Your systems are lying to you. Quietly.
Calls get recorded, deals get staged, candidates get noted, and none of it agrees. We connect read-only to what you already run, backfill 30 to 90 days of history, and show you exactly where deals, candidates and follow-ups are leaking between your tools, and what each leak is worth at your placement fee. Every finding is computed from your own data, or it doesn’t go in the report.
- A deal scored dead that was alive. A weekly review wrote a prospect off as gone quiet. The call recordings showed he had personally booked a follow-up twelve days earlier. No system held the date. The call went ahead as booked: a full placement fee walking back in the door.
- A CRM carrying a lost deal as live. The pipeline showed Proposal Stage. The recordings and the owner's own account showed the deal had been lost weeks before. The forecast was fiction and nobody had lied, the systems just never spoke.
- One prospect, two records. The same person existed twice under different names. One record was being aged toward write-off while the other sat mid-pipeline. Merged in an afternoon once the audit surfaced it.
- Terms drifting between calls. The same commercial guarantee was described one way to one prospect and a materially cheaper way to another, eight days apart. Caught by reading the calls side by side, before either signed.
All four happened in one pipeline in one week: ours. We built this to audit ourselves, it caught things we believed we were on top of, and that is the standard your report is held to.
Your calls on the left, your CRM on the right. Select any row to see the evidence behind it.
The report, live
Not a mockup and not sample data: this is our own pipeline being audited, anonymised, regenerated weekly. The KEPT row is a deal a weekly review once wrote off as dead.
Open the full report in its own tab · names replaced with neutral labels, calendar dates replaced with relative ages, no sectors or locations, counts and durations real. If a number in it looks bad, good: it was true, and it is why we built this.
The numbers behind that claim, ours
We ran the audit on ourselves before offering it to anyone. This is what 45 days of our own history surfaced, in a pipeline we believed we were on top of.
The arithmetic that matters: at a typical £20,000 to £30,000 placement fee, one deal rescued from a false “gone quiet” pays for this audit forty to sixty times over. We publish no customer savings figures yet because no customer has completed an audit yet; the founding cohort is priced at £99 a month partly in exchange for a permissioned, attributable result we can put here. The first real number replaces this sentence.
One desk, six stages, six places to leak
The audit scores every stage of how a desk actually runs and names what is leaking at each one. Where your data shows a gap, we recommend the specific agent that closes it, priced monthly, and nothing your data didn’t ask for.
Select a stage to see what the audit checks there, and what it found in ours.
The leak is industry-wide, and measured
Not our numbers, the industry’s, each one traced to its primary source. Your report shows yours.
Two figures you may have seen elsewhere are deliberately absent: the “£6,125 average cost per vacancy” often credited to the CIPD is not in their 2024 report, and a widely-quoted data quality cost could not be traced to its primary source. If we would not accept an unsourced number in your audit, we do not put one on our own page.
How the agents work, and why that way
Three rules govern every agent we build. They exist because each one is the fix for a failure we have watched happen.
Watch-only before anything acts
Every agent starts read-only: it reads, reconciles and reports, and cannot send, edit or delete. Acting agents come later, per agent, behind your named approval. Blast radius is a design input, not an apology.
Verbatim beats summary
Agents quote recordings and CRM fields, never AI meeting summaries. We watched an auto-summary turn a misspoken city on a sales call into a written targeting instruction. Our tooling cannot fetch summaries at all, so the rule cannot be forgotten under pressure.
Absence is a question, not a verdict
No recorded call does not mean a dead deal: the meeting may have moved to email, or the recorder never joined. Agents surface the gap and ask; they never write an outcome your data cannot prove. The same discipline resolves one person hiding under two names.
Seen enough? Start free.
Everything above was computed from two connected systems. Tell us yours and the next report like that is about your desk. Free for 14 days, then £99 a month, and no card to start.
Tick your stack
Systems marked connects now are read in the audit today. Ticking anything else moves it up our build list, and we say so rather than pretend.
Works with what you already run
Marked connectors are read in the audit today. The rest are built in the order agencies actually tick them on the form above, and we say which is which rather than showing you a logo wall of promises. No API? The audit runs on exports, any system that can produce a CSV qualifies.
Fireflies
Attio
Bullhorn
Vincere
JobAdder
Loxo
HubSpot
Calendly
tl;dv
Google Calendar
Outlook
Microsoft Teams
LinkedIn