What to Automate First
Beginner friendly · ~7 minute read · Updated August 4, 2026
The most common reason a first automation project disappoints is not the technology. It is the choice of workflow. Businesses tend to start with whichever process annoys them most, which is usually the messiest, least defined, highest-judgement work in the building — exactly the wrong place to begin.
This is a selection method rather than a sales pitch. Audit the processes you actually run, score them on four factors, check the shortlist against a few known-good and known-bad patterns, decide where a human must still approve, and then run one narrow rollout over 30 days. Nothing here quotes a benchmark, because the only number that matters is the one you record in your own business before you start.
1. Start with a real process audit
Spend one week writing down what genuinely consumes time, before you evaluate any tool. The goal is a plain list of recurring processes with a rough sense of how often each happens and who does it. Ask the people doing the work, not the people describing it from memory.
- 1.List every recurring process a customer or staff member triggers — calls, inquiries, bookings, follow-ups, quotes, invoices, records updates.
- 2.For each, note frequency per week and roughly how long one instance takes.
- 3.Note who does it, and what happens when that person is unavailable.
- 4.Note where the information lives: which calendar, inbox, spreadsheet, or system.
- 5.Mark whether the steps are the same every time or depend on judgement.
- 6.Mark what happens if it is done wrong — an inconvenience, a lost sale, or something serious.
If a process cannot be described in six steps by the person who does it, it is not ready to automate. Write it down first; that documentation is useful whether or not you automate anything.
2. A simple scoring matrix
Score each candidate 1 to 5 on four factors. Volume, repeatability, and value are added together; risk is subtracted. It is deliberately crude — its purpose is to stop you choosing on frustration alone.
- Volume (1–5): how often it happens. High volume means you learn fast and any improvement compounds.
- Repeatability (1–5): how identical each instance is. Same inputs and same steps every time scores 5.
- Value (1–5): what resolving it faster is worth — revenue captured, hours returned, or fewer errors.
- Risk (1–5): the cost of getting it wrong, including reversibility, sensitivity, and who is affected.
Scoring three candidates
- After-hours inquiry response — volume 5, repeatability 4, value 5, risk 2 → score 12. Strong first choice.
- Appointment confirmations and reminders — volume 5, repeatability 5, value 4, risk 1 → score 13. Strong first choice.
- Custom quote preparation — volume 2, repeatability 2, value 4, risk 4 → score 4. Not first.
Take the highest score, not the loudest complaint. If two candidates tie, pick the one that is easier to switch off without disrupting anything else.
3. Strong first workflows
These recur across most service businesses and share the same profile: frequent, well defined, reversible, and easy to observe.
- After-hours and weekend inquiry response — the contacts you currently never see at all.
- Appointment confirmations, reminders, and reschedule links.
- Answering published questions: hours, location, service area, service descriptions, published prices.
- First-line intake capture, gathering clean structured details for a human to act on.
- Routine booking of standard appointment types against live availability.
- Post-conversation record keeping — writing outcomes into your system of record.
- Simple internal handoffs: notifying the right person with full context.
4. Poor first projects
Avoiding these is worth as much as choosing well. Each one either cannot be undone, requires judgement you cannot specify, or hides its failures until they are expensive.
- Anything irreversible: payments taken, refunds issued, contracts sent, records deleted.
- Anything clinical, legal, or financial in substance — those belong with qualified people.
- Judgement-heavy work like bespoke pricing, negotiation, or complaint resolution.
- Processes nobody has written down, or where each person does it differently.
- Work touching your smallest number of highest-value relationships.
- Anything requiring identity verification you cannot yet perform reliably.
- A cross-department project needing four systems and three approvals to launch at all.
5. Where a human must approve
Automation does not mean unattended. Decide the approval points before launch and write them as explicit rules, because an unstated approval point is simply an absent one.
- Anything irreversible or financial: approval every time, no exceptions.
- Anything sent in bulk or to a whole list: approval before sending.
- Anything outside published information — a price, a promise, a timeline — goes to a person.
- Any stated urgency or safety concern: immediate handoff with no interpretation attempted.
- Exceptions the rules do not cover: stop and notify, never guess.
- First week of any workflow: observe-only where possible, with a human sending or confirming.
Also write down the off switch: who can disable the workflow, how, and how quickly. Staff trust a system far more when they know how to stop it.
6. A practical 30-day rollout
- 1.Days 1–5: run the audit, score candidates, choose one workflow, and record today's baseline numbers.
- 2.Days 6–9: document the six steps, the answers it needs, the never-do list, and the approval points.
- 3.Days 10–14: configure it, connecting only the systems this one workflow requires.
- 4.Days 15–17: internal testing with staff replaying real past cases, including the awkward ones.
- 5.Day 18: launch narrowly — one channel, defined hours — with the owner reviewing everything daily.
- 6.Days 19–25: fix the top three failure patterns, tighten rules, and log every change.
- 7.Days 26–28: widen hours or add one adjacent case, only if week one's failures are resolved.
- 8.Days 29–30: compare against baseline and decide honestly: expand, adjust and rerun, or stop.
7. Common mistakes
- Choosing the most irritating process instead of the highest-scoring one.
- Automating a process that exists only in one person's head.
- Launching four workflows at once, so no failure can be attributed to anything.
- Skipping the baseline, which makes the day-30 review unarguable and useless.
- Leaving approval points unstated and discovering them after something irreversible happened.
- No named owner, so transcripts and exception queues go unread.
- Treating day 30 as a formality when “stop” should be a legitimate outcome.
First-workflow selection checklist
- One week of process notes gathered from the people doing the work.
- Each candidate scored on volume, repeatability, value, and risk.
- Chosen workflow describable in six steps by its current owner.
- Baseline numbers recorded before any change.
- Never-do list written, covering claims, pricing, and advice.
- Approval points defined for irreversible, financial, and bulk actions.
- Documented off switch a non-technical staff member can use.
- Scope limited to one channel and defined hours for launch.
- Internal test cases replayed from real past examples.
- Day-30 review booked, with expand, adjust, or stop all valid.
Key takeaways
- Pick the first workflow by score, not by frustration.
- High volume, high repeatability, real value, low risk — in that combination.
- If nobody can describe it in six steps, document it before automating it.
- Avoid irreversible, judgement-heavy, and substantively professional work first.
- Write approval points and an off switch before launch, not after an incident.
- Launch narrowly, review daily for a week, and widen only once failures are fixed.
- Record a baseline so day 30 produces a real decision instead of an opinion.
Related resources
Want to Talk Through Your First Workflow?
Book a demo and we'll walk through scope, data, and handoffs for your business. Reading this guide requires nothing from you.
Free, no-obligation demo · Tailored to your business