Find the signal in the noise.Test it in real work. Then decide.

Bring a slow workflow, a decision without good evidence or an idea worth testing. You do not need a finished AI use case. If AI is not the answer, we tell you.

Talk about your problemSee the five steps

People start with a tool. We start with the question behind it.

  • We need an AI tool.

    Where does work wait while people copy between documents, inboxes and systems?

  • We need better reports.

    Which decision rests on scattered data or inconsistent analysis?

  • AI could help, but where?

    Where would it save time or improve a decision, and where would it not?

  • AI may change what clients expect.

    Which of our capabilities could become easier to copy?

Signal Check

Already looking at an AI tool? Check it on your own cases first.

We check one question against your finished work. We show where time or quality is really lost, compare your option with a simpler one and recommend one way forward. It is a fast track through steps 1 to 3 of our Signal to System method.

BuyNot yetSimpler first

The answer does not have to involve AI. Signal Check ends with that recommendation; architecture, vendor choice and pilots are Discovery work.

You bring
One decision question. For example: “Should we buy an AI tool to write our quotes?”
Your cases
3 finished cases with their emails and document versions, exported by you. We do not need access to your systems.
Two interviews
Up to 45 minutes each, with the person who owns the problem and the person who does the work.
You get
A Signal Card with one recommendation, and a review meeting of up to 60 minutes.
Price
2k Euro Credited in full to Discovery if you start it within 30 days.
See an example Signal Card

It starts with a free 30-minute call. We narrow the question and check that your cases are enough. If they are not, we tell you before you pay.

Signal to System

Five steps. Each one earns the next.

Every Jutro project follows these five steps, whichever phase you start with. Signal Check runs steps 1 to 3 for one question. Discovery and the pilot take a problem through all five. Meet the people behind the work.

  1. 1. Understand the real problem

    What is actually happening?

    We map the workflow, the decisions, the people and the systems involved, and what it costs when it goes wrong.

    You getOne shared view of the problem and who owns it.

  2. 2. Find the signal and baseline

    What does better mean, in numbers?

    We separate facts from assumptions and measure today's starting point. Any new approach has to beat it.

    You getA number to beat.

  3. 3. Choose the simplest credible move

    What is the simplest thing that could work?

    We compare better process, rules, search, analytics and AI, and pick the least complex option that can do the job. Sometimes that is not AI.

    You getOne recommendation, in proportion to the problem.

  4. 4. Test it in real work

    Does it hold up in daily work?

    A real workflow, one owner, one main measure. We agree the stop line before we start.

    You getEvidence, before you commit budget.

  5. 5. Decide what comes next

    Scale, revise or stop?

    We look at the value, whether people use it and how it can fail. Then we recommend one of three.

    You getA clear decision.

We say stop when the numbers say stop.

We write the stop line down before the test starts. In our example on the home page, approvals that still take more than 4 days after four weeks do not scale.

  1. Scale

    The system created enough value to justify a wider rollout.

  2. Revise

    The direction holds. The evidence shows what to change, or a simpler fix does the job.

  3. Stop

    The value, the evidence or the conditions do not justify going on.

Stopping weak work protects your budget and attention for a better idea.

A useful system is more than its AI.

Take the order approvals from our home page example. Four parts work together. Only the first one is AI.

  1. AI

    Reads the orders

    They arrive as emails and PDFs. In this example, AI does this one job.

  2. Rules

    Sorts the queue

    One shared queue. Simple priority rules decide what goes first.

  3. People

    Decide

    A person approves each order and stays accountable for it.

  4. Feedback

    Checks the result

    Every week we compare approval time with the 4-day stop line.

In this example, one of the four parts uses AI. The other three decide whether anyone uses it.

AI can play many roles in a system: predicting trends and supporting decisions, marking unknowns or anomalies, reading and drafting, comparing and classifying or automating a bounded step. Each role has to be justified. We do not push AI in for its own sake. We use it where it brings real value, and where it does not, we say so. When the evidence supports building, we carry the work through architecture, deployment and evaluation, and bring in specialists only where needed.

Want to see how else we use AI, and where it helps?

See our work

Bring the problem, not a polished AI brief.

Tell us what is not working or what you want to change. We will say whether the next step is Discovery, a focused test, a system or no build. Discovery starts from 7.5k Euro. After it you see the cost of the next phase and decide. Not sure about Discovery yet? Start with a Signal Check for 2k Euro.

Send us a noteSee how we start

Last updated: