AI does the work. Your people approve it.
WFK designs the approval layer — the screen where an AI’s work is checked, approved and put on the record before it counts. That’s what makes AI safe to trust with real customers and real money. Whether you’re shipping an AI product or bringing AI into your business: fixed scope, a fixed price, senior hands, and work you own outright.
- find where AI fits your people and your work
- what's costing you — and where the gains are
- every issue graded & prioritised
- evidence, not opinions
- senior hands, start to finish
- a plain-English findings report
- a prioritised fix roadmap
- delivered in weeks, not quarters
Which one sounds like you?
One method — find where AI genuinely helps, and keep a person in charge of it. Two very different starting points.
You build an AI product
Your model works, but users don’t trust it yet. The audit finds where trust breaks, then hands you a ranked fix map and one redesigned flow — ten days, fixed price.
See the auditYou run a business
Quoting, scheduling, suppliers, admin — you know AI could take some of the weight, but not where. The Business Diagnostic finds where the hours and margin leak, what to fix first, and what to leave human. Report and walkthrough video, delivered in days.
See the diagnosticThe model works.
But no one can tell when it's wrong.
No one can steer it back on track.
No one can undo the damage.
That's not a model problem. It's an architecture problem.
This is what “human in the loop” actually looks like.
An AI agent does the work. A person approves before anything happens. Every step is logged. Twenty seconds, no sound needed.
Silent · loops · ~13 seconds
Two ways in
Two ways to work together, depending on how much help you want — from a fixed-scope audit to hands-on design. All delivered by the architect: no account managers, no juniors, no theatre.
The Audit
A 10-day health-check of your AI product. We find where people stop trusting it, where it needs a human to check its work, and where users get stuck — then hand you a ranked list of fixes, one key flow redesigned, and a recorded walkthrough.
How the audit worksArchitecture
Hands-on design for AI that acts on its own. We work out how it shows its reasoning, when it should ask permission, and how people stay in control. For teams shipping AI that acts — and has to earn the right to. Scoped engagements, three to four weeks.
Start a conversationRun a business? Find where AI actually fits.
Quoting, scheduling, suppliers, admin — you sense AI could hand you back hours, but not where, and not without breaking the service your customers stay for. The Business Diagnostic finds where the time and margin leak, and what to leave human. And there may be public money to fund the fix.
What a typical audit costs, how long it takes, and how fast we get back to you.
You stop debating opinions. You start shipping decisions.
An audit is only worth what it changes about your next sprint. Everything WFK hands over is built to be acted on the day it arrives.
- A ranked fix map — your team ships in priority order, not committee order
- The top trust-breaking flow redesigned, not just flagged
- Findings tied to churn, support load and activation — not aesthetics
- Everything handed over in ten days. No retainer creep.
Evidence, not logos
No client logos. No testimonials yet. Just work you can use. We think that's the right order.
How WFK runs
The studio’s own operating platform — site, CRM, inbox, boards, email and AI agents behind human approval gates — written up as a working case study. The system you’re using right now is the exhibit.
Read the case studyThe 5-minute Readiness Check
A free, honest read on your UX and AI maturity. Twenty questions, an indicative scorecard, your risk flags — the same lenses the paid audit uses, self-serve.
Run it nowThe Oversight Console
An interactive study of an AI agent doing real multi-step work — with plan visibility, approval gates and a full audit trail. What "human in the loop" should actually feel like.
Notify me when liveHow we think
Long-form, practical, opinionated — the thinking behind the work, on trust, oversight and recoverable AI.
Trust is architecture, not a feature
Why trustworthy AI products need product architecture, not just better AI copy, explainability labels, or governance theatre.
Approval fatigue: why human-in-the-loop fails at scale
Human-in-the-loop AI fails when approval is overused, under-contextualized, and poorly placed. Here is how to design oversight that survives real workloads.
How to audit an AI product in 10 days: the five lenses
A five-lens AI product audit method for finding trust, oversight, and recoverability problems in AI features before they become adoption problems.
AI that acts without earning trust is a liability with a UI. We design AI that shows its reasoning, asks before it acts, and fails in ways humans can recover from. If that sounds slower — it's the only version users keep.
One engagement at a time.
The intake does the first conversation’s work.