Follow-up that runs from the CRM, not from memory

Follow-up runs reliably when the CRM decides what happens next, not when a person remembers to. That means every lead sits in a named stage, every stage has a rule for what fires and when, and a human only intervenes to do the parts that need a human — the call, the judgement, the reply. Memory-based follow-up isn't a lesser version of the same thing; it's a different thing that degrades every time the firm gets busy, which is precisely when the leads are worth the most.

What's actually wrong with follow-up from memory?

It correlates with workload in exactly the wrong direction. When the pipeline is quiet, people chase diligently. When delivery is heavy — which is when the invoices are being earned and the next quarter's pipeline should be built — chasing is the first thing dropped. The result is the feast-and-famine revenue curve most service firms treat as weather rather than as a design flaw.

Memory also fails silently. A missed task in a CRM shows up as an overdue item; a missed follow-up in someone's head shows up as nothing at all. Most firms stop at two touches while deals typically need five or more, and it's rarely a decision — nobody chose to stop, they just forgot, felt awkward, or lost the thread. The 90-Day Follow-Up Framework covers the full structure; this piece is about where that structure has to live.

What does CRM-driven follow-up look like mechanically?

It's a set of when-then rules attached to pipeline stages. When a new enquiry lands, then the CRM creates the contact, stamps the source, and starts the clock. When the lead is scored — fit and intent on entry — then the score routes them: high scorers trigger a call task due within the hour plus a sequence; low scorers enter a lighter automated track. When a lead replies, then all pending automated touches cancel and a task lands with the owner. When a task sits overdue for 24 hours, then it escalates — visibly. When a deal sits in one stage beyond its allowed age, then it flags for review rather than quietly fossilising.

None of this requires expensive software. It requires the rules to be written down, agreed, and configured once — after which the system asks nothing of anyone's memory, only of their execution.

Which parts should stay human?

More than automation vendors would have you believe. Automate the remembering, the timing, the routing, and the plumbing — never the judgement. A sensible split:

• CRM does: create tasks, send the scheduled nurture emails, cancel sequences on reply, stamp dates, escalate the overdue.
• Humans do: phone calls, replies to replies, proposals, and any message that benefits from actually having read the lead's last email.

The failure mode on one side is the firm where nothing happens unless the founder pushes it. The failure mode on the other is the lead who receives "just circling back" from a robot four times and correctly concludes nobody there knows they exist. The CRM's job is to make sure the human shows up on time — not to impersonate them.

How do you keep the system honest?

A follow-up system is a loop, and loops need instrumentation: you change something, the numbers move, you adjust. Watch three numbers weekly — median time-to-first-touch, the count of leads with no next action scheduled, and overdue tasks per owner. The second number is the killer: any lead without a scheduled next action is, operationally, dead — the firm has just not admitted it yet. A healthy system holds that number at zero by rule: no deal record may be saved without a next action and a date.

This is also why speed metrics belong on a dashboard rather than in anyone's impressions. The industry rule of thumb that contact rates fall roughly eightfold after five minutes is only actionable if you can see your own median response time moving. Instrumented this way, follow-up becomes one of the clearer examples of a feedback loop you can actually tune — most firms simply never close the loop.

Where do these builds go wrong?

Three places, in my experience. First, over-configuration on day one: forty automations nobody understands, half firing at the wrong moments; start with the five rules above and add only what a real failure justifies. Second, no single owner: when everyone owns follow-up, nobody does — every lead needs exactly one name on it, assigned automatically, which is a routing problem in its own right (covered in enquiry to assigned owner in seconds). Third, the founder exempting themselves: the moment the busiest person in the firm keeps their deals "in their head", the system stops being the source of truth and everyone quietly reverts.

Build it small, wire it to the CRM, watch the three numbers. The leads you already have stop dying of neglect — which is cheaper than buying new ones.

---

Next step: the Growth System Audit — £450, seven days, credited against any build — maps where your growth system leaks and what to build first.