ServiceNow administration

Most of our automation experiments are on ServiceNow. Common candidates include:

• Routine data checks. Scheduled checks that find stale CMDB records, duplicate configuration items or missing owners and report them to the right person.
• Repetitive configuration changes. Applying the same change across several catalog items or workflows, with a record of what was changed.
• Instance comparison. Comparing configuration between instances to catch differences before they cause a failed release.
• Access reviews. Producing clear reports of who holds which roles, so reviews are quick and accurate.
• Release preparation. Assembling change notes, test evidence and checklists from information the team already records.
• Failed job monitoring. Watching scheduled jobs and integrations and alerting when something stops running.

For each candidate we ask the same questions. How often is the task done? How long does it take? What happens if the automation does it wrong? Who checks the result? A task that is done rarely, takes minutes and could cause harm if it fails is usually better left manual.

Automation in ServiceNow delivery

We also experiment with automating parts of our own ServiceNow delivery work, such as generating first drafts of ServiceNow test cases from written requirements, which a person then reviews, or producing status summaries from ServiceNow task records.

More ServiceNow experiments we run

• ServiceNow catalog clean-up. Finding catalog items that nobody has requested for a long time, so the ServiceNow catalog stays short and useful.
• ServiceNow notification review. Listing every notification a ServiceNow workflow sends, so duplicates and noise can be removed.
• ServiceNow update-set checks. Checking update sets for missing dependencies before they move between ServiceNow instances.
• ServiceNow report validation. Comparing ServiceNow report results with the raw records, to confirm a dashboard means what its title says.

Each of these follows the same pattern: a written question, a test on a copy of the data and a short write-up of what the ServiceNow instance revealed.

Salesforce automation

We run a small number of experiments on CRM workflow and data-cleaning automation, usually to test whether a repetitive manual step can be safely removed.

How we keep automation safe

Test first. Every automation runs in a test environment before it touches anything real.

Log everything. Each run records what it did, so a mistake can be found and reversed.

Keep a human on the important steps. If an error would be costly, a person approves the result.

Measure the saving. We compare the time spent before and after, and say honestly if it was not worth it.