Most businesses have more automation opportunities than they can act on. The risk isn't a lack of ideas — it's starting with a workflow that's hard to automate, delivers little value, or breaks the moment something unusual happens.
Here is the framework we use to decide where to start.
Step 1: List the repetitive work
Ask each team a simple question: what do you do every day or week that feels like copying, checking or chasing? Typical answers include moving enquiries from WhatsApp into a spreadsheet, following up on quotes, re-typing invoices, compiling weekly reports and chasing approvals.
Step 2: Score each workflow
Score each candidate from 1 to 5 on five dimensions:
- Frequency — how often does it happen? Daily beats monthly.
- Rule clarity — could you write down the decision rules? Clear rules automate cleanly; judgement-heavy steps need AI assistance or a person.
- Cost of delay or error — what happens when it's late or wrong? Lost leads, unhappy customers and incorrect invoices score high.
- Data readiness — is the information digital and accessible through APIs, or on paper and in people's heads?
- Reversibility — if the automation makes a mistake, how easily can it be corrected?
The best first projects score high on frequency, rule clarity, cost of delay and reversibility, with at least reasonable data readiness.
Step 3: Map the workflow as it really runs
Documented processes and real processes are rarely the same. Sit with the people who do the work. Note every tool they open, every message they send and every workaround they use. The workarounds are where the value — and the hidden requirements — usually are.
Step 4: Design for exceptions
An automation that handles the typical case and silently fails on the rest is worse than no automation. Decide up front what happens when data is missing, a system is down or a request doesn't fit the rules. Usually the answer is: route it to a named person with the context attached.
Step 5: Measure before and after
Before building, record how long the workflow takes and how often it goes wrong. After launch, measure the same things. This keeps the conversation grounded in evidence and tells you where to go next.
Common strong first candidates
- Lead capture and assignment from website, ads and WhatsApp into a CRM.
- Appointment reminders and confirmations.
- Scheduled operational reports assembled from several systems.
- Invoice intake with extraction and routing for approval.
- Customer status updates for orders, deliveries or applications.
What to avoid at the start
- Workflows that change every few weeks.
- Processes owned by nobody in particular.
- Large, cross-department transformations as a first project — start with one workflow and expand.
Automation compounds. A first project that works builds confidence, clean data and reusable integrations that make the second and third projects faster.