Wireframe · illustrative
The decision lands where you already work.
We diagnose the moment a decision is needed, prescribe the call — then it arrives as a card in the tools you already have. You approve or reject in one tap; a record is kept. No new system to learn, nothing in your stack to rebuild.
A rival cut their Pro plan £39 → £29
Undercuts your best-converting tier by £10.
Hold price. Offer a 3-month loyalty credit to at-risk accounts.
You already have Slack. We add the card — not another system.
A rival cut their Pro plan £39 → £29
Undercuts your best-converting tier by £10.
Hold price. Offer a 3-month loyalty credit to at-risk accounts.
No portal login. The decision travels in the email itself.
A rival cut their Pro plan £39 → £29
Undercuts your best-converting tier by £10.
Hold price. Offer a 3-month loyalty credit to at-risk accounts.
Try it — approve or reject, and watch the record appear →
Audit trail
- W27 · Monapproved · Dana P.Logged
- W26 · Thurejected · Dana P.Logged
- W26 · Tueapproved · Sam R.Logged
Every decision — who, what, when, and the score that justified it — kept in one place. Export it, or point it at a Google Sheet you already own.
It sits beside what you already run.
Nothing to rip out, nothing to migrate. One flow, whatever the work — the only thing that changes per client is how we plug into your system, and that’s small.
If you run the business
What you actually get
- Nothing acts unattended. Your AI keeps its speed, but consequential actions wait for a person — approved in seconds, not a dashboard nobody opens.
- A record you can stand behind. Every decision: who, what, when, and why. If anyone ever asks “who signed this off?”, the answer exists.
- It runs on your accounts. Your Slack, your email, your sheet. If we walked away tomorrow, it keeps working. No lock-in.
- No new system to learn. The decision lands where your team already works. Adoption is a tap, not a training programme.
- Priced to the workflow. A fixed-scope pilot on one workflow first — you only expand it if the first one earns it.
If you run the systems
Where it plugs in
- Thin adapters, one core. An adapter turns whatever your system emits into a normalised item; the check → gate → log core never changes. New source = a small adapter, not a rebuild.
- Read-mostly, least privilege. We read what your system produces and write only the approved result back — narrow scopes, nothing more.
- Your infrastructure, your keys. Runs on your accounts and credentials. We hand over the config and the rubric, documented — no WFK-hosted black box in the critical path.
- No data retention by us. Items are processed in flight; the only thing persisted is the record — in your own store.
- Signed, single-use actions. Slack requests are signature-verified; email approve-links are signed and expire. A decision can’t be replayed or forged.
How it connects — whatever your system already does
The audit destination is just as flexible — a CSV, a Google Sheet, a database table, or a git commit — one line in the config.
Great fit
- Your AI drafts something a person should approve before it takes effect
- Discrete decisions: quotes, replies, refunds, publishes, outreach, data / config changes
- You need a defensible record of who decided what, and why
Not the right tool
- Fully autonomous flows where no human should be in the loop
- Ultra-high-frequency / low-latency decisions (thousands a second)
- Continuous processes with no discrete “decision point” to gate