Imagine Growth
Home/Built for revenue/Self-improving sales system
Built for revenue · Case study · AI

A sales system that improves itself, and recovered £100k+ of pipeline.

Four connected HubSpot and Claude systems that write proposals in minutes, keep the CRM honest, chase what goes cold and learn from every win and loss.

Proposal engine
Call recording
Branded proposal
Draft generated4 min
118 competitor proposals 17 of our own Priced & branded
£100k+
of pipeline recovered in the first sweep
11
previously unlogged opportunities created
Hours → mins
time to a branded, priced proposal
100%
of closed deals now carry a win or loss reason
The challenge

Sales depended on people remembering things, not on a system.

Five problems kept recurring: proposals took hours with inconsistent quality, marketplace leads needed fast custom-priced replies every time, and the CRM lagged reality with missing deals, wrong currencies and closed projects never logged.

On top of that there was no visibility into why deals were lost, with the reasons buried in emails and call recordings, and follow-ups fell through the cracks. The core issue was structural: nothing caught the gaps automatically.

What we built

Four interconnected systems across HubSpot and Claude.

This is the engine we run on our own business. We built it, so we know exactly what it takes.

  • A proposal engine. A Claude skill trained on 118 competitor proposals and 17 of our own, generating branded, priced proposals from a call recording in minutes.
  • CRM data integrity. Claude reads email threads and call records to add win and loss reasons to closed deals, and corrects amounts and currencies.
  • Automated hygiene tasks. Scheduled Claude tasks for opportunity discovery, cold-deal chase-up drafts and reporting, so nothing goes cold unnoticed.
  • A continuous learning loop. Monthly reviews feed insight back into the proposal skill, so the system gets sharper over time.
The blind spot it found

A dead zone in the middle of the pricing, hiding in plain sight.

With 100% win and loss coverage, the pattern was obvious once the data was clean. Deals in the £15k to £30k band were winning at just 25%, well below the bands either side. They are now automatically routed as phased offers.

Win rate by price band
Cleaned from 100% win/loss coverage
63%
25%
80%
Below £15k £15k to £30k Above £30k
The middle band now routes automatically as phased offers, so mid-size deals stop stalling in the dead zone.
Why this study is anonymised

This is our own sales engine.

The consultancy in this study is us. We built this system on our own HubSpot and Claude to prove it before we sold it, so there is no external client to name and no third-party quote to publish. The numbers are ours and they are real: £100k+ recovered, 11 unlogged deals found, 118 competitor and 17 of our own proposals in the model, and 100% win and loss coverage across closed deals.

If your sales runs on people remembering tasks, an AI sales system catches what they miss and gets better every month.

Talk to Rikki or Emma about building a self-improving sales system on your HubSpot.

Speak to us