Decision intelligence layer

One door for every signal. One question for the person who decides.

The same problem reaches your company in four places at once: a change in SAP, an alert in o9, a supplier's email, a planner's WhatsApp. Aidecis reads them all, sees that they belong together, and gives the person with the authority one clear question.

The layer connects your systems. The intelligence inside it is S-CLAIM, the Aidecis decision model. Your systems keep doing what they do.

See the layer   Meet S-CLAIM

The problem

Four channels, four people, one problem.

Each system sees only its own signals. SAP knows the delivery moved, o9 knows the plan breaks, the buyer has the supplier's email and the planner knows the line will stop. Nobody sees the whole case, so the decision is late, made twice, or made by someone without the authority.

  • Signals get lost

    An email or a chat message never reaches the person who decides, or reaches them after the deadline.

  • Decisions are made twice

    Two people act on the same problem from different signals, with different facts.

  • Authority is unclear

    Whoever sees the message first decides, not the person whose mandate covers it.

  • Nothing is traceable

    Afterwards nobody can show which facts were known, who decided and why.

The layer

From four signals to one decision, in minutes.

Every source comes in through the same door. Aidecis reads each signal, joins the ones about the same object, routes the case by your own authority rules, and asks one person one question.

Your channels
SAP · 07:12Supplier confirmation: PO 4500077103/00010 moves from 20 to 27 November.
Email · 07:40DeltaRevin: "We can deliver 8,000 kg on 27 November at €7.90/kg."
o9 · 08:05Supply shortfall: PE-410 at Plant 2 below safety stock from 25 November.
WhatsApp · 08:20Planner: "Line 2 stops Thursday without PE-410."

And your own agents, MES, TMS or EHS: one small adapter each.

One door
Signal intakeEach source's message becomes one standard signal: who, when, what, and which order or shipment it is about.
S-CLAIM reads itWhich decision this is, what it protects first, how urgent it is. Its four modules
One case
PO 4500077103/00010 Joined on the same purchase order line, by code
  • SAP: delivery 7 days later
  • Supplier: 8,000 kg on 27 Nov
  • o9: below safety stock from 25 Nov
  • Planner: Line 2 stops Thursday
Within €50,000: the procurement manager decides.
One decision
For Anna, procurement manager Accept 27 November and bridge Plant 2 with 2,000 kg from Rotterdam? Approved at 08:34. Recorded with every signal, the rule and the reason.

Illustrative example with fictional names and figures.

Four signals arrive in four places.

Swipe sideways to follow the flow.

  1. Every source, one doorSAP, o9, email, WhatsApp or your own agents. A new source is a small adapter, not a new project.
  2. S-CLAIM readsIt recognises the decision and its key figure. A figure the signal does not give is asked for, never guessed.
  3. Code joinsSignals about the same order or shipment become one case. A likely match is asked, never assumed.
  4. Your rules routeYour delegation of authority names who decides. Above a limit, a second person confirms.
  5. One person decidesOne question, the evidence one tap away. Nothing runs without that decision, and everything is recorded.

The intelligence: S-CLAIM

Four modules read every signal. One reading goes to the case.

S-CLAIM is the Aidecis decision model. When a signal comes through the door, four modules each read it from their own angle. Two of them are models; two are code working from your own rules and your own history. Together they give the case one reading the person who decides can check.

The signal
SAP · 07:12Supplier confirmation: PO 4500077103/00010 moves from 20 to 27 November.

The same four modules read every source: email, o9, WhatsApp, your own agents.

S-CLAIM · four modules
Model · trained by AidecisSupply-chain decision modelWhich of the 72 supply-chain decision types this is, from expediting freight to a product recall; its key figure, what it protects first and how urgent it is.A purchase-order exception: the supplier moved the date by 7 days. Time comes first; urgent today.
Code · your rulesYour policiesYour delegation of authority: who decides, under which limit, and when a case must wait for a person.Within €50,000: the procurement manager decides.
Code · your historyEarly-warning patternsWhat your own decisions and outcomes show, computed in your environment and kept there.DeltaRevin was late on 4 of its last 10 orders.
Model · your languageLanguage modelReads the signal as people write it, in English, Dutch or German, and writes the explanation in plain words.Writes the question for Anna, with the reason in one line.
One reading
On the case Each line says where it came from, so it can be checked
  • Decision: purchase-order exception, time first, urgent today
  • Decides: procurement manager, within €50,000
  • Warning: this supplier is often late
  • In plain words, in your language
A missing fact is asked for, never guessed.

