The first business process worth automating is usually one your team can explain, observe and support. It may be less dramatic than the idea that attracts the most attention in a meeting. A modest workflow with reliable inputs and a clear owner can provide a better starting point than a large process whose rules keep changing.
Choosing that first process is a practical investigation. You need to understand the work, compare suitable candidates and establish whether the proposed automation improves the complete task. This guide shows how to make the selection using evidence from everyday operations, without assuming that every repeated activity should be automated immediately.
Collect examples of repeated work
Ask the team which tasks they repeat and where avoidable effort occurs. Useful candidates might include copying approved enquiry details into another system, preparing a routine status message or creating a standard task after a confirmed booking. Ask for recent examples so the discussion stays connected to actual work.
Record the trigger, input, action and result for each candidate. Who starts the task? Which information is needed? What does the person do? Who uses the output? A task that sounds simple in conversation may depend on several decisions that the experienced employee makes without noticing.
Include the work around the visible action. Copying a record may take a minute, but checking incomplete information or correcting duplicates may take longer. The selection should consider the whole activity, rather than judging the benefit from the easiest step alone.
Check whether the process is understood
Watch a suitable sample of the work with the people who perform it. Compare the written procedure with what happens in practice. If different team members use different rules, establish whether those differences are intentional or signs of an unresolved process.
Write the rules in ordinary language. For example, a confirmed request might create a task for a particular team, while an incomplete request needs review. Identify the information that makes that distinction possible. If staff have to infer the answer from context that is not recorded, the automation brief needs more investigation.
Prioritise procedural tasks with a useful lifespan, and recognise that human decisions can remain within a wider workflow. For a first project, choose a part of the process with clear behaviour rather than trying to automate every judgement at once. Define the trigger, expected result, exceptions and approval points. Our workflow automation and approvals guide provides the related checks.
Measure the current effort fairly
Record how often the task occurs and how much active work representative examples require. Separate normal cases from exceptions. Note the effort involved in finding information, carrying out the action and checking or correcting the result. Use observed figures where possible and label estimates honestly.
Distinguish staff effort from elapsed waiting time. A request may sit in a queue for two days while needing only a few minutes of work. Automation might help with routing or reminders, but it does not automatically remove every cause of delay. Identify the particular problem the proposed change would address.
Look at mistakes and interruptions too. A low-volume task may still deserve attention if an error creates substantial rework. Conversely, a frequent task may be a poor first candidate if the information is unreliable and the team cannot yet define the correct result. Keep the evidence together so the decision reflects more than one number.
Compare candidates through benefit and readiness
Create a short comparison sheet. For each process, record the expected benefit, clarity of rules, quality of inputs, number of connected systems, exception frequency and available owner. Explain the observations behind the assessment rather than giving unexplained scores.
If you use a simple rating scale, define it before comparing candidates. “Ready” might mean that the input fields and rules are agreed, representative examples are available and the process owner can review tests. The rating should help discussion, not pretend to calculate an objective answer from subjective guesses.
Identify any condition that rules out immediate work. A missing system owner, an unresolved information handling decision or an unavailable integration may require preparation first. Keep those candidates on the list with their next action, while considering another process that is ready for a controlled pilot.
Ask whether simplification should come first
Review why the repeated activity exists. The team may be copying information because two forms collect the same details, or preparing a message because customers receive unclear instructions earlier. Removing an unnecessary step can be more useful than building an automation to perform it faster.
Ask whether an existing tool already provides an appropriate function. A supported setting or a clearer working procedure may address the need. Compare those options with an integration or a bespoke workflow, including the effort of maintaining each one.
Do not preserve a confusing process simply because it is familiar. Agree the desired workflow before implementing the automation. Document what changes for staff and customers so the pilot tests the intended process rather than reproducing avoidable problems.
Define a narrow first version
Choose one trigger, one clear input and one useful output where possible. For example, a confirmed website enquiry could create a review task with the approved details and assigned owner. That first version can be checked without also automating qualification, proposal writing and every follow-up message.
Specify where a person remains responsible. The workflow may prepare a draft, route a record or request a decision rather than make the decision itself. State what staff should review and what happens when the input does not meet the agreed conditions.
Giraffe Digital's digital strategy services provide a starting point for examining how website enquiries and working tools connect. The automation brief should support that wider journey while keeping the first implementation small enough to understand and operate.
