AI & Innovation Portfolio Vend.AI

03 · What’s next

From sales training to an intelligent sales coach.

An AI sales coach for a Latin American telecom operator. What it trains is not the script, but reading the person on the other side of the conversation.

Role
Direction of the product experience
Organization
Capgemini · Applied Innovation Exchange
Client
Latin American telecom operator
Focus
AI product · Learning experience

01 The work

I directed the creation of an AI-enabled sales coaching experience for a Latin American telecom operator, designed to support learning, knowledge development and commercial performance.

Product walkthrough
Instead of a course to complete, the product offers customers to practice on, each profile filterable by difficulty. Silent screen capture · interface in Portuguese

A salesperson picks a profile, works a live conversation against it and gets structured feedback on what actually happened in the exchange. The skill being trained is not the script. It is reading the customer in front of you.

Product walkthrough
A session in progress: the conversation on the right, what it is testing on the left. Silent screen capture · interface in Portuguese

The management layer closes the loop. It shows what teams practice, where they get stuck and how usage moves, which is what turns a training tool into something a commercial leader can actually run.

Product walkthrough
The admin view: sessions, feedback indicators, token consumption and per-user performance. Silent screen capture · interface in Portuguese
A synthetic customer profile in Vend.AI showing objectives, pain points and context.
Each profile carries objectives, pain points and context.
The end-of-session feedback form in Vend.AI.
Every session ends with structured feedback.
  • AI product vision
  • Learning experience
  • Knowledge support
  • Gamified development
  • Pilot definition

02 How I ran it

We trained the judgment, not the script.

Sales training products usually teach the pitch, because the pitch is the part that is easy to write down. I started this one by asking what a good salesperson actually does that a weak one does not.

  1. I named the skill before we designed the product

    What we are training is not the script. It is reading the person on the other side of the conversation and choosing what to say next. Once that was the stated skill, a course shaped product stopped being an option.

  2. I asked for practice against a profile, not a curriculum

    The product gives a salesperson customers to rehearse with instead of modules to complete. You can finish a curriculum without learning anything. You cannot fake a conversation past the point where it stops working.

  3. I made the feedback about the exchange, not the answer

    Sessions are assessed on what happened in the conversation. What was raised, what was missed, what was positioned. That is the level at which a salesperson can do something different next time.

  4. I put running cost next to performance in the same view

    The admin dashboard shows sessions, feedback indicators and token consumption together. An AI product whose running cost is invisible to the people approving it does not survive its own pilot.

We defined the pilot before we built, so there was a specific claim we could be right or wrong about. Without that you have a demo with a roadmap attached to it.