The five CRM automations worth building first

The first five CRM automations worth building are, in order: instant lead routing, stale-deal flags, follow-up tasks on stage change, automatic activity capture, and the self-compiling weekly report. Each one replaces a human remembering something, each is buildable in days with a mid-market CRM and a lightweight automation layer, and together they cover most of the gap between a CRM that stores data and a CRM that runs the sales process.

Why these five, and why this order?

Because they attack the failure points in the order money leaks through them. Leads die first at response time, then at follow-up, then at record-keeping, and finally at reporting — so the build order follows the lead's own journey. The end state is the one described in The MD Dashboard Blueprint: a commercial machine whose numbers compile themselves, which is only possible when the underlying actions are systematised too.

The selection principle matters more than the list. A CRM automation earns early build priority when it meets three tests: the trigger is unambiguous, the action would otherwise depend on someone's memory, and the cost of the miss is a deal rather than a typo. All five pass; most of what automation vendors demo does not.

What do the five actually do?

1. Instant lead routing. When an enquiry arrives — form, email, referral note — then a contact and deal are created, an owner is assigned, and a first-response task fires within minutes. Contact rates drop roughly eight-fold after five minutes, as an industry rule of thumb, so this automation has the fastest payback of anything in the stack.
2. Stale-deal flags. When a deal sits in a stage past its normal age with no activity and no scheduled next step, then its owner is flagged; when it stays stale past a grace period, then it is queued to close as lost with a reason code. This is the front line of the hygiene system detailed in data-hygiene robots.
3. Follow-up tasks on stage change. When a deal enters a stage, then the standard next action for that stage is created with a due date — proposal sent begets a follow-up call in three days, and so on. Most firms stop at two follow-up touches while deals typically need five or more; this automation is how the CRM, rather than willpower, closes that gap.
4. Automatic activity capture. When a call happens or a meeting ends, then its notes and outcomes land on the record without anyone typing them in later — the mechanics are covered in meeting notes into the CRM without typing. Every other automation depends on this one being honest, because triggers can only read what was captured.
5. The self-compiling report. When Monday arrives, then the week's numbers — deals created, won, stalled, plus exceptions — assemble and deliver themselves, per the weekly report that writes itself. This is the automation that makes the other four inspectable.

What do they need to run reliably?

Three unglamorous preconditions. First, stage definitions that are verifiable events, because every trigger above reads stages — automating on top of vague stages produces confident nonsense. Second, mandatory fields at deal creation (value, source, owner), or the routing and reporting automations run on blanks. Third, a named owner for the automations themselves: someone who reviews the exception digests and adjusts thresholds quarterly. Automation without an owner decays into ignored notifications within months.

Tooling is the easy part. Most of the five can be expressed natively in a mid-market CRM's workflow builder; the remainder — notes capture, formatted reports — typically takes a small automation layer alongside. Where AI components genuinely help (transcribing calls, summarising notes) and where they are dressing, is ground covered honestly in AI automation for B2B: what actually works.

What should you deliberately not automate yet?

Anything requiring judgement about people. Auto-merging duplicate records, auto-closing deals without human review, auto-sending "just checking in" emails on a timer — each saves minutes and occasionally costs a relationship or a data trail. The working boundary: machines detect and propose; humans decide and send. Also skip lead scoring in month one; with a small firm's volumes, a simple source-plus-behaviour flag outperforms a scoring model you cannot yet calibrate.

How do you know the five are paying back?

Measure them the way you would measure a hire. Response time to new enquiries, follow-up touches per deal, percentage of deals with complete records, hours no longer spent compiling — each has a before and after. The composite effect shows up in one number worth watching above the rest: cost per client, not cost per lead, because automations that convert existing enquiries better reduce it without any new marketing spend — the distinction unpacked in cost per lead is vanity; cost per client is sanity. When the five run, a firm typically finds it was never short of leads; it was short of a machine that refused to drop them.

---

Next step: the Growth System Audit — £450, seven days, credited against any build — maps where your growth system leaks, including which of these five automations you are missing, and what to build first.