The AI workspace for small business

Your business,
structured for AI.

An AI assistant that files what comes in, drafts what goes out, and chases what's due - on your own files, set up for you. If you can use folders and email, you can run it.

Book a free demo on your own material +32 471 769 661

Free, about an hour, on your own material. If it doesn't show clear value within 30 minutes, it stops there.

The sales-call recordingThe prospect's follow-up email

What it already knows: your rates, past proposals, what won before

The assistant

Pulls what was promised and what worried them
Researches the company online - who they are, what they do
Prices it from your rate card and past wins
Drafts the proposal and the cover mail

Out

A proposal in their language, ready to send

The first to follow up usually wins. You follow up in minutes.

The problem

Everyone's told you to adopt AI. Almost nobody says what that takes.

A chat window isn't AI adoption

It only knows what you paste in. Useful for a paragraph, useless for running a business.

Your business lives everywhere

Drives, inboxes, people's heads. Point an assistant at that and it can only search, at best.

Structure is the missing step

Organise the files once, give the assistant all of it, and it genuinely works with you.

That unglamorous first step, structure, is the service. It's why everything after it works.

Today

Scattered everywhere

Drives

Inboxes

People's heads

ERP

Jira / boards

Loose files

External systems

An assistant pointed at this can search, at best.

With AI Workspace

One structured home

Projects

Areas

Resources

A small business acting like a big one.

If any of these sound like you, a free demo on your own material will make the point.

You're the owner, and you're the bottleneck.

The quotes, the follow-ups, the who-said-what to whom - it all runs through you. You want a business that keeps moving when you step away, without hiring an office manager to make it happen.

Your business lives everywhere except one place.

Drives, inboxes, a messaging app, a few spreadsheets, and a lot of it in your own head. Answering "what did we agree with that supplier?" means twenty minutes of digging.

You've tried ChatGPT and hit its ceiling.

Handy for a paragraph here and there, but it doesn't know your business, so it can't actually run any of it. You suspect there's a bigger gear, and there is.

You're about to make a big move.

An acquisition, a new system, a major build. Structure your information first and much of the upfront analysis is already done - the project starts cheaper and sharper, and you run it from the same workspace.

You don't need to be technical, and you don't need to tidy up first. The mess is the starting point - that's what the work is for.

A worked example

Monday, 9am. One command.

Overnight, forty emails, a scanned letter, an order, two invoices and a signed contract landed. By the time your coffee's finished, each has been read, filed, and acted on. You didn't open a single folder.

Overnight, in

40 emailsSupplier letter, scannedA customer orderTwo invoicesA signed contract

The assistant reads each one and acts

It works out what each thing is and what to do with it by reasoning over everything you keep.

Your structured workspace - filed history, contacts, projects, past decisions

The live sources it's wired to - your webshop, calendar, accounting, drive

Minutes later, done

Filed where it belongsRenewal creeping up in price, flaggedObvious replies drafted for your OKOrder confirmation on your letterheadWhat's due today, laid out

No software gets built for any of this. It isn't a program written for each task - it's the AI reasoning over your organised data and the sources it's connected to, then acting.

The same inbox, different actions - it depends on what each thing is

A supplier renewal notice

becomes

A flag: the price crept up - confirm or switch, before it locks in.

A customer's order

becomes

Lines checked against live webshop prices, the buyer matched to your records, a confirmation drafted on your letterhead.

A signed contract

becomes

The deal promoted from prospect to a live project, every link repointed, the brief pulled together.

It goes further than filing

Filing is the floor, not the ceiling. The same assistant helps you plan and run the work itself: pulling a project brief together, checking a quote's numbers before it goes out, chasing the follow-ups a job depends on.

Beyond the routine

The part that surprises people

The routine work pays for the setup on its own. What people remember is the moments it does something nobody thought to ask it for. Four real ones:

A whole website, straight from the files

For a property-development partnership, a complete marketing website was generated straight from the workspace - no input from the owners needed, and they were shocked at how accurately it captured their business.

It checks the numbers behind a deal

Asked for a second pair of eyes on an investment case, it worked through a spreadsheet of complex financial calculations and caught the kind of subtle slips every complex model hides.

It finds money you're leaving on the table

My own financial admin is wired into Odoo, the accounting system. When a telecom or energy invoice lands, the assistant checks the market for a better deal and reports what switching would save per month - advice nobody asked for, from a bill it was only meant to file.

It steps into the technical work

In one working session it edited a technical drawing from a spoken instruction, caught and redid its own first attempt, then dropped the corrected version into a quote on the company's own letterhead.

Proof

Already in daily use

This isn't a concept. It's how I run my own life and business, and how a real multi-business operation runs day to day. A committed post-acquisition engagement at a Belgian manufacturer starts next. Every row below is real, and running today.

Who runs itWhat they run it for

SME owners and operatorsDeals, contacts, and key documents in one place - a scattered pile becomes clear projects and resources, linked back to the emails and meeting recordings behind them.

Family affairsSchool letters, insurance renewals, the contractor's quote - filed and linked to the emails behind them, so it spots the renewal creeping up in price and where you're overpaying.

Financial adminWired into the bookkeeping - "what did we spend with this supplier last quarter?" answered in seconds. It does the prep an accountant charges by the hour, cutting the bill dramatically.

An IT engagementMeeting transcripts dropped in the inbox get filed, their decisions linked to work items on the live board (Jira, Azure DevOps), and pull requests opened in real time.

And the method itself is public - you don't have to take my word for any of it:

Built by TrotStar Technologies

para-os

The open-source kit behind AI Workspace, free to use: PARA folders, plain-language house rules, and the skills that run them. Includes a fully worked example workspace you can poke at before you talk to me.

Want to see the shape first? The BelFoot example is a complete, fictional engagement you can read end to end.

