Book an intro call

AI workflow automation · companies of 5 to 50

You didn't start a company to retype spreadsheets.

I start with your process, not with a tool. What gets built runs on software you already pay for, and it's yours outright.

01 / Fit

Who this works for, and who it doesn't.

I work with a handful of companies at a time and go deep on the process. That's why an audit takes three weeks instead of three months.

This works if

  • A lot of your work happens away from a desk. Planning, timesheets, endless back and forth with clients and staff.
  • Five to fifty people on the payroll.
  • You're growing, and the admin is growing faster than the revenue.
  • You already run software you'd rather not replace.
  • You want to know what you're automating before anyone builds it.

This doesn't work if

  • You're past fifty people. The depth I work at doesn't scale beyond that, not with one person doing it.
  • You want a chatbot. That's a tool. I'll point you somewhere better.
  • You want AI but can't yet say what for. Get in touch anyway, just know it's a conversation and not a project.
  • You're choosing on price. There'll be a cheaper quote than mine, and you should probably take it.

Not sure? Book the call. You'll know inside thirty minutes, and it costs you nothing if the answer's no.

02 / Boring is beautiful

Most AI projects fail from overengineering, not from too little AI.

That's my position, and I'll defend it in a first call. Here's what it looks like on the ground.

A chatbot nobody opens, because it sits beside the work instead of inside it.

A dashboard nobody reads. The numbers were already in the system it pulls from.

An integration that broke the first time somebody changed the process.

03 / Method

Automation that survives contact with reality.

Workflow automation means handing a job people do by hand to a system. The trigger, the steps, the exceptions, the record afterwards. Building that takes a week or two. Making it survive the next twelve months is the actual work.

Most projects get built tool-first. You won't notice in month one. You'll notice in month twelve.

The usual automation project How I build it

One integration per problem, built on whatever tool was current that quarter.

One layer across the systems you already run, shaped around the process.

Works until somebody changes the process, or a vendor ships an update.

Survives both, because the process came first and the tool came second.

The knowledge lives in the builder's head. You're tied to their platform.

You own it. Documentation included. Walking away is possible.

Isolated wins, no direction. Five tools that don't add up to anything.

Each build stands on the last one.

04 / A real build

One that's running right now.

A few times a year, everyone on the payroll has to report their holiday dates. It used to go out by hand, one message at a time, in several languages, and every reply was retyped into the system afterwards. Client name and figures removed, mechanism unchanged.

Schedule RECURRING TRIGGERS Send PER PERSON, PER LANGUAGE PERSONAL LINK Form ONE LINK PER PERSON ON SUBMIT Look up person EMP. NO. → RECORD ID WRITES Write back SYSTEM OF RECORD NO REPLY Reminder AUTOMATIC SENDS AGAIN NO MATCH Review list A HUMAN DECIDES
The two copper parts are what a demo never has: a lookup step, and somewhere for the exceptions to go.

What took longest

Not the automation. It was finding out the target system wanted its own internal record id, not the employee number every human in the building uses. Two days, one undocumented endpoint. That's the sort of thing an audit turns up before a quote rather than after.

The exception path

The first reminder round went out on its own. A handful landed on the review list for a person to look at instead. Those aren't failures, they're the design: when a match isn't certain, a human gets it rather than the system guessing.

Still open

The write-back is built and tested, waiting on a signed data processing agreement before it goes live. Full numbers and the client's name land here once they approve publication.

05 / Patterns

Where the time goes.

Four things I keep running into. Recognise two of them and there's probably something here worth an hour of your time.

Back office

The recurring request

A few times a year everyone has to hand something in. An expiring certificate, availability for next quarter, whatever it happens to be this round. It goes out person by person, and the replies get retyped by hand.

BecomesOne round that sends and chases itself, feeding a form that writes straight into your system.

Sales & delivery

The same document, again

A quote, a report, a work order. Rebuilt by hand in your house style every time, and it comes out slightly different depending on who made it. Twenty minutes if they're quick.

BecomesThe template fills itself from data you already hold. Your people check the result rather than typing it.

Finance

The spreadsheet nobody touches

Cost price or margin, living in a workbook one person really understands. They go on holiday and it stops. Nobody's certain the rates in it are still current.

BecomesThe same logic, but traceable and tested, with its assumptions written into the file. Not cleverer. Just checkable.

Operations

Two systems, no conversation

The same detail typed in twice, sometimes three times. Planning, then the back office, then an email to be safe. The mismatch surfaces at invoicing, weeks later.

