Synaptwin

Synaptwin · AI pilots into daily work

Which AI pilot is worth it?You know in three weeks.

For COOs and CFOs of European enterprises: we name the cost your AI must take out and who owns it, and you end with a go or stop your CFO signs. Fixed price once scoped: €15,000–€25,000.

  • Pilot theatre

    Pilots run, demos impress, and none of it reaches the daily work. There is no gate that says ready, and no rule that says stop.

  • Security and legal say no late

    The CISO, data protection or legal see the pilot at the end, and block it. Their conditions were never part of the plan.

  • Value nobody signed

    The CFO sees a demo, not a business case. Nobody agreed what the pilot must be worth, so nobody can say it worked.

In plain words

What we do, for whom, and how it works

What do you do?
We find out which of your AI pilots is worth taking live, and take it there.
Before anything is built, we pin down three things: which cost the AI must take out, who owns the decision, and what your CFO signs. We don’t build the model; we make sure it pays and is governed.
For whom?
COOs, CFOs and heads of transformation at European enterprises.
Your company runs AI pilots, or pays for AI that came with its software, and nobody can yet say what it delivers, who owns it, or where it stands under the EU AI Act.
Which problem do you solve?
AI works in the demo and changes nothing on Monday.
The pilot runs, the slides look good, the cost stays where it was. Because nobody asked the hard questions before the build: which cost falls, who owns the decision, what the CFO signs.

How does it work?

A free call, then three paid steps. Each one ends in something a person on your side signs, not in a slide.

  1. Free · half an hour

    The call

    We hold your AI case against the three questions.

    You leave with: which of the three you can answer today, and which you can’t.

  2. 3 weeks · €15,000–€25,000

    The Scan

    We decide with you which pilot is worth taking live, and what stands in its way.

    Ends in: a go or stop signed by your sponsor, your CFO and the control owners.

  3. 90 days · priced on what the Scan finds

    The Last Mile

    One pilot into the daily work, with the operating model around it, measured against the floor your CFO signed.

    Ends in: a measured result and a signed decision to scale, fix or stop.

  4. Monthly · priced on scope

    Assurance

    Our platform keeps it governed and proven: the inventory, the registers, a reassessment every quarter.

    Gives you: the decision record, kept current for your auditor.

The Scan, week by week: 3 weeks, €15,000–€25,000, fixed once quoted

  1. Week one: agree and watch

    Your sponsor and CFO sign what has to be true for this to be worth it. We sit with the team, watch the work as it is really done, and list the AI already in use, approved or not.

  2. Week two: score and name the gaps

    Readiness across ten dimensions, on evidence. Every gap gets an owner and a test for when it is closed.

  3. Week three: the case and the decision

    Every use case with its owner and value target; every pilot with the cost it takes out, its risks and its EU AI Act classification. The readout ends in a go or stop.

Stop is a result. If a pilot should not go live, you hear it in week three, before the budget is spent on it.

What you hold at the end of the Scan

  • Platform access for one organisation
  • AI system and use-case inventory — including what nobody approved
  • EU AI Act classification per use case — an assessment, not legal advice
  • Readiness across ten dimensions, with the conditions that block scale
  • The blockers between each pilot and production: owner, data, controls, adoption
  • A business case for the pilot worth taking live, with a value floor
  • A named owner and a signed go or stop
  • Executive report the sponsor signs
Or check in three questions where your AI is stuck →

Half an hour with one of the two founders. Bring one AI case; nothing to prepare.

Try it on your own company

What do your own AI job postings tell the EU?

Type your company's name. We read its public job board, find the AI roles and show which EU rules their own sentences point to. You get a private page that is not listed and expires on its own.

Fixed rules read only what your company publishes on its job board; no AI model writes the page. We ask for no name and no e-mail. Jobs not on a public board? The questions below work for any company.

Where is your AI stuck?

