When deciding what to automate first, shortlist recurring work
Choose what to automate first by comparing recurring tasks your team can describe from beginning to end. Lead assignment, weekly reporting and client onboarding are useful candidates to investigate because each can have an identifiable trigger and finished result. Avoid starting with a department such as marketing or a broad ambition such as using more AI. For each candidate, record what starts the work, who owns it, which systems are involved and what completion means. Ask the people doing the work where they copy information, wait for an answer or repeat a check. Their examples should shape the shortlist before anyone chooses software.
Measure effort separately from waiting
Observe a representative set of recent cases. Record how often the workflow occurs, the minutes people actively spend on it, time waiting between steps and effort spent correcting mistakes. Include review and follow-up rather than timing only the easiest step. These measures explain different problems: a task may require little labour but remain unfinished because ownership is unclear. Note ordinary cases and exceptions separately so one unusually simple example does not become the baseline. If volume changes with campaigns or seasons, record that context. An estimate based on a busy week should not silently become a year-round forecast.
Remove unnecessary work before automating it
Challenge the current process while mapping it. Does anyone use the report being assembled? Is the same information being entered twice because two teams have never agreed on a shared record? Could a required field prevent a recurring clarification email? These questions can lead to a smaller and more useful project. Separate work that should disappear, work that needs a clearer procedure and work that is suitable for automation. A process audit should leave room for all three outcomes. Automating an unnecessary approval or duplicate spreadsheet preserves the original burden and adds another system someone must maintain.
Check whether the workflow is ready
A candidate needs usable inputs, an agreed sequence and a practical way to recognize a correct result. Check where information comes from, whether the business has the necessary access and which fields are regularly missing. Identify rules that can be stated explicitly and steps that require interpretation. For example, assigning an enquiry using its selected service may need a rule, while summarizing a long message may suit an AI-assisted step. Record what remains with staff in either case. If the process changes every week or nobody can agree on the expected output, resolve those gaps before estimating an implementation.
- A named person owns the workflow and can approve changes.
- The trigger, required inputs and completed output are defined.
- Ordinary cases and exceptions have representative examples.
- System access and integration dependencies are understood.
- Human review, a manual route and a way to measure effort are included.
Compare value before choosing what to automate first
Use a simple comparison sheet with recurring effort, clarity of the process, readiness of the information, implementation dependencies and consequences of an incorrect action. Write a short reason for each assessment rather than creating a precise-looking score from guesses. Consider an illustrative choice between assembling an internal report and sending personalized offers automatically. Both may save preparation time, but incorrect external messages create different consequences from a draft awaiting review. Choose the first project where useful effort can be reduced and an owner can inspect the output. Include the expected review burden when comparing the candidates.
Define a pilot with an owner and a decision
Write a short pilot brief naming the selected workflow, included cases, excluded actions, person responsible and evidence needed to decide what happens next. Keep a manual route available while the changed process is evaluated. Compare active staff time, corrections, completion and the usefulness of the result against the baseline. Count new checking and maintenance work as well as time removed. Decide whether to extend, revise or stop based on that evidence. GenRev's AI process audit can help prepare this brief and identify whether workflow automation, AI assistance or a clearer SOP is the most useful next piece of work.
Your decision about what to automate first should remain revisable. If the pilot needs more review or maintenance than expected, update the business case before extending it.