Skip to content
All projects
Case study · AI Adoption

AI Training for ALL

1,600 professionals. 8,000 training hours. AI taught through real business problems.

8,000 Training Hours

Context & Constraints

From 2023/2024 onward, KPMG intensified its AI adoption agenda. The firm wanted to change how people worked over the next years, not just introduce a tool.

A one-off lecture would not build capability at the scale of 1,600 professionals. The program had to work for people in the room and online, and it had to connect AI to real Tax work.

  • The program had to work for 1,600 people with very different AI maturity levels.
  • Hybrid delivery meant online participants needed active facilitation, not passive attendance.
  • Exercises had to come from real business problems collected before the event, or the training would feel generic.
We had to keep the people online from becoming just a little photo in the event.

Architecture

I helped design and facilitate recurring quarterly editions. Each edition had a theme: prompts, agents, or workflow opportunity mapping. Before the event, participants submitted real business problems through a form. Those became hands-on cases during the training. A dedicated online facilitator managed polls, chat and engagement so remote participants stayed active.

My Role

I idealized and executed the first event, built examples, forms and exercises, facilitated sessions, and organized results for leadership.

Results & Validation

1,600+
professionals reached
online and in-person participants across Tax
8,000
training hours
aggregate hours achieved before I left the program
Quarterly
program rhythm
themes included prompts, agents and opportunity mapping
Real cases
hands-on exercises
built from pre-event forms submitted by participants
The program reached 1,600 people and 8,000 aggregate training hours before I left. The metric mattered, but the design mattered more: adoption only works when the training uses the work people actually do.

Stack

EducationChange MgmtCARTS

Building AI that has to work every time?

I design, ship and operate systems like this one.

Let's talk

Building AI that has to work every time? Let's talk.

leonardo@leonardosa.pro

Madrid, Spain · from August 2026 · English / French / Portuguese / Spanish

© 2026 Leonardo Costa de Sá