Three questions, an answer straight away

Where is your AI today?
Do the people it was built for use it every week?
Is one named person accountable for its results?

Runs in your browser — nothing is sent or stored.

Answer the three questions and your next step appears here.

For COOs, CDOs and heads of transformation whose AI is stuck in pilots.

The AI Act already applies to every company that uses AI in the EU. See what it means for yours.

About us →For partners →

If you use AI in the EU, the AI Act already applies to you

Since 2 February 2025, any company using AI in the EU has two duties in force: no prohibited practices, and AI literacy for the people who use it. What else applies depends on the class of each use case, and each use case needs a named owner to answer that.

The dates a company using AI is on

  1. February 2, 2025Prohibited practices apply, and the AI literacy duty (since the omnibus: measures to support its development)in force
  2. August 2, 2026Article 50 transparency duties applyin force
  3. December 2, 2027Annex III stand-alone high-risk systems386 days · moved from August 2, 2026
  4. August 2, 2028High-risk systems embedded in regulated products630 days · moved from August 2, 2027

In financial services, credit scoring of people and risk assessment and pricing in life and health insurance are named high-risk use cases (Regulation (EU) 2024/1689, Annex III, point 5).

Dates: Regulation (EU) 2024/1689, as amended by Regulation (EU) 2026/1744 — Digital Omnibus on AI, in force since July 27, 2026. A statutory calendar, not legal advice.

What the Scan does about it

  • Finds every AI system you run, including the ones nobody approved.
  • Classifies each use case against the Act. The result is preliminary; your counsel reviews it and signs the final call.
  • Turns each obligation into an owner, a control and the evidence that proves it.
  • Assurance keeps the register and the evidence current as the dates and the rules move.

See it before any call, free, no sign-in

We are not lawyers, and the engine gives no legal advice. It gives you the register, a classification to review and the evidence. Your counsel decides.

Your AI works in the demo. It stalls before the daily work.

Five problems, each with the step that fixes it.

  1. Pilot theatre

    Pilots run, demos impress, and none of it reaches the daily work. There is no gate that says ready, and no rule that says stop.

    Fixed in the Last Mile
  2. AI nobody approved

    Ask which AI systems run today and who approved them, and nobody can answer for the board.

    Fixed in the Scan
  3. Security and legal say no late

    The CISO, data protection or legal see the pilot at the end, and block it. Their conditions were never part of the plan.

    Fixed in the Scan
  4. Value nobody signed

    The CFO sees a demo, not a business case. Nobody agreed what the pilot must be worth, so nobody can say it worked.

    Fixed in the Scan
  5. The AI Act, unanswered

    Nobody has classified which use cases fall under the EU AI Act, or kept the evidence a regulator would ask for.

    Kept in Assurance

Each one is fixed by a decision someone signs, at a step you can see below. That is the whole method.

From the first check to AI that runs in the work

Every step below ends in something you can open later: an answer, a signature, a record. You can stop after any of them.

  1. Before you commit

    No cost until you sign the Scan

    1. You check where your AI is stuck

      Three questions on this page, or ten on the full self-assessment. Your answers stay in your browser.

      You get: your own answer, before any call.

      Take the three questions
    2. A free call

      Five questions: which pilot, who can scale it, who can stop it, by when, and whether you would stop a pilot that does not earn its keep. We show the engine on a labelled demo.

      You get: a fit, or a plain no and the reason.

    3. You sign the Scan online

      Scope, fixed price and terms on a private page. You sign it there; the invoice follows by rule.

      You get: scope and price in writing, signed.

      What the Scan includes
    1. The Pilot-to-Production Scan

      3 weeks · €15,000–€25,000

      AlignDiscoverDiagnosePrioritize

      We find every AI system you run, including what nobody approved, see what blocks each pilot, and pick the one worth taking live.

      • Your AI inventory and EU AI Act classification
      • What stands between each pilot and production
      • A business case with a value floor and a named owner

      Ends in: a business case your CFO signs — go or stop.

