Set the pilot boundary before anyone logs in
Start with a short list of account types, branch teams and approval points. A pilot works best when the scope is narrow enough to watch closely, but still close to the live path your clients will use in Johannesburg, Pretoria, Cape Town, Durban or Gqeberha.
Decide which business line will go first, which cases stay outside the pilot and who signs off each stage. Keep the rules plain so the team does not improvise on the day.
- Choose one or two customer journeys only.
- Limit the pilot to a small internal group or selected clients.
- Set the exact documents and session steps to be used.
- Name the person who approves changes during the pilot.
- Keep a record of every exception and manual intervention.
Map the session flow from invite to close
The session should move in a steady line: invite, consent, document capture, face comparison, liveness checks, decision and record keeping. Your client team must know where the handoff sits, especially if the customer starts with support staff and finishes with a verification agent in another location.
Use a simple flow chart, then test it with one real pilot case at a time. If the process requires extra explanation, shorten the wording before you add more customers.
- Confirm the invite message and join instructions.
- Check that consent is recorded before the session begins.
- Review how the ID document is captured on camera.
- Confirm what happens when a photo is unclear.
- Note the point where the account can move forward.
Build the document and face checks into one routine
A good pilot does not treat the ID image and the person on screen as separate chores. The agent should compare the document details, the face image and the live presentation in one controlled routine, with enough time to pause if the picture quality is poor or the document edges are cut off.
For South African onboarding, keep the checklist aligned to your internal FICA due diligence process and your POPIA record keeping practice. The pilot is also where you decide which checks are automatic, which need a second look and which need escalation to a supervisor.
- Test front and back capture for the document type in use.
- Check that the face is visible in steady light.
- Watch for blur, glare, cropped edges and shadowing.
- Use a liveness step that matches your platform flow.
- Keep notes on every mismatch or retry.
Train the people who speak to customers and review files
Training is stronger when it matches the pilot, not a generic script. The operations lead, verification team lead and training coordinator should work from the same session script, the same escalation ladder and the same language for explaining consent, document handling and pause points.
Client teams also need to know what to say when a customer fumbles the camera, uses the wrong document or leaves the session before the checks finish. Keep the script short enough to read under pressure and specific enough that new staff can use it without guesswork.
- Walk through one full session before live testing begins.
- Teach agents how to explain each camera step clearly.
- Show staff where to log failed captures and escalations.
- Use the same wording for consent and privacy prompts.
- Refresh the script after the first few pilot cases.
Watch the pilot with live notes, not loose guesses
The pilot should produce notes that show where the process holds and where it slows down. Track practical issues such as repeated captures, unclear instructions, drop-offs during the session and delays caused by missing documents or a poor internet connection on the customer side.
Operational reporting should stay plain. A short daily summary is often enough for the pilot group, provided it shows session status, exception types, escalations and any training gap that needs fixing. That makes it easier to keep Rosebank, Sandton and Midrand teams reading the same picture.
- Count completed sessions and failed sessions separately.
- List the most common reasons for intervention.
- Record any script changes made mid-pilot.
- Note where the customer journey breaks or stalls.
- Capture feedback from agents and supervisors together.
Decide what changes before the wider rollout
When the pilot ends, compare the working parts with the parts that caused friction. Some changes will sit in the workflow, some in the agent script, and some in the integration layer that connects your customer journey to the verification steps.
Do not widen the rollout until the service team has agreed the final process, the escalation path and the training refresh. That is the point where change request support matters most, because even small adjustments can affect how the next group of customers moves through the session.
- Lock the final workflow map.
- Update integration notes and field names where needed.
- Revise the training guide and pilot script.
- Set the quality monitoring points for live use.
- Keep a short handover note for the launch team.
