I build systems that take the manual work off your team: data scattered across systems lands in one place, reports are always up to date, and with documents you are left checking rather than reading from scratch. I start by understanding how your process really runs and finish at deployment and handoff, entirely on your own infrastructure.
Most of the time lost in a firm is not a technology problem but a process one. Fix the process first, then automate it. That's the only order that works.
Manual credit-file reading gave way to AI-supported analysis where every figure can be checked.
Reading a credit file drops from hours to a matter of minutes — the AI reads it, the lawyer checks the numbers, and the offer goes out the same day.
Scale: ~50 files/mo
Every extracted number links back to the page it came from, so checking one takes seconds instead of re-reading the file.
Runs: On the client's servers
From 13 runaway spreadsheets to one source of truth.
Thirteen spreadsheets holding roughly a million records were cleaned, matched up and merged into one database — with a daily backup and a restore path someone has actually tested.
Spreadsheets merged: 13
Call-centre data now reaches a report on its own — a report that had never existed before.
Refresh: Every 15 min
Paper contract signing became a 5-minute tablet flow that archives itself.
A recruiter used to spend about 20 minutes signing one driver contract on paper. It now takes five, and the contract files itself the moment it is signed — nothing printed.
Signing one contract: 20 min → 5
Went from the team's bottleneck to the tool they rely on.
Every order, quote and conversation with a client now sits on one timeline the team actually works from, and the ERP keeps itself in sync.
Tables merged: 15+
Hand-built quotes became a guided, on-brand builder.
A branded PDF now leaves the catalog in minutes, priced the same way on every single quote.
Margin levels: 3, applied identically
Production planned by feel became a measured, data-backed decision.
Stocking is decided on measured data instead of gut feel, once a single size profile turned out to cover almost every order.
Of 1,011 orders: 99.3%
Every extracted field carries its source document, page number, the exact snippet it came from and a confidence score. Verifying a figure means glancing at the quote — seconds, not re-reading the file.
Poor scans and ambiguous clauses get flagged rather than silently guessed. You see where the system was unsure, so attention goes exactly where it is needed.
The system produces a summary for review — it never sends anything to your client on its own. A person stays the last step, just with hours of reading and adding up removed from the first one.
The whole application, AI model included, runs on your own servers. No client document is ever sent anywhere outside, which is what makes the whole thing viable under GDPR.
Process first, automation second: I map how the work actually flows before writing a line of code, then build the tools and automations that run without you watching.
30 minutes online. You describe the problem, I ask questions. No pitch, no commitment.
Within 48 hours I send a brief and a fixed number. You decide — nothing starts until you approve.
Weekly check-ins. You test before anything goes live. Docs and handoff included.
Free 30-min call · Individual fixed quote within 48h · One engineer, no agency markup
Depends on the scope. Pulling data out of portals into a dashboard is considerably faster. A full process rebuild — like the transport case here, design through rollout — took six months. Either way, you get a concrete timeline with the fixed quote within 48 hours of our call.
One fixed number, not an hourly rate times an unknown — quoted within 48 hours of the call, once we know what we're actually building. No team margin, no office, no account manager baked into that number, which is why the same scope comes out cheaper than at an agency. Nothing starts until you approve it.
That's exactly why everything runs on your side by default — app, AI model, all of it, on your own servers or your own cloud, with logins limited to your company. No client document ever leaves. GDPR is handled by how the system is built, not by a clause in a contract.
Usually not — replacing it is rarely the point. Two cases here started with exactly that: a CRM the team had written off became the tool they rely on, and 13 spreadsheets at their row limits became one backed-up database. I work with what's already running wherever that's the cheaper answer.
You test everything before it hits production, and documentation plus handoff come with the build, not as an upsell. Since you've dealt with me directly from the first call, there's no team to meet later — you talk to the person who wrote the code.