Skip to content
Thnkers

AI automation

Your orders live in Excel. Your pricing lives in one person’s head. Let’s fix that.

We turn manual routine into working systems — in weeks. Automation included where it earns its place.

What’s missing

Orders, quotes and statuses live in Excel and email. Every file is someone’s personal universe.

When the person who owns the file is away, the process stops. When two people update the same file, something gets overwritten. When the director asks for a status update, someone opens three spreadsheets and produces a summary by hand. This isn’t a technology problem. It’s a structural one, and it gets worse as the business grows.

The critical formulas and rules exist in one person’s head. Their holiday is a crisis.

The pricing logic. The assembly coefficient. The exception for one product category. The workaround introduced two years ago because of a supplier change. All of this is real operational knowledge that lives in one place: a person. When that person is sick or leaves, the knowledge goes with them.

The director can’t see the pipeline.

How many deals are active. Where they’re stuck. What they’re worth. The information exists — spread across files, inboxes and memories — but nobody has a single view of it. Decisions get made on instinct because the data isn’t assembled anywhere.

“Maybe AI can help?” — but nobody says where to start or what it will cost.

The question is reasonable. The answers usually aren’t. Vendors sell transformation programmes. Consultants produce assessments. What’s missing is someone who will look at the actual operation, find the two or three places where automation makes a measurable difference, and build those first.

The promise

We don’t sell automation. We remove the manual work that consumes your best people. Sometimes that needs a vision model reading an engineering drawing. Sometimes it’s a properly structured database and a role-based workflow. What you actually need becomes clear from an audit — and we’ll tell you honestly, including if the answer is that automation isn’t the right move yet.

What we do

  • Process audit

    One to two weeks. We map where hours and money are being lost, what can be automated and what can’t, and produce a prioritised plan with estimated ROI for each item. The audit is a standalone deliverable — you can take it to any team.

  • Pilot on one process

    Two to four weeks, on real data. One process, automated end to end, with measurable before-and-after numbers. The decision about whether to continue is made on facts, not a vendor’s promises.

  • Automation programme

    Processes addressed in sequence, each with a defined outcome and a way to measure it. Not a transformation roadmap that lives in a presentation — a series of working systems delivered one at a time.

  • Automation components where they pay for themselves

    Document and drawing analysis. Submission checking against structured criteria. Draft generation for repetitive responses. Where a simple rule does the job, we use a rule. Where a vision or language model saves hours a rule can’t, we use that — and explain why.

  • Compliance by design

    For clients in the EU we build with the AI Act in mind from the start: risk classification, decision logging, human oversight at the right points, documentation in the delivery. Internal automations are usually not high-risk — we give you a written classification with reasoning.

  • Retainer

    The system runs, the metrics are visible, and new processes get added as the business identifies them. We’re available to maintain and extend without starting over.

Compliance by design

For AI work in the EU we build to the AI Act from the start — risk classification, decision logging, data documentation and human oversight, with the technical file included in delivery.

How we work

Audit first — one to two weeks — to find the actual bottlenecks. Then a pilot on the highest-priority process, two to four weeks, on real data. Then a programme if the pilot proves the value. Then a retainer if you want ongoing development. Each step is a separate decision. If the pilot doesn’t work out, we part without complications and you keep a working pilot.

Where we have done it

Questions

  • Which processes should we start with?

    The ones with high repetition and clear rules: order intake, quoting, document processing, status reporting. These automate cleanly and produce measurable results fast. The audit will rank them.

  • Will our data go to external clouds?

    No. Production runs on your servers or your cloud. API calls to external services happen behind your own keys, the same way the Alfa Homework system works. Your data stays in your infrastructure.

  • What about the EU AI Act?

    We build with the requirements in mind from the start: risk classification, logging, human oversight where required, technical documentation included. Internal automations are usually not high-risk. We give you a written classification with reasoning, not a verbal assurance.

  • Our people are worried about being replaced.

    We automate the routine, not the people. In both case studies here, the team got more time for the work they were actually hired to do — not fewer people. That’s the outcome worth measuring.

Order a process audit

Download the automation guide for mid-sized businesses — with a checklist for assessing whether a process is ready to automate.

Order a process audit