The invisible colleague: automating processes with AI

I'm a project manager. My job is keeping the overview, getting people aligned, and making sure things get done on time. I don't write code. I don't design interfaces. I make sure the whole picture works.

And it's precisely from that role that I see something happening that many organizations still underestimate: the quiet revolution of process automation with AI.

Not the big, flashy AI demos you see on LinkedIn. Not the chatbots answering customer questions. I'm talking about the invisible things. The processes nobody finds sexy, but that keep a business running — and that are now, one by one, getting smarter.

The problem nobody names

Every company has them: processes that work, but don't really work. They cost too much time, are error-prone, and depend on that one person who "knows how it works."

A few examples I encountered at clients over the past year:

• An employee who spent two hours every Monday manually transferring data between two systems
• Invoices arriving by email, manually retyped into the accounting software, with regular errors creeping in
• Status updates traveling through eight different channels — Slack, email, WhatsApp, verbally — so nobody has the full picture
• Onboarding new employees going slightly different each time, because the process lives in someone's head

None of these are big problems individually. But combined: hours per week. Frustration. Errors. And the feeling that you're spending time on things that should just happen by themselves.

Automation isn't new — but AI makes it accessible

Process automation has existed for decades. Large companies invest millions in ERP systems, workflow engines, and RPA solutions. That works, but it's expensive, slow, and complex.

What AI changes isn't the idea. It's the barrier to entry.

Where you used to need an entire IT project to automate a workflow, you can now build something with AI tools in days that delivers immediate value. Not a proof of concept that disappears into a drawer, but a working solution that can go live tomorrow.

At Tech_42, we experience this daily. Clients come with a problem they've had for years, assuming it's "just the way it is." And then we show them it can be solved in a fraction of the expected time.

That surprise — that moment when someone realizes a process that cost five hours per week now runs automatically — that never gets old.

Three patterns we see everywhere

After dozens of projects, I keep seeing the same three categories. Not every automation fits neatly in a box, but these three cover the vast majority:

1. Data that needs to go from A to B

It sounds simple, and it is. But it's astonishing how many companies still manually copy data between systems. From CRM to accounting. From webshop to inventory system. From form to database.

AI makes this not just faster, but smarter. It recognizes patterns, corrects errors, and warns when something doesn't add up. Where a traditional integration blindly copies, an AI-powered integration thinks along.

One client had an employee who extracted invoice data from PDFs and retyped it into their accounting software every day. Twenty minutes a day, five days a week. Now AI reads the PDFs, extracts the right fields, and books them. That employee now spends those twenty minutes on work that actually matters.

2. Communication that needs structure

Most companies communicate via everything simultaneously. Email, Slack, Teams, WhatsApp, phone, sticky notes on the desk. The information exists, but it's scattered across ten channels.

AI can structure that chaos. Not by adding yet another tool, but by smartly connecting existing channels. A Slack message that automatically creates a task. An email that reaches the right person based on content. A status update that comes together in one place, regardless of where it originated.

It's not about replacing communication, but about unburdening it. So people can do what they do best — communicate — without worrying about whether the information ends up in the right place.

3. Knowledge that needs to be captured

This is the hardest, but perhaps the most important category.

Every company has unwritten knowledge. How does that one process work? Where's that documentation? What are the agreements with that client? It lives in heads, in old emails, in folders on a shared drive that nobody opens anymore.

AI can unlock that knowledge. Not by neatly organizing everything in a wiki — we've tried that ten times already, and it never gets maintained. But by adding an intelligent layer over existing information. An assistant that knows where things are. That answers questions you'd otherwise ask that one colleague who's been there for fifteen years.

That's not science fiction. That's now.

Where do you start?

The question I get most often: "Sounds good, but where do I start?"

My answer is always the same: start with the frustration.

Not with the big strategic vision. Not with a twelve-month AI roadmap. Start with that one thing that frustrates your team every week. That one process everyone says: "Surely this should be easier?"

That's your starting point.

Make it small. Make it concrete. Solve it. And when it works, take on the next one. Ring by ring, as Chris would say.

The pitfalls

Honesty compels me to also mention where things go wrong. Because they sometimes do.

Thinking too big, too fast "Let's automate everything!" No. Start with one process. Prove it works. Build trust. Then the next one.

Putting technology first Automation isn't a goal, it's a means. The question isn't "where can we deploy AI?" but "what problem do we want to solve?" Sometimes the answer is a smart piece of software. Sometimes the answer is a better form. And sometimes the answer is a good conversation.

Skipping the people The best automation fails if the people who need to work with it aren't brought along. Involve them early. Let them watch. Give them ownership. An automated process that nobody trusts doesn't get used.

The invisible colleague

I like to call AI the invisible colleague. Not because it's unimportant, but because the best work is invisible.

When an invoice gets booked automatically, nobody notices. When a status update lands in the right place by itself, it doesn't stand out. When a new employee has all the information they need on day one, nobody thinks: "How clever of that AI."

And that's exactly the point. The best automation is the automation you don't see. That just works. That makes room for the work that does matter: thinking, deciding, collaborating, creating.

That's what I ensure as a project manager. Not that the technology works — we have developers for that. But that the process is right. That it fits how people actually work. That it becomes simpler, not more complex.

The future isn't spectacular

And that's a compliment.

The future of AI in business processes isn't a robot taking over your office. It's an invoice that books itself. A schedule that updates itself. An employee who doesn't search anymore, but finds.

Small. Quiet. Effective.

Exactly as it should be.

— Jeroen