TASK TAMERS

The decisions your business repeats all day — made in under a second.

Refunds, leads, support tickets, applications. Your systems send the case, the Decision Engine returns a decision and how confident it is, your software runs the action. Anything it isn’t sure about goes to a person.

Book a discovery call

No cookies. No trackers. Nothing on this page phones home.

The Decision Engine. Your systems send the case. It makes the judgment. Your software runs the action.

The case, as it arrived

Returns form

Order
#48213, delivered 9 days ago
Account
First return in two years
Message
“The jacket arrived with a broken zip. Worn once before I noticed. I’d rather have a refund than an exchange.”

What the Decision Engine scored

  • Claim plausibility91

    Specific defect, consistent with the item

  • Fraud risk6

    No return history, no address mismatch

  • Policy match88

    Inside your window, defect category

  • Route confidence93

    Refund, not exchange

Approve the refund

Refund issued and a return label sent. No agent ever opened the ticket.

Median decision: under 250 ms. The case is decided and routed before anyone has opened it.

Illustrative output. The scores show how the Decision Engine reasons — they are not measured results from your systems.

Use cases

Twenty decisions, four departments.

Every one of these is the same mechanism with different questions: your case in, a decision and its confidence out, your own rules choosing the action. None of it is a prebuilt integration — the questions and thresholds are configured against your past cases.

E-commerce

Returns, orders, goodwill

  1. Refund or exchangeIs the return inside policy, is the defect plausible, and does the account show fraud signals?
  2. Order cancellationCan it still be stopped before it ships, and does the reason qualify?
  3. Delivery complaintsCarrier fault, warehouse fault, or a claim that needs evidence first.
  4. Goodwill and discount requestsDoes this customer's history justify it, and up to what amount.
  5. Review and photo moderationPublish, hold for a person, or reject against your content rules.

Anything above your value threshold is approved by a person, not the engine.

Sales

Inbound leads and replies

  1. Lead qualificationA real buyer, someone browsing, or a competitor looking at your pricing.
  2. Routing and ownershipWhich segment, which queue, which rep — by your rules, not round-robin.
  3. Spam and bulk filteringA genuine enquiry or another agency pitching you their services.
  4. Reply intentInterested, not right now, wrong person, or asking to be left alone.
  5. Pipeline hygieneDoes what the rep wrote actually match the stage the deal sits in?

The engine decides order and ownership. It never contacts anyone.

Customer support

Triage, routing, escalation

  1. UrgencyIs work blocked, for how many people, and is there a deadline in the message?
  2. Department routingBilling, access, technical, account management — read from what was written.
  3. EscalationDoes this need a senior or a manager now rather than at the queue's pace?
  4. Churn and frustration signalsIs this the third message about the same thing, and is the tone changing?
  5. Duplicate detectionA new issue, or the same one arriving through a second channel.

Tickets mentioning legal action or safety go straight to a person, whatever the score.

Recruiting

Screening, never rejecting

  1. Requirement matchDoes the application meet the must-haves you actually wrote down?
  2. Bulk application detectionWritten for this role, or the same letter sent to two hundred companies.
  3. Role routingWhich of your open positions this person actually fits, if any.
  4. Seniority bandJunior, mid or senior against your ladder — not against the job title claimed.
  5. Screening answersAre the answers complete, consistent, and do they contradict the CV?

Every one of these is a summary for a person to read. The engine never rejects a candidate.

Your numbers

What these decisions cost you by hand.

Nothing here is a benchmark or a claim about your business. It’s arithmetic on figures you enter. Change any of them.

Your decision figures
20 hHours a month on decisions that follow a rule
€900Cost a month
€10,800Cost a year

This counts handling time only. It doesn’t try to price what slow decisions cost you in churned customers, cold leads or candidates who took another offer — we have no honest way to estimate that for you.

Delivery

Live in a week.

Days 1–2

We test on decisions you’ve already made

You hand over a few hundred past cases along with the call you made on each one. We run them through and show you, case by case, where the engine agrees with you and where it doesn’t. You see it being wrong before you pay attention to it being right.

Days 3–5

You set the rules

What a good outcome looks like, your categories, who receives what, and how unsure a decision has to be before a person takes it. These are your thresholds, not ours, and they stay editable.

Days 6–7

It goes live

Real cases flow through. Every decision is logged with its confidence so you can audit any one of them. Unsure cases go to a person from the first day, not as a later upgrade.

Be honest with yourself

This is not for you if…

  • You make fewer than a few hundred of these decisions a month. The engine won’t pay for itself, and a written checklist probably will.
  • Every case is genuinely different. This works on decisions that repeat — one-of-a-kind judgement calls stay yours.
  • You want the final say automated on high-stakes calls: hiring, credit, anything with a legal consequence. It won’t do that, and we’d talk you out of it if you asked.
  • Nobody owns what happens next. Deciding faster changes nothing if no one acts on the decision.
  • You can’t share past cases. The pilot has nothing to test against without them.
  • You need it running tomorrow. It takes a week, and most of that week goes on your rules, not our code.

Before you book

The questions people actually ask.

Which decisions can it actually handle?

Ones that repeat, that a competent person could make from the information in front of them, and that you could explain to a new hire in a morning. Refund approvals, lead routing, ticket triage, application screening, document classification. Not one-off strategic calls — those need context the engine doesn’t have.

What happens when it gets one wrong?

Every decision is logged with the confidence behind it, so you can go back and look at any single case. Wrong calls are how the thresholds get tuned — you’ll see them in the weekly report rather than finding out from a customer months later.

Does this replace my staff?

No. It handles the decisions that repeat and hands your people the ones that don’t. Nobody gets contacted by the engine.

What happens to our data?

Cases are sent to a third-party model provider to be scored. Before a single real case is processed we’ll name every sub-processor in writing and sign a data processing agreement with you. If that’s a blocker for your legal team, raise it on the call rather than after.

Can we try it before committing?

That’s the pilot, and it’s the only way we start. Your real past cases, your rules, and you see the decisions before anything runs for real.

We already have rules for this. Why this?

Rules work on fields — a dropdown, an amount, a checkbox. This reads what the person actually wrote. If your existing rules already catch the awkward cases, you don’t need us.

What languages do you handle?

German and English today. Anything else, ask on the call and we’ll tell you honestly whether we’ve tested it.

How much is it?

It depends on how many decisions you run and how complicated your rules are, so quoting a number here would be a guess. You’ll get a real one on the call, and it won’t change afterwards.

Start here

Start with the pilot.

Your own past cases, your own rules, and you see every decision before anything runs for real. If the pilot doesn’t convince you, there’s nothing to cancel.

Book a discovery call

One call, about 30 minutes. If you’re not a fit, we’ll say so on the call.