Plan exceptions before the pilot
List the cases that should stop or divert the workflow. Missing details, an unexpected service choice or an unavailable connected system are useful examples. Decide which issues staff can resolve and which need the supplier. Give every exception a visible destination and an owner.
Agree how repeated processing is handled. A task should not appear several times because a message was delivered twice or a staff member retried a failed step. Ask the developer to demonstrate the expected behaviour with controlled examples and explain the record that shows what happened.
Write a fallback for a stopped pilot. Staff need to know how to continue the work without accidentally running both the manual process and the automation against the same records. A first project should leave the team able to handle the task even when the new workflow is unavailable.
Test benefit and operating effort together
Run a controlled sample containing normal cases and agreed exceptions. Compare the output with the result an informed staff member expects. Include the work of reviewing, correcting and supporting the automation when assessing its benefit.
Ask somebody other than the builder to use the operating instructions. Can they find the record, understand a failed run and take the correct next step? That check reveals whether the workflow is ready to become part of everyday operations.
Review the original evidence after the pilot. Did the change reduce the identified effort or improve the intended handover? Did it create a new queue or move work to another team? Use those observations to decide whether to extend, adjust or stop the project. Avoid declaring success from the number of automated runs alone.
Keep the first success maintainable
Record the approved trigger, fields, rules, connected tools and owner. Include what should prompt another review, such as a changed form or service process. The person responsible for the workflow needs to know when the assumptions behind it have changed.
A good first automation teaches the team how to select, test and maintain further improvements. Start with a well-understood task, compare it fairly with alternatives and verify the complete working result. That creates a dependable foundation for automation that serves the business.
Frequently asked questions
How should we collect examples of repeated work?
Ask the team which tasks they repeat and where avoidable effort occurs. Useful candidates might include copying approved enquiry details into another system, preparing a routine status message or creating a standard task after a confirmed booking. Ask for recent examples so the discussion stays connected to actual work. Record the trigger, input, action and result for each candidate. Who starts the task? Which information is needed? What does the person do? Who uses the output? A task that sounds simple in conversation may depend on several decisions that the experienced employee makes without noticing. Include the work around the visible action. Copying a record may take a minute, but checking incomplete information or correcting duplicates may take longer.
How should we measure the current effort fairly?
Record how often the task occurs and how much active work representative examples require. Separate normal cases from exceptions. Note the effort involved in finding information, carrying out the action and checking or correcting the result. Use observed figures where possible and label estimates honestly. Distinguish staff effort from elapsed waiting time. A request may sit in a queue for two days while needing only a few minutes of work. Automation might help with routing or reminders, but it does not automatically remove every cause of delay. Identify the particular problem the proposed change would address. Look at mistakes and interruptions too. A low-volume task may still deserve attention if an error creates substantial rework.
How should we ask whether simplification should come first?
Review why the repeated activity exists. The team may be copying information because two forms collect the same details, or preparing a message because customers receive unclear instructions earlier. Removing an unnecessary step can be more useful than building an automation to perform it faster. Ask whether an existing tool already provides an appropriate function. A supported setting or a clearer working procedure may address the need. Compare those options with an integration or a bespoke workflow, including the effort of maintaining each one. Do not preserve a confusing process simply because it is familiar. Agree the desired workflow before implementing the automation. Document what changes for staff and customers so the pilot tests the intended process rather than reproducing avoidable problems.
How should we plan exceptions before the pilot?
List the cases that should stop or divert the workflow. Missing details, an unexpected service choice or an unavailable connected system are useful examples. Decide which issues staff can resolve and which need the supplier. Give every exception a visible destination and an owner. Agree how repeated processing is handled. A task should not appear several times because a message was delivered twice or a staff member retried a failed step. Ask the developer to demonstrate the expected behaviour with controlled examples and explain the record that shows what happened. Write a fallback for a stopped pilot. Staff need to know how to continue the work without accidentally running both the manual process and the automation against the same records.
How should we keep the first success maintainable?
Record the approved trigger, fields, rules, connected tools and owner. Include what should prompt another review, such as a changed form or service process. The person responsible for the workflow needs to know when the assumptions behind it have changed. A good first automation teaches the team how to select, test and maintain further improvements. Start with a well-understood task, compare it fairly with alternatives and verify the complete working result. That creates a dependable foundation for automation that serves the business.


