A paid audit of one process, with the fee credited in full to the build.
£750 to £1,250 to map what you do by hand, quote the fix, and prove it on one slice of the process before you commit to anything bigger. Plus candidate shortlisting for recruitment agencies, and workflow and AI automation built end to end.
An Automation Opportunity Audit is a paid, fixed-scope look at one process that is costing you hours. You get a written map of what happens now, a fixed quote for automating it, and a small working proof on a slice of it, so you can see it before you commit to anything bigger.
What actually happens, step by step: the tools involved, who does what, and roughly how many hours a week it costs.
A written breakdown of what can be automated and what cannot, a fixed price for the build, and a small working proof on one slice of the process.
Go ahead, and the full audit fee comes off a build starting at £3,000. Decide not to, and you keep the map.
Describe what is eating your time. I will reply with whether it can be automated, what the audit would cost, and what I would need to get started.
Send it ↓Bullhorn's own research says 85% of the candidates in an agency's database go uncontacted for a year. Not for want of trying: nobody has the hours to read thousands of CVs against every new spec. This does, and it shows its working on every line.
A job description as it is, and a folder of PDFs and Word files or a CSV out of your ATS. Nothing migrates. You keep Bullhorn, Vincere, JobAdder, whatever you run.
The spec is read and each requirement proposed with the sentence it came from. You check the list, delete what a CV cannot answer, and confirm. The list is frozen from then on.
Every candidate is ordered by relevance in seconds. The top N are judged criterion by criterion. The rest are reported as not judged, never as rejected.
Each "met" carries the line of the candidate's own CV it rests on. Each "unmet" carries a reason. An invented claim is a structural impossibility, not a policy.
For the candidates you name, a short note you can send to the client: only evidenced points, no adjectives, nothing about protected characteristics, gaps stated plainly by code.
Send one live job description and a CV export, and I will map what the matching would look like and quote it, the same way as the audit above. If it is useful, it becomes a monthly fee for a set number of live roles.
Send a spec ↓Most of the work is n8n and Make.com wired into Claude and the OpenAI API: webhooks in, data cleaned and de-duplicated, the right API called with retry and rate-limit handling, and a sensible branch when a call fails. Anyone can demo a happy path. The dedupe, the error branches, the idempotency, the scheduled runs: those decide whether it is still standing in three months.
Lead capture and routing, CRM sync, data enrichment, scheduled pipelines. Built in n8n (cloud or self-hosted) and Make.com, plugged into what you already run.
Claude and the OpenAI API doing real work: enrichment, classification, structured extraction over thousands of records. Pipelines with token budgets, confidence scoring and output you can validate.
Apify and custom scrapers at six-figure record counts, with the deduplication and schema work that turns a scrape into something searchable rather than a pile of HTML.
When a workflow is not enough: Next.js, Supabase, Stripe, Vercel, Python. Whole products shipped solo, front end to database.
The system behind the shortlisting above. Criteria are frozen and identified before anything is judged; every match carries a verbatim quote and an out-of-list claim is rejected by the schema before it can be written; retrieval, matching and summaries each refuse rather than repair.
700+ testsEvery guarantee has a test that would fail without itAdrian's own product, not a client build: an AI-guided reflective-practice tool for UK clinicians. A structured interview asks the questions a good supervisor would, then produces an audit-ready Word document mapped to regulator standards: NMC, HCPC, GMC, GDC, GOC.
5 regulatorsStandards mapped, audit-ready outputAdrian's own in-house pipeline, not a client project: it turns still photographs into cinematic vertical video with no manual editing. AI motion, AI voiceover, captions burned in, the whole thing assembled with ffmpeg and rendered unattended.
Zero manual editsStills in, finished video outTwenty years in the NHS. When a workflow has to withstand scrutiny, I already know where the risk sits.
Cleddau Automation is founder-led and built solo by Adrian Swales: a frontline paramedic with the Welsh Ambulance Service since 2004, a union representative, and for two years, most evenings and weekends, a self-taught engineer.
That combination is the point. Twenty years in a regulated, high-stakes service is twenty years of governance, record-keeping and audit trails, and of seeing the real cost when a record says something that is not true.
You get the person who built it, not an account manager. Give me a well-defined problem and I will go deep and ship it.
Tell me what the job is. I will reply with whether I can automate it, what it would cost, and how long it would take. If I cannot help, I will say so.
Or email adrian@cleddau-automations.com, or find me on LinkedIn.