Confirm that the problem needs software.
A request can stall because nobody has accepted ownership, the service definition is unclear or staff cannot find the current policy. Replacing the tool may leave those problems intact. Map one recurring service from intake to checked closure and identify the exact point that needs improvement.
Record request volume when available, the people involved, repeated entry, incomplete information and the consequences of a failed handoff. Do not invent a time-saving estimate before there is a measured baseline. The workflow guide and workbook help make those decisions explicit.
Choose the initial process with the automation priority matrix. For assistive model-generated steps, use the distinct AI review and approval controls.
Compare four implementation choices.
| Choice | A useful fit to test | Tradeoff to examine |
|---|---|---|
| Improve the process | The volume is manageable and the missing step is ownership or clarity. | Staff still maintain the record and follow-up manually. |
| Configure an existing tool | The current platform already supports the necessary fields, roles and rules. | Configuration limits, licenses and existing support obligations. |
| Buy a supported product | A standard workflow fits with acceptable changes. | Implementation effort, permissions, integration and exit constraints. |
| Build a focused tool | A specific handoff remains unsupported and a team can maintain the solution. | Long-term ownership, testing, security review and support capacity. |
On a small screen, scroll the table sideways to see every column.
A hybrid can be appropriate: keep the system of record and add a small supported intake or routing component. Compare that approach with using the existing product more effectively before replacing a working system.
Test the ordinary case and the awkward case.
Give each option the same scenarios. Include a complete request, missing information, a duplicate, a declined transfer, waiting for another party and a reopened case. Observe whether staff can understand the current owner and next action without a side conversation.
Use sample records, not confidential resident details, during early demonstrations. Record whether a capability is demonstrated, needs configuration, needs custom work or remains unknown. Include staff effort and any step that occurs outside the tool.
The GOV.UK guidance on commercial products (opens in a new tab) emphasizes discovering the problem and testing alternatives before committing to technology. Its Service Standard applies in its own government context; it does not decide a municipality’s procurement route.

Evaluate the team that will operate the choice.
Name the process owner, technical owner, account administrators and escalation route. Ask who updates a rule when a department changes responsibilities, who tests a release and who notices a failed integration. A custom tool needs a realistic maintenance arrangement; a purchased tool still needs accountable configuration and use.
Include data export, documentation, staff training, access review and continuity during absence. An attractive subscription or build price cannot establish the total operating cost without those assumptions.
Use a bounded pilot with a decision at the end.
Choose one category, a defined group of staff, a stated test period and an agreed set of nonemergency scenarios. Decide how test records are separated, what would stop the pilot and how staff revert to the approved process if it fails.
Assess both fit and usability. Can staff find the owner? Do they create duplicate records? Can a reviewer inspect why a case closed? Use the measurement guide to define the baseline and what the pilot can actually demonstrate.
Keep the decision tied to evidence.
Summarize the need, options considered, test results, unresolved dependencies, operating costs and reasons for choosing an approach. A decision record makes it possible to revisit the choice when the service or platform changes.
For connected systems, work through the integration planning checklist. For a concrete demonstration, try the fictional resident request walkthrough. To discuss implementation, review government workflow automation and staff tools.
Use a pilot record with a real stop and decision point.
| Field | Define before starting |
|---|---|
| Boundary | Included request category, staff group, test environment and observation period. |
| Baseline | Comparable cases, start/end events and data-quality limitations. |
| Hypothesis | Which specific handoff or task should become more reliable or easier. |
| Scenarios | Ordinary case, missing details, duplicate, decline, waiting, reopening and failed connection. |
| Stop and fallback | Named pause authority, unacceptable conditions and manual recovery process. |
| Decision | Continue, revise, stop or gather more evidence, with owner and review date. |
On a small screen, scroll the table sideways to see every column.
Fictional planning example. Cedar Bay tests one nonemergency internal request category using fictional records. The pilot covers intake and accepted ownership. It excludes real submissions, payments and public eligibility decisions. The test group observes whether each request retains an owner through a declined handoff.
A missing request or incorrect permission triggers a pause and reconciliation. Staff keep a manual record of pending items so they can restore the approved route. A successful normal case does not cancel an unresolved failure scenario.
At the review point, compare the same task definitions and inspect unfinished work. Include setup, review and support effort. Synthetic tests demonstrate behavior under those scenarios; they do not prove live adoption or service improvements. A live pilot requires separately approved data, access and service arrangements.
Select a candidate with the priority matrix and document the full cost assumptions before a larger commitment.


