Inbound lead responder
A form or enquiry arrives, gets qualified against your criteria, lands in the CRM with the right owner, and receives a useful first reply within minutes rather than the next morning.
Case studies
Two builds, in full.
Different industries, the same shape of problem: a person spending their week doing something a system should have been doing. Client names are withheld at their request; everything else is exactly as it happened.
Auto & import · Portugal
Auction sourcing, rebuilt as a weekly shortlist
- SectorVehicle import & resale
- TeamOwner-operated, small back office
- ProblemManual sourcing across auction platforms
- Build timeWeeks, not months
Custom scraperScheduled jobsScoring rulesEmail digestChange alerts
The situation
Sourcing was the business. Finding the right car at the right price, before someone else did, was the entire margin — and it was done by hand. Every week meant working through several auction platforms, filtering by model, year, mileage, damage category and reserve, opening dozens of listings, and keeping a mental note of which ones were worth a bid.
It ate the best part of the week. Worse, it scaled badly: the only way to look at more inventory was to spend more hours looking, and the listings that got missed were invisible — you never find out about the car you didn't see.
What we built
A sourcing engine that does the browsing. It sweeps the major auction platforms on a schedule, applies his actual buying criteria rather than the crude filters the sites offer, and scores every vehicle it finds against what he's historically been willing to pay.
What lands in his inbox is a single curated shortlist: the vehicles worth a look this week, ranked, each with the numbers that matter and a direct link to the bid page. The judgement stayed with him. The searching didn't.
What changed
Sourcing went from a multi-day slog to a ten-minute review with a coffee. The coverage went up rather than down — the engine doesn't get bored on page four of the results — and the decision he's actually good at is now the only part of the process he spends time on.
Days → 10 min
Weekly sourcing time
All platforms
Swept every cycle, not just the ones there was time for
1 shortlist
Ranked, scored, with direct bid links
The interesting part of this build isn't the scraping — it's the scoring. Encoding what the owner already knew about a good buy is what turned a list of results into a shortlist worth acting on.
Healthcare · child therapy clinic
A month-end close that closes itself
- SectorPaediatric therapy
- TeamMultiple therapists, shared admin
- ProblemSessions, attendance and billing in spreadsheets
- Build timeWeeks, not months
Internal platformRole-based accessAutomated billingScheduled closeAudit trail
The situation
Every therapist recorded their sessions their own way, and all of it ended up in spreadsheets. Attendance lived in one file, session notes in another, billing in a third, and the relationship between them existed mainly in one administrator's head.
Month-end was the reckoning. Two days of cross-checking who saw whom, for how long, and what should be invoiced — with the errors that inevitably survive that process turning up weeks later as a billing query from a parent.
What we built
A single platform where each therapist logs their own sessions as they happen, from wherever they are, with only the fields they actually need. One record, entered once, at the moment the information is accurate.
Everything downstream runs off that record. Attendance is a view of it. Billing is generated from it. The month-end close is a scheduled job rather than a person with a spreadsheet and a long afternoon. Each therapist sees their own caseload; administration sees the whole clinic.
What changed
The two-day close disappeared entirely — it now happens on its own, on schedule. Billing queries dropped because the numbers come from the person who was in the room rather than from a reconstruction. And the clinic can add a therapist without adding administrative load, which is the part that actually changes what the business can do.
2 days → 0
Month-end reconciliation
1 record
Entered once, at source, by the therapist
Automatic
Attendance, billing and close, on schedule
Nothing here required AI. It required someone to design the record properly and then build the thing. A good share of automation work looks like this — and we'd rather tell you that than sell you a model you don't need.
Repeatable builds
The ones we've built often enough to build fast.
These are patterns from our own library rather than named client stories — the builds we've done enough times that the design work is already finished, which makes them cheap to deploy and quick to prove.
Quote & proposal generator
Inputs in, branded document out — priced from your own rules, written in your own language, ready for a human to check rather than to write.
Monday morning report
Numbers gathered from every system they live in, reconciled, laid out the way the management team reads them, and delivered before anyone sits down.
Invoice & document intake
Documents read as they arrive, fields extracted and validated against what you already hold, exceptions flagged for a person, the rest filed automatically.
Two systems that never spoke
The unglamorous one, and often the most valuable: a reliable bridge between the two tools your team has been keeping in sync by hand every morning.
Team knowledge assistant
An assistant loaded with your policies, pricing and process documentation, so the same question doesn't get asked of the same senior person every week.
Your turn
What would yours look like?
Both of these started with a thirty-minute conversation about where the week was going. That call is free, and you'll leave it with two or three specific ideas whether or not we work together.
What both builds had in common
- One person carrying a process in their head
- Work that repeated on a predictable cycle
- Judgement that stayed human, effort that didn't
- A first version live in weeks, then improved