonymised example.
- Who talks to our teams, and how many people do you interview out of how many are concerned?
- Where does each figure come from: reported, measured or estimated? Show me one line of the calculation.
- Who reviews and signs off the conclusions, and what do they know about my business?
- Which deliverables exactly, in what form? Can I see an anonymised example?
- How do you score use cases, and what do you do with the ones you set aside?
- Are results given by task and process, or by person?
- What do you plan for adoption: communication to the teams, managers, use of the time freed?
- Is the assessment tied to buying what comes next? Can I stop afterwards?
- In which cases will you tell me AI adds nothing?
Formats, timing and budget
The right format depends on the question: where do we stand, or where should we invest? The Diag Data IA terms were taken from Bpifrance’s website on 15 October 2026 and may change.
For what comes next, SHIFTLABS offers Platform + Support, from €3.5k excl. VAT per month, annual commitment; the Pilot does not require it.
| Format | Duration | What you get |
|---|---|---|
| SHIFTLABS express diagnostic | 3 minutes, 12 questions | An indicative maturity reading across six dimensions. Nothing is sent. |
| Diag Data IA (Bpifrance) | 8 days of expert time, over 3 months at most | Review of the current situation, concrete use cases, prioritisation and a roadmap, by a Bpifrance-approved expert. Quote capped at €10,000 excl. VAT, 40% subsidised: €6,000 excl. VAT left to pay. SMEs and mid-sized companies (ETI) with 10 to 2,000 FTEs and at least €1 million in turnover. |
| SHIFTLABS Pilot | 2 weeks | Map of the real work for one scope, 3 to 5 quantified use cases, a first case ready to launch, readout and business case. Fixed fee from €12k excl. VAT, no commitment to continue. |
Frequently asked questions
How long does an AI assessment take, and what must we commit?
Two weeks for the SHIFTLABS Pilot, on a focused scope. On your side, plan for a sponsor on the executive team, a contact person and 30 to 45 minutes per person interviewed.
How much does an AI assessment cost?
It depends on the scope (teams, sites, functions): compare like for like. At SHIFTLABS, the Pilot starts at €12k excl. VAT, as a fixed fee, for two weeks, with no commitment to continue. Bpifrance subsidises 40% of Diag Data IA for eligible SMEs and mid-sized companies: check your eligibility.
AI assessment, AI audit, maturity test: what is the difference?
The words cover very different services. A maturity test places the organisation on a scale. An audit checks practices against a rule or a standard. An assessment, as used on this page, looks for where AI is useful, with numbers, and ends in a roadmap. Judge by the deliverables.
Will employees feel evaluated? Is it a tool to cut jobs?
The aim is to remove tedious, repetitive tasks, not people. At SHIFTLABS, the interview agent introduces itself as an AI, consent is individual, and results are reported by task and by process, never by person. Tell your teams before the first interview.
What happens after the assessment?
You leave with a roadmap and a first use case ready to launch. The Pilot carries no commitment to continue: you decide once you have seen the numbers. If you carry on, an online cockpit tracks gains and adoption (Platform + Support offer).
Sources
Further reading
- Express diagnosticAssess your AI maturity in 3 minutes.
- ContactBook a 30-minute call with SHIFTLABS.
- Example interviewAn AI interview replayed step by step.
- Gain simulatorEstimate the gain from a first AI agent using your own assumptions.
And in your organisation?
In two weeks, AI agents interview your teams and domain experts validate every conclusion: you leave with costed use cases and a roadmap.