Document extraction
Invoices, delivery notes, certificates, forms. Pulled into structured data automatically instead of somebody retyping them into a spreadsheet every Friday afternoon.
Free · No sign-up · Runs in your browser
The AI Lab. Take it for a drive.
Most agencies talk about AI. This page hands you three working tools and asks nothing in return. If they are useful, imagine what the versions built against your own data would do.
Straight talk
AI is not a feature. It is a way of removing work.
A chat box bolted onto a dashboard impresses nobody after week two. These are the places where it genuinely changes the economics of a business.
Triage and routing
Inbound enquiries, tickets, defect reports read and routed to the right person with a suggested reply already drafted. Response times drop, nobody has to sift.
Natural language reporting
"How many jobs ran over on the Bristol contract last quarter?" answered in seconds against live data, rather than a request that sits in someone's queue for a week.
Predictive scheduling
Job durations, travel time and skill matching predicted from your own history. Fewer overruns, tighter routes, less time spent on the phone rearranging the day.
Evidence checking
Certificates read on upload, expiry dates extracted, gaps flagged before the auditor finds them. The compliance file maintains itself.
Embedded copilots
An assistant that lives inside your own portal, knows your data and your rules, and can actually do things rather than just describe them.
Diagnostic
Five questions. An honest verdict.
Including, quite often, "not yet — fix the data first". We would rather tell you that now than eight weeks into a build.
AI Readiness Score
Scored on data, process, integration, culture and ownership
Generator
Marketing copy, generated on the spot.
A structured template engine running client-side. On a live project this same interface runs against the Claude API with your brand guidelines loaded, and the output is a different class entirely.
Copy Generator
Headlines, about copy, calls to action and meta tags
The business case
Does the build actually pay for itself?
Model it honestly. The tool assumes only sixty-eight percent of the identified admin gets automated, which is deliberately conservative — real projects usually beat it once the second workflow moves across.
ROI Modeller
Payback period and three-year net position
Under the bonnet
How we build AI features that survive contact with users.
Structured output, always
The model returns validated JSON against a schema, not free text somebody has to parse with a regular expression. If it fails validation, it retries. Users never see a malformed response.
Grounded in your data
Retrieval over your actual records, with citations back to the source row or document. If the model cannot find the answer it says so rather than inventing one that sounds right.
Cost controlled at the gate
Prompt caching, request queuing and a cheaper model for the routine calls. AI features should cost pounds a month, not hundreds, and we design for that from the first line.
A human can always override
Anything consequential is a suggestion until a person confirms it. The audit trail records who accepted what and when, which is what your insurer and your regulator will ask for.
Next step
These tools took a fortnight. Yours could too.
Everything on this page runs client-side with no backend. The versions built against your own data are considerably more useful, and not much harder to build.