Skip to content

Case study · AI operations

Automating appointment confirmations without losing operational control

Voice AI, customer identity checks, structured outcomes and CRM-ready updates for a UK estate-planning team.

Client and sector

A UK estate-planning business with a field-based advisory team. Advisors visit customers at home, so every missed or unconfirmed appointment costs travel time as well as the slot itself. The operation runs on Activ CRM, with appointments booked by an in-house team.

The operational problem

Every appointment needed a confirmation call a day or two beforehand. The volume was manageable when the team was small; as bookings grew it became a daily calling task that competed with new enquiries for the same people.

The calls themselves were simple. Most customers confirmed. A minority needed to rearrange, and a smaller group had questions that only an advisor could answer. The difficulty was not the conversation, it was the consistency: who had been called, what they said, and whether the CRM had actually been updated afterwards.

Why the existing process wasn't working

  • Confirmation calls were worked from a list, so outcomes were recorded after the fact, sometimes in bulk, and sometimes not at all.
  • Rearrangements were handled in the call but not always reflected in the advisor's diary until later in the day.
  • There was no reliable way to distinguish "no answer" from "declined" from "left a message", which affected whether a second attempt was made.
  • Customers with questions were promised a callback that then had to be manually tracked.

The process did not need reinventing. It needed running the same way every time, with the outcome captured as data rather than a note.

What was designed

We designed the workflow before we designed the agent. The team mapped each possible path through a confirmation call: right person or not, good time or not, confirm, rearrange, decline, question outside scope. Each path ends in a defined outcome.

The Activ Agent voice workflow then follows that map:

  1. The agent introduces itself as an automated assistant calling on behalf of the business and confirms it is speaking to the named customer.
  2. It confirms the appointment date, time and advisor.
  3. It handles confirmation, rearrangement within available slots, or a decline.
  4. Anything outside the workflow is captured as a question and routed to the booking team as an exception with the call context attached.

Customers are told at the start that they can ask for a person at any point.

Technology used

  • Activ Agent for the voice workflow, outcome capture and escalation rules.
  • Activ CRM as the system of record for appointments, customers and advisor diaries.
  • Workflow automation to schedule calls inside the confirmation window and inside each customer's contact preferences.
  • Microsoft Azure for hosting and for the integration services between the agent platform and the CRM.

Integration architecture

The agent reads appointment and customer context from the CRM through its API immediately before each call, so it is never working from a stale export. Structured outcomes are written back through the same API as the call completes: appointment status, any new slot, a short outcome summary, and a link to the recording and transcript for review.

Exceptions become CRM tasks assigned to the booking team, with the context needed to pick the conversation up. No spreadsheet sits between the two systems.

Human and process considerations

  • The team reviewed transcripts daily in the first weeks, and adjusted phrasing where customers hesitated.
  • Retention periods for recordings and transcripts were agreed up front and are enforced by the platform.
  • The booking team owns the exception queue, so there is one place where "the agent couldn't handle this" lands.
  • Customers who ask not to be called by an automated assistant have that preference recorded, and the workflow respects it.

Outcome

Confirmation calls now run as a scheduled workflow rather than a daily task. The booking team's work has shifted from calling everyone to handling the exceptions the agent hands over, and every appointment carries a structured confirmation status the advisors can rely on.

We are not publishing volumes or percentages for this engagement. The operational change is the point: the process is consistent, auditable and no longer dependent on who was available to make calls that day.

What happened next

The same workflow pattern is being applied to document chasers, where the agent confirms a document has been received or explains what is still outstanding. The identity check, outcome capture and escalation rules are reused as-is.

Have a similar problem?

Tell us about the workflow. We'll tell you what we'd do.