Worked examples

What we build, and what we'd measure.

These are patterns, not client results. We're newly established, and we would sooner show you the shape of the work than dress an illustration up as a measured outcome. Each one sets out a problem, what we would build for it, and the figures we would baseline before starting.

Building services contractorIllustrative

Job sheets that stop being driven back to the office

The problem

Engineers fill in carbon-copy job sheets on site. The top copy comes back to the office in a van, sometimes days later, where administrators type them into the job system and chase the ones that never arrive. Invoicing waits on the paperwork, so cash comes in late.

What we'd build

A Power Apps job sheet on the engineers' phones, working offline in plant rooms with no signal, with photo capture and a customer signature. On sync, a flow writes the job to Dataverse, files the photos, and raises the invoice draft automatically when the job is marked complete.

What changes

Paperwork reaches the office the same day rather than the same week, and invoicing stops waiting on a van. Time released from rekeying could be used for credit control.

What we'd baseline first

Administrator hours a week spent typing job sheets
Days between a job being completed and an invoice raised
Proportion of jobs sheeted on the day they were done

Power AppsDataversePower AutomateSharePoint

Food wholesalerIllustrative

Supplier invoices read by machine, not by eye

The problem

Purchase invoices arrive by email from hundreds of suppliers, in every layout imaginable. One person opens each, reads it, matches it to a purchase order, and keys it in. Queries stack up behind the keying, and early-settlement discounts are missed because approvals run late.

What we'd build

An extraction model trained on your own eighteen months of invoices, pulling supplier, PO reference, net, VAT and due date. A flow performs the three-way match against the purchase order and goods receipt, routes anything over an agreed value to the budget holder in Teams, and drops anything the model isn't confident about into a human exception queue.

What changes

The exception queue becomes the job, not the whole postbag. Nobody opens an invoice the machine was sure about.

What we'd baseline first

Hours a week spent keying invoices
Proportion of invoices passing straight through untouched
Average days from invoice received to approved
Extraction accuracy per field, tracked monthly for drift

AI BuilderPower AutomateDataverseTeams

Recruitment agencyIllustrative

Compliance documents that chase themselves

The problem

Every placed contractor needs right-to-work evidence, insurance, qualifications and references before starting. Consultants chase each document by email, track it on a shared spreadsheet, and find out about a lapsed certificate when a client audit finds it.

What we'd build

A portal where contractors upload documents, with expiry dates read from the certificates themselves. Flows chase at 30, 14 and 3 days before expiry, escalate to the consultant, and block placement confirmation until the file is complete. A live compliance dashboard replaces the spreadsheet.

What changes

Chasing becomes something the system does overnight rather than something a consultant does between calls, and an expiry becomes a scheduled event instead of an audit finding.

What we'd baseline first

Consultant hours a week spent chasing documents
Expired documents found at client audit
Days from offer accepted to fully compliant placement

Power PagesAI BuilderPower AutomateDataverse

Domiciliary care providerIllustrative

Timesheets, mileage and payroll on one thread

The problem

Carers record visits on paper and post timesheets weekly. Payroll spends days each fortnight reconciling them against the rota, and mileage claims are estimated because nobody can reconstruct the route.

What we'd build

Visit check-in and check-out on the carer's phone, with mileage calculated from the actual visit sequence. Exceptions surface to the coordinator in real time instead of at the end of the fortnight: a visit missed, a long overrun, a shift not covered. Payroll receives a reconciled export.

What changes

Payroll preparation stops being a multi-day reconciliation, and a coordinator hears about a missed visit while they can still do something about it.

What we'd baseline first

Payroll preparation hours a fortnight
Variance between claimed and calculated mileage
Time between a visit being missed and the coordinator knowing

Power AppsPower AutomateDataversePower BI

How we measure

The baseline is written down before we build anything.

Nobody remembers how long a task used to take, and without a baseline every claim of success is a matter of opinion. During the audit we time the work with the people doing it and agree the figure in writing. The after figure is taken from the same measurement, at least sixty days post-launch, once people have stopped doing it the old way out of habit.

Hours returned means hours a named person no longer spends. It does not mean headcount removed. The point is to put those hours back into work the business already wanted doing.

When an engagement completes and those numbers exist, they will appear here as a case study with the client's written permission, and they will be that client's measured figures, not our estimates of them.

Next step

Yours would start with a ninety-minute call.

Bring one process. We'll walk it end to end and tell you what it costs you today.

Book a discovery call