Agree the process before configuring it.
Begin with a simple map of how a lead becomes a customer. Identify the entry points, qualification questions, ownership rules, pipeline stages and closure outcomes. Ask the people doing the work to explain where context gets lost today.
Write the meaning of each stage in plain language. “Qualified” should have a shared definition. “Won” should identify what has been confirmed and which values must be recorded. A clean interface cannot resolve an undefined operating process.
- Name the owner of the overall sales process.
- Define entry and exit criteria for each stage.
- List the exceptions and escalation routes.
- Agree what managers need to see in reporting.
Prepare the data and its responsibilities.
Map your existing fields to the new model. Separate people, companies, leads, opportunities and properties where the production design calls for it. Decide which system owns each field and how conflicting updates will be handled.
Test a small, anonymised import first. Check date formats, country codes, duplicate matching, missing required fields and ownership. Do not treat a successful file upload as proof that the data is ready.
- Confirm consent and communication preference fields.
- Review duplicate and merge rules.
- Validate record counts and sample field values.
- Define retention, export and deletion responsibilities.
Connect channels deliberately.
Each channel needs more than credentials. Agree the inbound event format, source attribution, duplicate handling, delivery errors and ownership. Webhooks should be verified and processed consistently, even when an event is retried.
Use sandbox or test accounts where available. Confirm what happens when an email fails, a call event arrives late or a website submits the same enquiry twice. Make the exception visible to the team rather than silently losing it.
Validate access and operational controls.
Review the role matrix with the people responsible for the data. Test what an agent, manager and administrator can see and change, and confirm that access is enforced by the server rather than only hidden in the interface.
Prepare monitoring, backups, restoration checks and incident ownership before the live rollout. Confirm that AI is optional and that the manual route remains available when external services are unavailable.
Pilot a complete journey, then expand.
Choose a manageable group and a representative set of scenarios. Walk through a new lead, a duplicate, an unassigned enquiry, an agent absence, a viewing, a lost opportunity and a confirmed sale. Include an AI-off scenario even if assistance is part of the planned rollout.
Collect specific feedback: where did the user hesitate, what information was missing, and which action was unclear? Fix those issues, update the playbook and expand in stages. Keep a named owner for training and for decisions about future changes.
- Train using realistic, anonymised scenarios.
- Record issues with steps and expected behaviour.
- Agree acceptance criteria and sign-off owners.
- Review adoption and data quality after launch.
Your practical checklist.
Use this as a starting point. Your ticks are remembered on this device.