How it works

Three layers

In one line: your business files, organised so an AI assistant can work across all of them - set up for you, on your own storage.

Any "AI on top of your stuff" setup is three layers stacked. Most tools live in one or two of them and stay quiet about the rest.

How you work with it

The window a human drives or reads it through

VS Code · Obsidian · a chat app

The assistant

The AI that reads and writes the files

Claude Code · Cursor · Codex · Claude Cowork

Your files

This is AI Workspace

The files and how they're organised

PARA folders + house rules the assistant re-reads each session

Layer 1

Your files (the part I build)

This is the layer almost everyone skips, because it feels like filing, not engineering - and it's the one that makes every assistant above it actually work. It's a folder you copy, not a program you install - plain files on your own storage, organised so every "where does this go?" has one answer. Under the hood it's para-os, my free, open-source kit. I set it up for you; you keep it.

The folder structure

The folders follow PARA, a simple filing method sorted by how you act on things, not what topic they're about:

Projects

Time-bound work with a deliverable or deadline. One folder each.

Areas

Ongoing responsibilities and the people you deal with.

Resources

Reusable reference, and ideas not yet committed to projects.

The assistant

The AI that reads and writes those files. Claude Code is the pick today because it works best on this kind of workspace, but the files are plain text any capable assistant can read. The whole field is moving this way - and the edge belongs to whoever structured the files, not the model.

It's also the layer that reaches the rest of your world, wiring the workspace into the systems your work already lives in - ready-made connectors where they exist:

Google Workspace or Microsoft 365 - mail, calendar, and drive.

Shared folders and document systems - linked straight in.

Project and dev tools - Azure DevOps, Atlassian / Jira, and the like.

No connector? It writes its own. An ERP like Odoo, your webshop, a legacy database - if it has an API, the assistant sets up a connector on the spot and starts using it in the same session.

Bring your own agent

You're never locked in. As assistants get better, clean files get more valuable, not less - and when a better one arrives, you point it at the same workspace and keep everything.

Layer 3

How you work with it

The window you drive or read it through - in practice, VS Code with Claude Code, simplified so it's no harder than working with folders and documents. You ask in plain language, the assistant edits the files, and the document refreshes in front of you. If you can use folders and email, you can run it.

An AI Workspace running in VS Code with Claude Code - a daily brief assembled live from the BelFoot example workspace, bucketed into overdue, this week, and next 30 days

The window most owners use: VS Code with Claude Code - here, a morning brief assembled live from the workspace, bucketed into overdue, this week, and the next 30 days (running on the BelFoot example).

The rule

Structure first, software second

A clean picture of your whole company shows what software you actually need: a new ERP, something lighter, or nothing new at all. "Nothing new" comes up more than the software industry likes to admit - most point tools are just structured information plus questions, and an assistant over clean files answers them without another subscription.

It doesn't replace your ERP - it sits alongside it.

The ERP keeps the rigid finance, orders, and inventory core; the workspace is the flexible day-to-day layer around it. Structuring first also makes the ERP decision better: you commit knowing what you actually need.

Structure before decisions: once your information is legible, every call that matters gets made on a clear picture.

The engagement

Phased, hands-on, and safe to stop.

Every phase ends production-ready, so you can stop at the end of any one and keep something live. Continuing is a choice earned by results, not a contract. Expect weeks, not months.

Taste

Free

It starts with a conversation - no pitch deck, no sales process. Then a free demo built live on your own material: something concrete in hand by the end of the meeting, at no charge.

Goal

Free

Agree the high-level outcome the engagement points at, so there's a named destination before money changes hands.

Discovery

The first paid phase: map the landscape - licences, mailboxes, shared drives, tools, where data lives and who owns it. Its first output is a realistic estimate for everything after.

Connecting

Wire the workspace to the sources discovery surfaced: mail, shared folders, document systems, your webshop - ready-made connectors where they exist, ones the assistant writes where they don't.

Interviews & triage

Get what lives in people's heads into the workspace, and process the existing material source by source into usable outputs: customer and supplier profiles, searchable archives, process flows.

Future projects

With the clean picture in place, the follow-on work chooses itself - the software decision, agents, automation, training, and a light monthly review to keep the workspace clean.

What it costs

Structuring work: no fixed price is honest before discovery - nobody knows yet how much data there is or where it lives. So it runs in small phases at an SME-friendly hourly rate with a cap: you hear the rate in the first conversation, pay only for time used, and can stop at the end of any phase and keep everything built so far.
Defined deliverables (an agent, a build, a training day): fixed scope, agreed up front.
Either way: no open-ended commitments, and the workspace is plain files on your own storage - nothing stops working when an engagement ends.

About

I'm Vincent Weijburg. Twenty-plus years across every role in the software delivery lifecycle, from developer and business analyst to enterprise architect and team lead, for organisations ranging from global consultancies and energy companies to international institutions. The thread through all of it: structure first, then build - and being direct about what's needed, even when it's not what people want to hear.

Then, six months ago, I discovered Claude Code, and it changed my view of what IT can do. Work that used to take a team and a budget now happens in days, when the foundation is right. That's why I built this: big-company leverage, set up for businesses run by their owner.

How I work

Show, don't tell

Every conversation starts with a working demo on your own material. If it doesn't show clear value within 30 minutes, it stops there.

Integrity over revenue

No pressure to oversell, extend engagements, or recommend complex solutions when simple ones work. Sometimes the honest answer is that you don't need new technology at all.

Agile by design, never lock-in

Work runs in phases, and every phase ends with something concrete and production-ready in your hands. Stopping is always safe; continuing is earned by results.

Independence as the goal

Every engagement ends with something you run yourself - plain files on your own storage that keep working whether or not I'm around. Retainer dependency is a sign of failure, not success.