When the spreadsheet stops holding
Most businesses don't suffer from a shortage of technology but from a surplus of files. A good system puts the process in one place, calculates what is calculated by hand today, and warns you before something falls — not after.
A system is built from the process, not the screen
Almost every system that fails, fails for the same reason: it was built around screens rather than around how the work actually moves between people. So I start by mapping — who enters what, who approves, what happens when someone's on leave, and exactly where the information breaks today.
My background is a degree in Information Systems and a Certified Systems Analyst credential, and that's precisely this work: translating a business process into a system real people are willing to use. AI goes in where it shortens something — classification, summarising, filling gaps, spotting outliers — not so we can say there's AI.
Two systems like this are running at clients today: one managing funding calls and budgets worth millions, and one managing hours, staff and payroll for hundreds of employees. Both were built on an existing process that spreadsheets could no longer hold.
What goes into a system
Every system is assembled differently, but these are the parts that come back almost every time.
One place for everything
Clients, projects, tasks, documents and dates — without five files and three versions of the truth.
Automatic calculations
Hours, rates, budgets and balances calculate themselves, so a typo doesn't become a payroll error.
Statuses and alerts
Every process knows where it stands, and an approaching deadline announces itself before it passes.
Role-based permissions
Who sees what, who approves and who only enters — with a record of who changed what, and when.
Reports on tap
Reports you pull in a click instead of rebuilding every month in a separate sheet.
Integrations and AI
Connected to whatever already works for you, with AI components for classification, summaries, gap-filling and outlier detection.
How it gets built
From the map to a system the team genuinely works in — in stages, with something working early.
- 01
Map the process
Sitting with the people who do the work, not only with management. Out of that comes a document describing the process as it is, and the gaps.
- 02
Scope and data model
What gets stored, how it links, what calculates automatically and what stays manual. This stage decides whether the system still holds in two years.
- 03
Build in versions
Start with the part that hurts most, put it into real use, fix on the strength of what happened — then carry on.
- 04
Rollout, training, support
Data migration, team training and a close support period until the system becomes the default way of working.
Why it holds
A system is judged in its second year, not in its launch week.
- A data model built right up front — so adding a field or a new process doesn't mean rebuilding everything.
- Built with the people who'll actually use it, because a system the team works around is a system that doesn't exist.
- Released in versions, so risk is spread and you can stop or change direction at any stage.
- Documentation and training that stay with you, alongside a clear line on what you can change yourselves.
Questions that come up
Why build rather than buy off the shelf?
If there's an off-the-shelf product that fits, I'll tell you — and it'll be cheaper. A custom system earns its place when your process is precisely what makes you different, or when you're already paying for the gap in salaries.
What about our existing data?
It comes across. Part of the work is cleaning and converting what has accumulated in sheets and files, and it's worth budgeting for up front because it's almost always bigger than it looks.
How long does it take?
The first part in real use — usually weeks. A full system is a months-long project delivered in stages, so you're not waiting until the end to see value.
Where exactly does AI come in?
Where it shortens work: classifying enquiries, summarising documents, filling missing data, spotting outliers and drafting text. Where the result must be exact and identical every time, a plain calculation is better — and there I won't put a model.
Who maintains it afterwards?
You get a system with documentation and training, and monthly support for changes and additions is available. I don't build deliberate dependency — it isn't in my interest either.
Do you work with public bodies?
Yes. One of the systems I built serves a public-sector consultancy managing funding calls and budgets, and I work with local authorities too.
