Why AI Adoption Is a Workflow Problem, Not Just a Training Problem
- Mar 24
- 3 min read
Updated: May 29
Over the past year, I have spent a lot of time helping colleagues explore how AI might fit into their work.
Some people picked it up quickly. Others attended the same sessions, received the same guidance, and still struggled to make AI useful in their day-to-day work.
At first, I assumed the problem was training.
Maybe people needed more practice. Better examples. More exposure to effective prompting techniques.
The more I worked with people, however, the less convinced I became that training was the main issue.
Instead, I started noticing a different pattern.
The people who struggled were not necessarily the people who knew the least about AI.
Often, they were the people trying to fit AI into workflows that had never been examined closely in the first place.
The Common Approach to AI Adoption
Many AI adoption efforts begin with training.
Organizations run workshops, share prompting tips, demonstrate use cases, and encourage employees to experiment with AI in their work.
There is nothing wrong with this approach. People do need to understand how AI works before they can use it effectively.
The challenge is that understanding the tool does not automatically change how the work gets done.
Knowing how to write a better prompt does not necessarily help someone decide:
What information matters
What context should be included
What risks should be highlighted
What makes an output useful
When human review is needed
These considerations already exist inside the workflow.
AI simply makes them more visible.
What Changed My Thinking
One of the projects that influenced my thinking started with a simple question from a colleague.
Before each coaching session, he reviewed previous coaching notes, participant goals, ongoing commitments, and recent progress. The preparation process was not particularly difficult, but it was repetitive and happened multiple times each week.
On the surface, this looked like a straightforward efficiency problem.
Could AI help save time?
To answer that question, I first looked at the existing workflow.
The coach's preparation process involved:
Reviewing participant goals
Reviewing notes from previous coaching calls
Identifying unfinished commitments
Deciding what should be discussed next
Building the agenda for the upcoming session
As I unpacked the process, something interesting became clear.
The real work was not creating the agenda.
The real work was understanding everything that happened before the agenda was created.
What I Learned
I eventually built an AI-assisted workflow to support the coach's preparation process.
The original goal was to save time.
The most valuable outcome turned out to be something else.
Before building anything, we spent time understanding how the coach prepared for a session. We mapped the steps, identified the information being reviewed, and looked at how the agenda was created.
That process revealed that preparing an agenda involved much more than simply generating a document.
Once the workflow became visible, it became much easier to identify where AI could help and where it was unnecessary.
The lesson I took away was simple: The AI was not the starting point, understanding the work was.
Why Training Alone Often Falls Short
This experience made me look at AI adoption differently.
Many organizations focus on teaching people how to use AI.
Far fewer spend time understanding how work is actually being done today.
If the workflow remains invisible, people often get inconsistent results. AI outputs may look reasonable, but important information, assumptions, or priorities may never make it into the process. Employees lose confidence in the tool and adoption slows down.
The issue is not always that people need more AI training.
Sometimes the issue is that nobody has taken the time to understand the work itself.
A Different Question
I still believe training matters.
People need opportunities to learn how AI behaves and where it can be useful.
But increasingly, I find myself asking a different question.
Instead of asking:
"Are people using AI?"
Maybe we should also be asking:
"Do we understand how this work actually happens?"
The coaching project started as a question about AI.
It ended as a conversation about workflow.
The more projects I work on, the more I find that understanding the work itself is often harder, and more valuable, than choosing the AI tool.
I don't think I have all the answers yet.
But I am starting to suspect that successful AI adoption begins long before the first prompt is written.
It begins with understanding the work.
Comments