Machinegeist
Vision

We build organizations where intelligence compounds.

Most firms that sell you AI have never run on it. We started the other way around.

the operating model, read top to bottom ↓
01 — The founders' story

We built the operating model for ourselves first.

Before Machinegeist took a single client, we built our company to work the way we believe every company will soon have to. Our entire operating model is legible to AI. Visual design, backlogs, policies, strategy, pricing, CRM, finance. Everything our company needs is codified, and Machinegeist's agents read and act on it every day.

Our competitors think we never sleep. That's because we have agents, and they have prompts.

Two convictions got us here. Christophe spent twenty years inside companies watching the same thing happen over and over: the larger an organisation becomes, the dumber it gets. Laurens spent fifteen years stringing systems together in places where a wrong answer has dire consequences, and learned that the wiring is more important than the model.

Put those two together and you get the bet this company runs on. AI by itself does not make an organisation intelligent. All companies have to be rebuilt to accommodate both humans and agents.

Design system · design-tokens.json
~/machinegeist/brand/machinegeist/design-tokens.jsonread daily
{
"color": {
"ground": "#FAF6EF"
"ink": "#1A1A2E"
"slate": "#4A5568"
"accent": "#307878"
},
"type": {
"display": "IBM Plex Mono"
"body": "IBM Plex Sans"
"scale": [16, 22, 30, 46, 56]
},
"radius": { "max": 8 },
"grid": { "base": 4, "cols": 12 }
}
02 — The intelligence is a utility now

Once everyone can buy the same intelligence, the model stops being the edge.

As AI agents arrive, your competitor can buy the same intelligence you can. Whatever edge it hands you, it hands to everyone on your street at the same moment.

If the model is not the advantage, where is the moat? We are betting one kind of company races ahead: the firm that takes what it alone knows, its institutional memory, its data, its workflows, and rewires that to play well with AI.

The model has no idea why your company behaves the way it does. The only durable edge is your firm's tacit operating knowledge, like why you walked away from that deal last spring. The job is to capture that expertise and build the rails that run on it.

You can buy tokens, but you can't buy change.

The moat · swap the model, keep the rails
Your operating modelcodified, legible to AI
AGENTS.mdcodified
policies/codified
pricing/codified
crm/codified
finance/codified
design/codified
Context / harness layer we engineer this
the rails above run on a foundation model you can swap
model claude swappable
03 — One file, many lenses

One source. Every team reads its own view.

While our vision has deep roots, we are practical people, and here is what it looks like on the ground.

Think of a signed contract. Sales cares about the deal value, finance about the payment terms, legal about the liability cap, and operations about what was promised. In most organisations, that means four people inputting one document into four systems. In our world, the contract stays one source. Each team sees its own view of it, and because every team reads from the same source, nothing drifts. The renewal flags itself 10 months later without anyone having to think about it.

The same trick works on a call transcript: it contains an expansion signal for sales, a satisfaction read for support, a feature request for product, and a compliance mention that interests legal. The agents can do the reading for them.

Stop using your best people as the wiring between your software.

One source · four lenses
Services Agreement
Deal value€67,500
PaymentNET 15
Liabilitycap 100% of fees
Deliverablesmodel + use case
Term12 mo, auto-renew
Sales€67,500deal value
FinanceNET 15payment terms
Ops2 itemswhat was promised
renewal flags itself in 10 months
04 — The work, and who does it

We sit with your team and build it, until the system runs without us.

AI tool vendors tend to gloss over the hard part of how their point solution will integrate in the larger system. You can't fix this with prompting, nor with a smarter model.

What it takes is a real grasp of how your company decides, and the patience to draw that out of your people and set it down in plain words a machine can act on. Or plain code, an LLM is not always the answer. Our forward deployed engineers sit with your team and build it with you, until the system runs without us.

Embed one of them and within weeks your company feels different on an ordinary Tuesday. Who they are and how they work is its own page →

Almost nobody can get your company out of its own head and onto the page. That is the whole job.

Forward deployed · contract-reviewer
~/finance/contract-reviewershipping
scaffold reviewer
run on localhost:3000
deploy to the cloud
pentest passed
status: runs without us
The people building it
Christophe RosseelCo-Founder · Antwerp
Track record/Operator & Exec turned founder. Two decades rebuilding operating models at technology companies, long before AI made the work urgent.
Previously/Netlog · In The Pocket · imec · Robovision
Writes/Complexity Matters newsletter · De Tijd column
Laurens Van AckerCo-Founder · Ghent
Track record/Industrial computer engineer. Fifteen years across enterprise integration, SaaS, and large-scale data.
Previously/CTO & Chief AI Officer at BOEMM!, one of Belgium's fastest-growing HR-tech scale-ups.
Domain/Policing · public sector · DAM · building-energy coordination

We built that for ourselves first. Come see it running on us, then decide whether you want it running on you.

Start a conversation

Tell us what you need.