Workflow prioritisation guide
Which business processes should you automate first?
Start with a process that combines meaningful business value, repeatable rules, accessible data, clear ownership and manageable exceptions. The biggest or most frustrating process is not always the best first automation.
The direct recommendation
Choose a process with enough volume or delay to matter, a clear beginning and verified outcome, rules that can be explained, data the project can access, and an accountable business owner. Prefer a workflow where automation can remove handoffs or re-entry while a person remains available for material exceptions.
Do not start with the most politically visible, ambiguous or irreversible decision. The first project should establish trust in delivery, data, support and measurement.
01
Evaluate value, suitability and readiness together
Frequency
How often does it run?
Repeated work creates more opportunity to reduce touches and waiting.
Friction
Where does it wait or fail?
Measure delay, re-entry, missing data, rework and unresolved exceptions.
Rules
Can the normal path be explained?
Stable criteria are easier to validate than judgement hidden in experience.
Data
Are inputs and systems accessible?
Confirm quality, identifiers, interfaces, permissions and source ownership.
Consequence
What happens when it is wrong?
High-impact decisions need stronger controls, evidence and human authority.
Ownership
Who owns the outcome?
A named sponsor and operational owner must resolve policy and exceptions.
02
Use a simple priority scorecard
Score each candidate from 1 to 5. Treat the result as a structured conversation, not an automatic investment decision.
| Criterion | 1 means | 5 means |
|---|---|---|
| Business value | Minor convenience | Material capacity, service, revenue or risk outcome |
| Frequency and volume | Rare | Frequent and repeated across the team |
| Rule clarity | Highly ambiguous | Normal path and decisions can be stated and tested |
| Data readiness | Unowned or inaccessible | Known sources, identifiers, quality and supported interfaces |
| Measurement readiness | No baseline or outcome | Time, touches, errors and completion can be measured |
| Exception safety | Errors are irreversible or severe | Exceptions can pause, route and recover safely |
Shortlist high-value candidates with strong readiness and safe exception handling. Investigate any score based on assumption rather than evidence.
03
Good first automation candidates
- structured request intake and assignment;
- document collection, validation and follow-up;
- quote or service-request routing;
- scheduled reporting collection and consolidation;
- reconciliation with an exception queue;
- customer onboarding steps and reminders;
- approval flows with clear decision rights; and
- status notifications triggered by verified system events.
Each can begin with deterministic rules. AI may help classify text, extract documents or draft communication, but it should not be the starting assumption.
04
Do not choose these as the first project without deeper discovery
| Candidate | Why it is risky first | Better next step |
|---|---|---|
| Rare executive exception | Low volume and high judgement | Improve information and decision support |
| Process changing every week | Automation will chase unstable policy | Stabilise ownership and the normal path |
| Poorly understood source data | Automated decisions inherit hidden quality problems | Profile, reconcile and assign data ownership |
| Irreversible high-impact decision | Consequence exceeds current control maturity | Automate preparation, keep final human authority |
| Organisation-wide replacement | Too many dependencies to prove quickly | Select one valuable workflow boundary |
05
Measure the current workflow before changing it
- cases started and completed per period;
- end-to-end completion time and waiting time;
- manual touches and re-entry steps;
- error, correction and rework rates;
- backlog size and age;
- exception types and responsible teams;
- service, customer or revenue consequence; and
- current software, support and labour cost assumptions.
Microsoft's Power Platform value guidance similarly recommends measuring business value at the application or solution level rather than trying to attribute every organisational benefit at once.
06
Prove one complete path
- Select the highest-confidence candidate, not simply the highest score.
- Define the trigger, verified outcome, owner and success measures.
- Map the normal path and top exceptions with frontline staff.
- Test the least certain data and integration dependency.
- Automate one end-to-end slice with human review where needed.
- Pilot with representative volume and deliberate failures.
- Compare the measured outcome and decide whether to improve, expand or stop.
Use the workflow automation ROI guide to turn the baseline into a financial case.
Sources
Primary references
Questions
Frequently asked questions
Which business process should be automated first?
Start with a frequent, rules-based process that creates measurable delay or manual effort, has a clear owner and outcome, uses reasonably accessible data and has manageable exceptions.
Should we automate the process with the highest labour cost?
Not automatically. High labour cost can signal value, but unclear rules, poor data, severe consequences or constant change can make it a risky first project. Balance value with readiness and consequence.
Should we automate a broken process?
Remove unnecessary steps and clarify ownership first. Automation can enforce a sound process, but it can also make waste and bad decisions happen faster.
Is AI required for workflow automation?
No. Deterministic rules, forms, APIs, queues and notifications often solve the first problem more reliably. Add AI only where interpretation or unstructured information creates a clear bounded need.
How many workflows should the first project include?
Usually one end-to-end workflow or a tightly related group. It should be complete enough to reach a verified business outcome and include at least one representative exception.
Choose the first workflow
Bring three candidate processes and the evidence you already have.
LCR can score the options, map the strongest candidate and define a focused first release with measurable outcomes.