Choosing the first automation project shapes more than one investment. It influences confidence, internal capability, management attention and how future proposals will be judged.
The loudest problem is not always the best starting point. A highly visible manual operation may have low economic impact, unstable inputs or product variation that makes automation fragile. A less dramatic bottleneck may offer a clearer baseline, simpler intervention and faster organisational learning.
Create candidates from losses
Walk the value stream with production, quality, maintenance, material flow and planning. Look for recurring conditions rather than collecting product ideas. Useful candidates may include:
- repetitive safety or ergonomic exposure;
- a constraint that regularly limits output or causes waiting;
- repeat defects caused by presentation, sequence or missed checks;
- manual recording that delays action or breaks traceability;
- material movement with stable routes and predictable demand; or
- inspection work with clearly observable defect characteristics.
Describe each candidate as an operational problem. “Install a robot” is a solution. “Operators wait for completed parts while the constraint remains unattended” is a problem that can be measured and compared with other losses.
Screen every candidate consistently
Operational impact. What changes in safety, quality, delivery, cost, capacity or responsiveness if the problem is reduced?
Frequency and repeatability. Does the condition occur often enough, with sufficiently consistent inputs and decisions, to justify automation?
Process stability. Is the current method understood, or would automation reproduce variation and hidden workarounds?
Technical tractability. Can parts, interfaces, space, cycle, sensing, recovery and safeguarding be addressed without disproportionate complexity?
Ownership and support. Is there a responsible process owner and a practical maintenance, training and escalation route?
Learning value. Will the project build reusable knowledge, data, interfaces or capability for later opportunities?
Prioritisation should make judgement visible. It should not turn weak assumptions into false precision.
Use scores to compare, then use gates to decide
A simple scoring matrix can help a cross-functional team compare candidates. Use a small scale, agree what each score means and attach the evidence behind it. Weighting impact more heavily than novelty usually keeps the discussion connected to operating value.
Do not allow averages to hide a fatal condition. Apply explicit gates after scoring:
- unresolved safety or regulatory responsibility;
- no agreed problem definition or baseline;
- process inputs too unstable for the proposed concept;
- no accountable owner for operation and recovery;
- a critical interface or product variant not yet understood; or
- a business case dependent on an assumption that cannot be tested.
A gated candidate is not permanently rejected. It returns to the opportunity list with specific readiness actions.
Validate the shortlist at the process
Take the highest-ranked candidates back to the workplace. Observe representative shifts, products, operators, changeovers and abnormal conditions. Reconcile what the records say with what actually happens.
For each shortlisted process, collect enough evidence to answer:
- Where is the loss created and where is it detected?
- How much of the work is truly repetitive, and how much depends on judgement?
- Which product, material or environmental conditions change the task?
- How does the team currently recognise, contain and recover from failure?
- What upstream or downstream dependency could simply move the bottleneck?
This step often changes the ranking. A recorded machine stop may be caused by inconsistent presentation. A slow inspection may be receiving avoidable variation from an earlier operation. The correct first intervention may sit outside the initially nominated workstation.
Compare the simplest credible interventions
For the final candidates, compare more than one route. Standard work, layout, fixtures, poka-yoke, sensing, visual control or better information may remove enough loss without full automation. A staged solution can also create the stable conditions needed for a later automated step.
Evaluate each route using the same requirements and total-cost boundary. Include engineering, integration, safeguarding, fixtures, programming, training, spares, maintenance, ramp-up and internal time. Benefits should include the operating effects that can be credibly measured—not only direct labour.
Select a first project that can teach
A strong first project has meaningful value, bounded scope and a visible evidence trail. It gives operators and maintenance teams a role in design and acceptance. It also produces reusable learning about suppliers, controls, data, safety, recovery and project governance.
Document why it was selected and why other candidates were deferred. Define the assumptions, pilot or acceptance test, owner and review date. If the evidence changes, be willing to stop or redesign the project before larger commitments are made.
The right process to automate first is the one that combines operational value with sufficient readiness and a manageable learning step. That is a more durable starting point than choosing the most impressive technology or the most visible manual task.