Bot Design, Build, and Deployment
What you must know
This domain tests engineering discipline: modular framework design (separation of business logic, config, and UI selectors), robust selector strategy, credential and asset management via a secure vault, comprehensive logging, and structured exception handling (business vs system exceptions). Candidates should know the PDD/SDD artefact chain, code review and peer review gates, version control, and packaging for release.
Deployment concepts: environments (dev/test/UAT/prod), CI/CD pipelines for bots, configuration promotion without code change, and deployment strategies such as pilot/ring rollout and blue-green or parallel-run cutover. Understand scheduling, queue configuration, bot capacity/utilisation, and monitoring/alerting with runbooks.
Traps
- Hard-coded selectors, URLs, or credentials - questions reward externalised configuration and vaulted assets.
- Conflating business exceptions (expected, route to queue/review) with system exceptions (unexpected, retry then escalate).
- Deploying straight to production without UAT or parallel run.
- Logging sensitive data in plain text.
- Treating bot capacity as unlimited; oversubscription causes queue latency that looks like a performance bug.
Rule of thumb
Anything that changes between environments must be configuration, never code. If promoting to production requires editing the script, the design is wrong.
Worked example
A bot reads a policy system. Wrong: URL and password in the workflow, one catch-all exception that silently continues. Right: endpoint and credentials in config/vault, a business-exception branch that routes 'policy not found' to a review queue, a system-exception branch that retries twice with backoff then alerts on-call with the transaction ID, and structured logs that record the ID but mask the policyholder name.