How to turn early enthusiasm into organization-wide Artificial Intelligence (AI) fluency and measured adoption.
Gallup reported that 45% of United States employees used AI at work at least a few times a year by the third quarter of 2025, up from 21% in 2023, yet only about one in ten used it daily. Access is spreading faster than fluency. The difference between occasional curiosity and a daily habit is rarely the tool itself. It is the people around the user.
This is Part 2 of a nine-part series on running AI transformation. Over the past few years, I authored 35 AI training courses that more than 6,000 employees completed, with over 21,000 course completions, and I served on a firm wide AI Accelerators Committee. The lessons below come from that work and apply in any industry, from banking and insurance to healthcare, manufacturing and government.
In this article, you will learn:
Why central training alone plateaus
How to choose champions who change behavior
How to design role-based learning paths
How to turn champion ideas into approved use cases
How to measure whether the network is working
Why Central Training Alone Plateaus
Training opens the door, but trusted peers are the ones who walk colleagues through it.
Central training works well for the first wave of enthusiasts. Then participation flattens, and usage settles at occasional experimentation.
The reasons are predictable:
Generic examples: People struggle to translate a classroom demo into their own spreadsheets, reports and customer conversations.
No follow-up: Questions appear on day three of real work, long after the course ended.
Fear of looking foolish: Many people will not admit in public that they do not know how to start.
Unclear permission: Employees are unsure which tools and which data they are allowed to use, so they wait.
A champions network solves these problems by placing a trained, trusted and local guide inside every team.
Courses create awareness at scale, but daily habits form one conversation at a time inside each team.
Choosing Champions for Credibility, Not Title
Choose champions people already trust, and that trust will transfer to the tools.
The best champions are rarely the most senior or the most technical people. They are the colleagues others already ask for help.
Respected practitioner: Someone who knows the real work of the team and is listened to by peers.
Curious and patient: Willing to experiment, and willing to explain the same thing five times without frustration.
Well connected: Has relationships across teams, so good practices travel quickly.
Protected time: Champion work is recognized in goals and given dedicated time, not squeezed into evenings.
Balanced coverage: Every function, location and level is represented, not only technology teams.
As a practical rule of thumb, aim for one champion for every few dozen employees, adjusted for how much each team's work will change.
The right champion is the colleague everyone already asks for help, now equipped to help with AI.
Role-Based Learning Paths
Teach each role what will change its own work tomorrow morning.
One curriculum for everyone produces shallow learning. Role-based paths connect AI skills directly to daily work.
Three core paths
Leaders: Strategy, value, risk appetite, how to sponsor use cases and how to role model AI use in front of their teams.
Everyday users: Effective prompting, verifying outputs, data handling rules, approved tools and the task patterns most relevant to their role.
Builders: Workflow automation, retrieval over company documents, agents, evaluation methods and the governance steps every build must pass.
The extra layer for champions
Coaching skills: How to guide a colleague through a first real task instead of doing it for them.
Governance literacy: What data is allowed, which tools are approved and when to escalate a question.
Use case framing: How to describe a problem, its users, its data and its expected value in a few sentences.
Converting courses into certifications on the Learning Management System (LMS) gave learners a credential they valued and gave the program a reliable signal of progress.
One curriculum for everyone teaches no one; learning paths built around real roles turn knowledge into daily practice.
From Idea to Approved Use Case
Every good idea deserves a clear path, a fair review and a fast answer.
Champions surface ideas constantly. Without a pipeline, those ideas die in inboxes and the champions lose energy.
Capture: A short intake form records the problem, the users, the data involved and the expected value.
Qualify: The champion and a business lead confirm the idea is valuable, feasible and not already solved elsewhere.
Review: A cross-functional committee reviews value, data, risk and implementation architecture.
Decide: The committee approves, redirects to an existing solution or declines with a clear reason.
Deliver and share: Delivered solutions are shared back through the network so other teams can reuse them.
Give credit and recognize: Acknowledge the employee who originated the idea for the new use case.
Speed matters as much as rigor. An idea that waits three months for an answer teaches champions to stop submitting ideas.
A transparent intake path turns champions from enthusiasts into a steady pipeline of governed, valuable use cases.
Measuring the Network
What gets measured gets supported, and what gets supported keeps growing.
Measure the network in four tiers, from reach to value:
Reach: Enrollments, course completions and certifications by team.
Usage: Weekly active users of approved tools and the variety of tasks they use them for.
Pipeline: Use cases submitted, approved and running in production.
Value: Hours released, cycle time, quality improvements and employee sentiment.
Report results by team so support goes where adoption lags, and recognize champions publicly. Recognition costs little and keeps the network alive.
Completions show reach, usage shows habit and production use cases show value; a healthy network reports all of them.
Key Takeaways
Adoption spreads from person to person, so build the network that carries it.
Training is not enough: Central courses create awareness, but peers create habits.
Pick trusted champions: Credibility inside the team matters more than title or technical depth.
Design by role: Leaders, everyday users and builders each need their own learning path.
Build the pipeline: A fast, fair intake process turns ideas into approved use cases.
Measure four tiers: Track reach, usage, pipeline and value, and recognize champions publicly.
The organizations that win with AI will be the ones that turn a few enthusiasts into a network that teaches everyone.
What has worked best in your organization to move people from trying AI to using it every day?
Previously, Part 1: Strategy to Execution.
Next in this series, Part 3: Change Is the Product: Driving AI Adoption with ADKAR.Explore more: Start Here · Topics · Archive · About