1. The exception has nowhere to go
Every rule meets a case it was not designed for: the enquiry with no email address, the duplicate, the customer who is also a supplier. If the automation simply stops, the record sits in limbo and someone eventually notices weeks later. Exceptions need a queue, an owner and a visible count.
2. Duplicate records
Automation multiplies whatever your data quality already is. Two records for the same company means two owners, two follow-up sequences and a customer receiving conflicting messages. Matching and merging rules are not glamorous work, but they decide whether anything downstream can be trusted.
3. The rule encodes a process nobody agreed
Automations are often written from one person’s description of how things should work. The rest of the team never agreed, so they work around it, and the CRM data stops reflecting reality. Getting the process agreed on paper before it is built is faster than rebuilding it twice.
- Who owns a lead at each stage, explicitly
- What counts as qualified, in words everyone accepts
- What happens when nothing happens for a fortnight
- Which field is authoritative when two disagree
4. Nobody owns it after launch
Processes change. Pricing changes, a team is reorganised, a form gains a field. Without an owner, the automation quietly drifts out of alignment with the business and the failure looks like a technology problem.
An automation without an owner is a piece of technical debt that sends emails.
Rebuilding trust
If confidence has already gone, do not start by fixing every rule. Pick the process people complain about most, map it properly, rebuild that one with exception handling and a visible measure, and let the improvement do the persuading.