    1. The Last Mile

      90 days · priced on what the Scan finds

      DesignProveScale

      We work inside your team to take that one pilot into the daily work, with the operating model built around it.

      • Owners, decision rights and governance
      • A test bar your business owner agrees
      • Use measured week by week

      Ends in: scale or stop, signed — measured against the floor.

    1. Assurance

      Monthly · after the handover

      After the handover, the record keeps proving that it still works.

      • Seven health checks, none averaged away
      • A reassessment against your own baseline
      • Evidence your auditor can open

      Ends in: proof every quarter — or a named problem.

You can stop after any step. And to be plain: the Scan is new — our first clients are its first cases. What each step includes →

For AI that exists — and hasn’t reached the work yet

A fit if

  • You run AI pilots, and few or none of them are used in the daily work.
  • You have a reason to move now: a board mandate, a budget round, an audit or the AI Act timeline.
  • Your CISO, legal or compliance lead can say no — and you want them in the room from the start.
  • You would rather stop a pilot than scale one that doesn’t pay.

Not a fit if

  • You want a chatbot built quickly, or a licence to buy.
  • You have no pilots yet. The three questions above will say so, and point you to where to start.
  • You need a result promised before anyone has looked.
  • Nobody on your side will put their name to a decision.

Our first focus is financial services, where the AI Act and an auditor are already in the room. Outside it, the method is the same; the industry playbook is newer, and we’ll say so on the call.

Every step proven, not presented

  • Adoption, not pilots

    Success is AI used in the daily work, measured by adoption and realised value — not by how many pilots were launched.

    How we work →
  • Proof built in

    Every number comes from a rule or a source you can follow. What isn’t known is shown as unknown. Every decision is signed by its owner.

    See how it’s checked →
  • Built, not just advised

    We have shipped what AI needs before people trust it to act: identity, memory and voice — Riaan’s own products.

    What we have built →
  • The same rules for our own marketing

    What we publish passes the checks we ask of clients: sourced figures, no buzzwords, no bait, one honest limit in every piece.

    Our rules →

The model is the smallest part of the work

An AI programme is mostly people and process. That is where pilots stall — and where we work.

10 %
20 %
70 %
  • 10 % · AlgorithmsChoosing, prompting or tuning the model.
  • 20 % · Technology and dataPipelines, integrations and the data the model needs.
  • 70 % · People and processesRedesigned workflows, decision rights, training and adoption.

Source: Boston Consulting Group (BCG), the 10-20-70 principle.

Safe to scale, because it is checked

  • Every decision is signed by the person who owns it.
  • Every AI use case is risk-classified before it goes live.
  • Every number can be traced to where it came from — what isn’t known is shown as unknown.

1559 automated checks passed on 2026-11-11. · See how it’s built

Straight answers

What do you do differently?
We get AI out of pilots and into daily work. Every step is proven rather than presented: numbers from rules or sources, decisions signed by their owners, unknowns shown as unknown.
Is this another AI tool?
No. It’s a way of getting the AI you already have into daily work. The software you see here makes the method visible; you don’t buy a licence.
How quickly will we see something?
The Scan takes 3 weeks. At the end you know where you stand and which use case deserves a pilot.
What does it cost to start?
The Scan costs €15,000–€25,000, set by the scope and fixed once quoted. After it, you decide whether to go further.
Do we need clean data or new software first?
No. We start with the tools and data you have. Where the data isn’t ready, that becomes a gap with an owner and a date — not a reason to wait.
What about the EU AI Act?
Every use case is classified before it goes live, and the record is ready for your counsel. It isn’t legal advice — your lawyers decide.
Who does the work?
We do, with your teams. Most phases are supported by the engine; bringing people along is hands-on work.

Start with the Scan

€15,000–€25,000, fixed once scoped, 3 weeks — and you know which pilot is worth taking live, and what it is worth.