Response times cut from 48 hours to under two, without new tools.

The firm was doing well enough to have a problem: enquiry volume had roughly doubled over two years and the qualification burden had landed on junior consultants. Response times had drifted to 48 hours on a good day, and higher-value leads were quietly going elsewhere.

Leadership had considered hiring, buying a bigger CRM and 'doing something with AI', in roughly that order. None of those options were bad — they just weren't the smallest useful move.

The engagement started with a short discovery to define what a good lead actually looked like inside their CRM — not in the marketing collateral. That produced a scoring rubric the sales lead was happy to sign off in writing.

The build sat entirely inside their existing CRM: incoming enquiries were enriched, an LLM produced a first-pass score and a short internal summary, and a human always approved the outbound reply. No new customer-facing tools. No new logins. Nothing left the CRM.

Rollout was staged: two weeks running silently alongside the human process to validate scores, then progressive handover with clear override rules and weekly sanity checks.

The interesting outcome wasn't speed — it was calm. The team stopped context-switching between qualification and delivery every twenty minutes, so the delivery work improved too. That was harder to put on a slide than the response-time number, but it's the change the partners noticed most.

If we ran this again, I'd push harder on the initial rubric. Getting that right up front cut days of tuning downstream.