Illustrative example with fictional names and figures.

A signal reaches S-CLAIM.

Swipe sideways to follow the flow.

How S-CLAIM is trained

Trained on how supply-chain decisions work. Your data only if you opt in.

General AI models learn from whatever text they can find. S-CLAIM learns from a body of supply-chain decision knowledge that Aidecis built and owns, in five steps.

  • 1 · The decision library72 decision types in 12 families, from expediting freight to a product recall. For each type: the signal that starts it, the evidence it needs, who holds the authority and which limits apply.
  • 2 · Rules judged by practitionersFor every type, when a decision escalates, when it gets reopened and what it protects first is written down as a rule, and each rule is judged by Aidecis's supply-chain leadership. No rule goes in without that verdict.
  • 3 · Realistic casesFrom those rules come thousands of cases in English, Dutch and German, with values checked against public freight and procurement data. Who approves each case is worked out by the same policy engine that runs in Aidecis, so S-CLAIM learns the route your policy takes.
  • 4 · Knowing when to stopSome cases leave out a deciding fact, or carry one that is out of date or unconfirmed. There the right answer is to hold and ask for that fact. S-CLAIM is trained to say what it needs first, instead of guessing.
  • 5 · Tested before releaseEvery version sits a fixed exam of 300 cases that Aidecis's supply-chain leadership answered blind, and that S-CLAIM never sees in training. A version is released only when it beats the general model it started from.

How each version gets better

  1. Thousands of casesRealistic supply-chain decisions in English, Dutch and German, across all twelve families.
  2. Checked by senior expertsAidecis's senior supply-chain leaders judge the rules and check answers by hand.
  3. Trained in the EUAn open-weight model, fine-tuned on Aidecis's own servers. No client data without a written opt-in.
  4. A blind examA fixed set of test cases, safety signals among them, answered blind by those same experts.
  5. Released only when betterA version replaces the last one only when it scores better on that exam.

Reviews from fictional demo cases feed the next round. A client's cases train S-CLAIM only with that client's written opt-in.

Why it is different

Your systems stay. Aidecis connects the decisions across them.

Aidecis does not replace SAP, o9 or your communication channels. It reads from them, read-only, and adds what none of them has on its own: the whole case, the authority, and the record.

  • Read-only by design

    Aidecis reads your systems and never writes into them. Your own system carries out an approved action.

  • Your authority, applied by code

    The model reads and explains. Who decides, and whether a case waits for a person, is set by code from your own rules.

  • Every signal kept with the decision

    The record shows what each system and each person reported, who decided, and what happened next.

  • Your data stays in your environment

    Hosted in the EU. Your data trains S-CLAIM only if you opt in, in writing.

  • Owned, not borrowed

    The decision library, the rules and the cases behind S-CLAIM are Aidecis's own work, built from years of running supply chains.

  • European by design

    S-CLAIM is built on an open-weight model and runs on EU-owned infrastructure. No third-party AI vendor is in the path.

Where it stands

What you can see today, and what a pilot adds.

  • Available now

    SAP reads for purchase-order and reroute decisions. Messages pasted into the signal inbox. Signals joined into one case on the same order line or shipment.

  • In the demo, simulated

    o9 alerts, WhatsApp messages and supplier emails coming in through the same door, in each source's own format.

  • In a pilot

    Connecting your own o9, the mailboxes of the key people and your WhatsApp Business account, with their consent.

See it work

Bring a decision that reaches you through four channels.

We show how Aidecis joins the signals, who should decide, and what the record looks like afterwards.