Start with a task, not a tool
Choose a repeated task with a clear input and a result that someone can review. Preparing an enquiry summary is a more useful starting point than asking for an AI transformation. Write down who does the task, how frequently it occurs, which systems it touches and what makes a result useful.
Define the baseline
Record how the task works today before adding automation. Consider time spent, corrections required and the delay before the next person can act. Pick one primary measure and a few quality checks. Faster output is not a success if your team spends longer fixing it.
Prepare representative examples
Collect examples of ordinary work alongside ambiguous requests, missing information and cases that should be escalated. Use only data you are authorised to process. Agree on which information the system can access and which material must remain outside the pilot.
Keep approval explicit
Separate drafting from acting. A pilot can prepare a response without sending it, or suggest a record update without changing the database. Name the person responsible for review and define what happens when the output is uncertain or incomplete.
Make the expansion decision
Compare the pilot with your baseline, including review effort and running costs. Inspect failures rather than relying only on averages. Expand when the task is useful and manageable; narrow the scope when the edge cases require too much supervision.
Your checklist
- One task and a named owner
- Approved inputs and access limits
- Representative test examples
- Human review and escalation
- Success measure and a stop decision
YOUR NEXT STEP
Have a project in mind?
Tell us what you want to build or improve. We will help you define the scope and a practical next step.
Start your project hello@bombstart.com
