Agentic Architecture Diagnostic · Paris & France

Know within two weeks whether your AI agents can reach production

An architect with twenty years behind them examines your code, your information system and your agentic chains - with AI tooling that reads the entire repository, not a sample. After two weeks you have a reasoned answer to a single question: can this project hold up in production, under what conditions, and at what cost. Fixed price, fixed scope, restitution date announced when you order.

Discuss your project See the methodology
They trust us
Fixed price, all included
€6,000

Excl. VAT. No overage, whatever we find.

Duration
2 weeks · restitution on day 10
Format
Remote or on your premises
Book the diagnostic

Reply within 24h. We tell you first whether it is relevant.

What we do

  • AI-tooled analysis of the entire repository, not a sample
  • Mapping of your IS and its integration points (APIs, data, events)
  • Review of your agentic chains: orchestration, guardrails, error recovery
  • Audit of the accesses and permissions actually granted to agents
  • AI Act compliance and decision-traceability review
  • Inference cost estimate at production volume
  • Interviews with your technical and business teams
  • Costed build: scope, effort, sequence, budget

What you keep

  • A written report, delivered before the restitution
  • The technical map of your information system
  • A costed, sequenced roadmap
  • A 90-minute restitution with your teams

What is not included

  • No code delivered - this is a diagnostic, not a build
  • No rework of your data foundation
  • No penetration testing
  • No change management

The challenge

Why most AI proofs of concept will never reach production

The POC convinces in a demo and breaks in production, almost always for the same reasons - and none of them is the model. Gartner forecasts over 40% abandonment by the end of 2027, and only 31% of companies have an agent in production today. Here is what we find every time:

  1. 01 The agent was wired to a data extract, never to the real information system
  2. 02 Nobody knows what the agent is allowed to do, or on whose behalf
  3. 03 Inference cost was never measured at production volume
  4. 04 No test set detects a regression in answer quality
  5. 05 Traceability does not survive an AI Act vendor questionnaire
  6. 06 The team that built the POC is not the team that will have to run it

Our methodology

How we deliver

Architects with twenty years behind them, and a delivery process instrumented by AI: scoping, code review, quality and security tooled at every phase. Each step produces concrete deliverables, and we stay through the run.

01
Days 1 to 2

Scoping

We settle the question the diagnostic must answer, and we obtain access. A sharp scope beats a broad audit that concludes nothing.
Deliverables
  • 2-hour scoping workshop with technical and business sponsors
  • Read access to the repository, environments and documentation
  • NDA signed before any access
  • Exit question written down and validated
02
Days 3 to 6

Tooled analysis

The AI tooling covers all the code and documentation while an architect runs the interviews. The combination is what matters: the machine covers, the human interprets.
Deliverables
  • Full reading of the repository and its dependencies
  • Mapping of the IS and its integration points
  • Review of agentic chains: orchestration, guardrails, error recovery
  • Inventory of accesses and permissions granted to agents
  • Interviews with technical and business teams
03
Days 7 to 9

Target architecture & compliance

We describe the architecture that would hold in production, the gap with what exists, and what that gap costs. The AI Act review happens here, not in an appendix.
Deliverables
  • Target architecture: agents, orchestration, guardrails, observability
  • RAG chains and MCP connectors needed into your business domains
  • Inference cost estimate at real volume
  • AI Act compliance and decision-traceability review
  • Costed build: scope, effort, sequence, budget
04
Day 10

Restitution

Ninety minutes with your teams. The report reaches you beforehand, so the session is for deciding rather than discovering. You leave with the costed build - free to hand it to whoever you want.
Deliverables
  • Written report delivered 48h before the session
  • 90-minute restitution, open questions
  • Costed, sequenced roadmap
  • A reasoned recommendation: go, defer, or drop it

Business value

What you concretely gain

Expected results

An answer, not another report

A costed build

Risk capped at €6,000

  1. 01

    An answer, not another report

    The diagnostic answers one question: does this project hold up in production. The recommendation may be to drop it - which is precisely what makes the others credible.

  2. 02

    A costed build

    You leave with a scope, an effort estimate, a sequence and a budget. Enough to decide in committee, or to consult other providers on a solid basis.

  3. 03

    Risk capped at €6,000

    Fixed price, fixed scope, announced date. You commit the price of two consulting days rather than launching a six-figure programme on a hunch.

  4. 04

    An architect, not a junior consultant

    The diagnostic is run by an architect from the network with twenty years or more behind them - the same one who would lead the build. Nothing is delegated.

Frequently asked questions

Your questions, our answers

01 Why is the diagnostic paid?
Because a free audit is not an audit, it is a sales device in disguise - and everyone knows it. Charging for it lets us spend two weeks of senior architect time on it, and above all lets us conclude "do not do this" without it costing us a sale. That independence is what you are buying.
02 Is the €6,000 deducted from the build if I hire you?
No, deliberately. A deductible diagnostic becomes a sales argument again: we would have an interest in concluding that you should build. The diagnostic is paid for because it has value of its own, whether or not you give us what follows.
03 What if you conclude we should do nothing?
That is a result, and probably the most profitable one. It will have cost you €6,000 instead of the two hundred thousand of a programme launched on a hunch. We will then tell you what has to be fixed before the question is worth asking again.
04 Do you need access to our code?
Yes, read-only, after an NDA is signed. That is what lets the tooling cover the whole repository rather than a chosen sample. If access is impossible for security reasons, we work in shared sessions on your premises - scope and price do not change.
05 Two weeks - is that not short for an audit?
It would be short for a general audit. The scope is deliberately bounded to one question, and that is what makes the timeline hold. An architect with twenty years behind them sees in ten days what a junior team takes six months to discover - and the tooling reads, in the meantime, what no human would read.
06 What happens after the restitution?
Nothing automatic. You hold the report, the map and the costing. If you give us the build, it starts on scoping already done. If you consult elsewhere, you do it with a serious specification - which is already a good reason to have run the diagnostic.

Two weeks to know

Tell us where your project stands. If the diagnostic is not relevant in your situation, we will say so before selling it to you.

Schedule a call See all our solutions