BecomesIt gets entered once, and an integration updates the rest. A review list replaces the inbox.

06 / Approach & pricing

Fixed price, agreed before anything starts.

No hourly billing, no open end. Every engagement is quoted for what it is, so the figure you agree is the figure you pay.

00

Workflow teardown

One day · fixed

Not ready to commit to a full audit? Pick the one process that's eating the most time. I take it apart in a day and you get it back in writing: where the hours actually go, what could be dropped outright, and what a build would cost if it turns out to be worth it. Credited in full against the audit if you carry on.

€1,500fixed, excl. VAT

01Start here

Audit

3 to 4 weeks · one day a week

Fixed feequoted per engagement

I map your processes with the people who run them, then put every step to three questions: can this go, can a system do it, can someone else do it? You get a roadmap: what's worth automating, and what each one is actually worth.

  • Sessions with leadership and 3 to 5 key people
  • A process map per department
  • 5 to 10 automation candidates, prioritised
  • A recommendation for the first build

No obligation to build after it. If the answer is that you shouldn't automate anything, that's the answer you'll get.

Half on signing, half on delivery. If the deliverable isn't worth it to you, the second half is waived.

02

Build

Typically 4 to 8 weeks per phase

Fixed per phasequoted on scope

The roadmap gets built one phase at a time, each with its own price and acceptance criteria. It runs in your environment, on your own data. Nothing leaves the EU. A phase that starts growing gets split in two rather than stretched.

  • Built in your environment, not on a platform of mine
  • Tested against real data before go-live
  • Handover with documentation your team can read
  • Fixed price, so scope creep is my problem
03

Care

Ongoing · cancel per quarter

Monthly retainerquoted per engagement

A system that works is invisible, and invisible work starts to feel like nothing. So Care is built around something visible every month rather than hours on standby.

  • Monthly report: volumes, hours saved, errors caught
  • One visible improvement a month, unprompted
  • Monitoring, and repair when something breaks
  • A day each quarter to find the next candidates for automation
  • First call on my calendar for new phases

You get the full figure in writing before anything is signed, and it doesn't move afterwards. Existing clients can book single days; new work starts with a teardown or an audit.

07 / Questions

What people usually ask.

Do we have to replace our current software?

No. The whole premise is that your system of record stays put. What I build sits on top and takes the handwork out from between your tools. If replacing something really is the right call, it'll be in the audit with the reasoning, and you decide.

Will we be locked in to you?

No, and that's deliberate. Nothing runs on a platform of mine. The build sits in your environment, on your accounts, documented well enough that your own people can follow it. Want to continue with someone else next year? You can, without starting over.

What about GDPR and the AI Act?

Anything touching personal data runs on EU hosting, with a language model in an EU region. We sign a data processing agreement per engagement, plus one with every external party in the chain.

The AI Act is in force now. Article 4 requires any organisation deploying AI to make sure the people working with it are sufficiently AI-literate. I cover that in the audit, including the record that shows you comply.

Why an audit first? We know where it hurts.

Usually you do, roughly. What's almost never clear up front is what a step really costs, where the time actually goes, and which step turns out to be droppable after two questions. That's the difference between spending twelve thousand on the right build and spending it on the wrong one. The audit is the cheap way to find out.

Do you work on site or remotely?

Remote-first. Sessions, build and handover all work fine at a distance, and it's far easier to get twenty minutes of the right person's time when nobody has to block out a morning for it. If you'd rather have me in the building one day a week, that can be arranged.

Outside the Netherlands it's fully remote, in English. The work is in your processes and your systems, so where I sit matters less than you'd think.

How long before something is actually running?

Audit is 3 to 4 weeks. A first build phase after that usually reaches production in 4 to 8 weeks, depending on how many systems are involved and how fast access and approvals come through. Those approvals are the delay more often than the building is, which is why we chase them from day one.

Can we start small with one automation?

You can, and there's a cheaper way in than the audit: the one-day teardown. Pick the process that hurts most, I take it apart, and you get it back in writing with a price if it turns out to be worth building.

What I'd push back on is building one automation without a picture of the process around it. That's exactly how these projects run aground. It works for six months, then the process shifts and it fits nothing.

Get in touch

Let's talk.

Emailinfo@devaware.nl Phone+31 6 38 73 29 32 LinkedInTaylen Doesburg
WorkingRemote, from Goes (NL)
LanguagesEnglish, Dutch