How Imagine Growth helped a lean B2B consultancy turn HubSpot, Claude and their own call recordings into a sales system that corrects itself, recovering over £100,000 of unlogged pipeline in the first sweep.
|
At a glance Client: A lean B2B consultancy winning work through a marketplace and referrals. Stack: HubSpot, Claude, Fathom, Google Workspace. Headline result: Over £100,000 of unlogged pipeline recovered in the first automated sweep, proposals cut from hours to minutes, and a pricing blind spot the team had never spotted. |
The client had a problem that is familiar to almost every small consultancy. The selling was working. The sales system behind it was not.
Work was coming in through a marketplace and through referrals. But the engine behind that work, the proposals, the CRM, the follow-ups, the pricing decisions, was running on memory, goodwill and whatever time was left at the end of the day.
That is fine until it is not. As volume grows, the cracks start to cost real money.
Imagine Growth partnered with the consultancy to turn a manual, memory-based sales process into a connected system, built entirely on tools they already owned.
The consultancy had reached the point where its sales process depended too much on people remembering to do things, and not enough on a system that did them automatically.
Before the project, the team faced five recurring problems:
• Proposals took hours each, and the quality varied depending on how much time was free that day.
• Marketplace leads needed fast, well-priced replies, written from scratch every single time.
• The CRM was behind reality. Deals were missing, closed deals had no loss reasons, amounts sat in the wrong currencies, and at least one won project had never been logged at all.
• Nobody could say why deals were lost. The answers existed, buried in email threads and call recordings, but no one had time to dig them out.
• Follow-ups slipped. Deals went quiet, and stayed quiet.
None of these were tooling problems. The consultancy already had HubSpot, Gmail and recorded calls. The problem was that nothing connected them, and nothing kept them honest.
The real issue was not a lack of software. It was a lack of system.
Every part of the sales process depended on a person remembering to act. Writing the proposal. Updating the deal. Logging the loss reason. Chasing the follow-up. When the team was busy, which was always, those were the first tasks to slip.
A growing consultancy cannot scale on memory. It needs a sales process that captures its own data, learns from its own outcomes, and prompts the next action without being asked.
So the goal was clear: turn a manual sales process into a connected system that runs quietly in the background and gets sharper over time.
Imagine Growth connected Claude to the consultancy’s HubSpot, Gmail and Fathom call recordings, then built four things on top of them.
We started with the most painful, most repeated task: writing proposals.
We analyzed 118 competitor proposals from the client’s marketplace alongside 17 of their own, then distilled the patterns that win into a Claude Skill. Think of it as a reusable instruction pack that carries the consultancy’s proposal structure, brand, approved testimonials, case studies and pricing rules.
Now the process is simple. A call recording goes in. A branded, priced, quality-checked proposal comes out in minutes. The same skill also writes marketplace lead replies designed to do one job: book the discovery call.
The business impact: proposals stopped being a time tax. Quality stopped depending on how busy the week was. And reply speed, which matters most on marketplace leads, went from hours to minutes.
A sales system is only as good as the data underneath it. The consultancy’s CRM was behind reality, so we fixed it at the source.
Claude read the email threads and call records behind every closed deal and wrote plain-English win and loss reasons onto each one. It corrected wrong amounts and currencies. It tagged every deal by lead source and engagement type.
Then it swept 90 days of Gmail and found 11 opportunities that had never been logged, worth over £100,000 between them. One of those was a project that had been won and delivered without ever touching the CRM.
The business impact: leadership moved from guessing to knowing. Every closed deal now carries a real reason, mined from real conversations rather than assumed after the fact.
Cleaning a CRM once is useful. Keeping it clean is where the value compounds.
Three scheduled Claude tasks now run unattended:
• A fortnightly hygiene sweep that catches new opportunities from email and calls, writes chase-up drafts into Gmail for deals going cold, and closes out the dead ones.
• A fortnightly win/loss report with pipeline stats, so the team always knows where things stand.
• A monthly review that mines recent outcomes for lessons.
The business impact: the sales admin that used to depend on someone finding the time now happens on its own. Follow-ups are drafted before deals go cold, each one referencing the actual last conversation.
This is the part that turns a clean CRM into a system that learns.
Each monthly review writes what it learns back into the proposal skill. On its very first run, it surfaced something the team had never seen.
Deals priced between £15,000 and £30,000 were winning just 25% of the time. Below that band, the win rate was 63%. Above it, 80%. That middle band was quietly costing the business.
The pricing rules now route anything in that range into a phased offer with a smaller committed first step. Every future proposal is priced by that evidence automatically, without anyone having to remember the insight.
The business impact: a pricing blind spot that had been losing winnable deals is now corrected on every proposal, by default.
Within the first sweep and the weeks that followed, the system delivered measurable outcomes:
|
£100,000+ pipeline recovered in the first sweep |
11 lost opportunities reclaimed in HubSpot |
Hours → minutes to produce a branded proposal |
100% of closed deals with win/loss reasons |
• Over £100,000 of pipeline recovered. 11 unlogged opportunities created in HubSpot, including a won project that had never been recorded.
• Proposals in minutes, not hours. From call recording to a branded document, with pricing, proof and a follow-up plan included.
• Every closed deal explained. 100% of closed deals now carry a written win or loss reason, mined from real emails and calls rather than guessed.
• A pricing blind spot found and fixed. The £15,000 to £30,000 band is now automatically proposed as phased work.
• Follow-ups that write themselves. Chase-up drafts appear in Gmail before deals go cold, each one referencing the real last conversation.
Why this matters
Many growing consultancies hit the same wall. The selling works, but the system behind it does not.
At first, running the sales process on memory feels flexible. Proposals get written when there is time. The CRM gets updated when someone remembers. Follow-ups happen when a deal comes to mind.
But as the business grows, that approach starts to leak. Pipeline goes unlogged. Deals go cold. Pricing decisions get made on instinct rather than evidence. And nobody has the time to find out why.
The fix is not more software. It is a system that captures its own data, keeps itself honest, and learns from every outcome.
None of this needed custom software.
HubSpot held the data. Fathom held the conversations. Gmail held the history. Claude connected them, learned the patterns, and now keeps the whole system honest.
The result is a sales process that largely runs itself. The admin happens in the background. The follow-ups draft themselves. And every deal, won or lost, makes the next proposal a little sharper.
That is what a sales system that learns looks like in practice.