Field checklists

Checklists exist because experienced people forget obvious things under pressure. Use these at each stage of an engagement — copy them into your project notes and tick them off with the customer where appropriate.

Qualification

• There is a specific business problem, not just interest in AI
• There is an executive sponsor with budget or access to it
• The data needed plausibly exists and can be accessed
• The deployment model (SaaS, customer cloud, on-prem) is known
• There is a realistic timeline and a reason it matters now

Discovery

• Interviewed the sponsor, a process owner and at least three frontline users
• Shadowed the workflow, or walked through real recent examples
• Captured volumes, handling times and error rates for the baseline
• Looked at a real sample of the data, not just the schema
• Identified IT, security and compliance constraints
• Sent a discovery summary and had it confirmed

Scoping

• Problem statement with metric, baseline and target
• In-scope and out-of-scope lists agreed
• Dependencies listed with owners and dates
• Evaluation set and acceptance criteria agreed, including who labels data
• MoSCoW priorities set within a fixed timebox
• Change process agreed
• Security review and access requests started

Proof of concept

• Uses real customer data in an approved environment
• Built on the same core services as production will be
• Evaluation results reported against the acceptance criteria
• Exit criteria for the pilot agreed before the POC ends

Pilot

• Defined user cohort, duration and review date
• Outcome, quality and adoption metrics instrumented from day one
• Single feedback channel and a named support contact
• Weekly review of metrics and feedback with the champion

Go-live

• Security sign-off received
• Monitoring and alerts for errors, latency, cost and quality
• Runbook written and reviewed by support
• Rollback plan tested; feature flag in place
• Cost forecast at production volume, with budget alerts
• Customer-side owner named and accepted
• Staged rollout plan, starting in assist mode for AI features

Handoff

• Customer team operated the system without help for an agreed period
• Architecture overview, runbook, operating guide and user guide delivered
• Hands-on enablement sessions completed
• Success review held and written up
• Next opportunities identified and prioritised with the sponsor
• Field report filed with your product team

Written by Bipin Singh

Senior Full-Stack Engineer · AI & AWS. I turn ambiguous customer problems into production systems — cloud, data and AI.

Work with me