A founder spends three weeks in her first year setting up an elaborate system. It connects her forms to her records to her invoicing to her calendar.
She has eleven customers. The system handles a volume she does not have, encodes a process she is still changing every fortnight, and needs rebuilding twice as the business finds its shape.
The one thing she does not automate is answering enquiries, which happens forty times a month, interrupts everything and is identical every time.
Automate: the questions asked repeatedly
Price, availability, coverage, process and what happens next. High volume, no judgement, and they arrive at the worst moments.
Automate: the enquiry capture
Every enquiry recorded consistently with its source and outcome, which is the habit described in handling your first enquiries properly.
Automate: the follow up
A single timed message after a quotation, which founders almost never do manually and which recovers real work.
Automate: the appointment offering
Letting somebody choose a time rather than exchanging six messages about availability.
Do not automate: anything requiring judgement
Pricing unusual work, difficult conversations and decisions about which clients to take. These are the founder's job and doing them builds the business.
Do not automate: a process still changing
Anything being redesigned monthly will have to be rebuilt every time, which is why early operational automation is usually wasted effort.
Do not automate: low frequency tasks
Something happening twice a month takes longer to automate than to do, for years.
Do not automate: the relationship
Early customers are the source of everything the business learns, and the founder should be in those conversations personally.
Test: how often does it happen
Daily is a candidate, monthly is not. Frequency is the single best indicator of what is worth the effort.
Test: does the answer change
If the answer is the same every time, it should be written down once regardless of whether anything automates it.
Test: does it interrupt paid work
The strongest argument of all, since the cost is not the task but everything it displaces.
Test: would you hire somebody to do it
If the answer is obviously no because it is too small, it is a candidate. If it is yes, it probably needs a person rather than a system.
The order that actually works
Write the answers down, put them where customers find them, let something answer the repetitive ones, and leave everything else until the process has stopped changing.
Test: can you describe the process in one paragraph
If not, it is not stable enough to automate. A process a founder cannot state plainly is a process still being designed, and building around it locks in a shape that will be wrong within two months.
The trap worth naming
Building systems feels like progress and is considerably more comfortable than selling. A founder three weeks into tooling has usually spent three weeks avoiding the thing the business actually needed.
Automate the answering and leave the rest
An assistant handling the forty enquiries a month gives a founder back the time she was about to spend building something her eleven customers do not need. The broader question is covered in deciding what to automate in year one, the practical concerns in first year trading questions new companies ask, and the method in batching automating and delegating with ai assistants. See it working on the demo, compare on pricing, and the sector view sits under startups.