I recently attended an NAW executive roundtable with leaders from large distributors, and one thing stood out to me.
Very little of the most interesting conversation about AI was actually about the technology.
The discussions kept coming back to people. Who owns the initiative? How is it introduced to employees? How do you build trust with customers? What happens when the first version is not perfect? How do you keep changes from overwhelming the organization?
Several leaders in the room described change management as at least half of the work involved in an AI initiative. That tracks closely with what we have seen ourselves.
Getting an AI system technically ready is one challenge. Getting a business to actually use it, trust it, improve it, and ultimately change the way work gets done is another.
For many distributors, change management is becoming the harder part.
The Technology Can Be Ready Before the Organization Is
Distribution businesses are particularly sensitive to change because so much operational knowledge has accumulated over many years. A customer service employee who has worked with the same accounts for 15 years knows things that may never have been formally documented. Salespeople have built relationships around particular ways of working. Customers may have called the same number, sent orders to the same email address, or spoken with the same person for years.
Then a new technology arrives and changes part of that process.
Even if the technology works, people have reasonable questions. Will it make mistakes? What happens when something unusual occurs? Is this eventually replacing my job? Will my customers accept it? Who is responsible if something goes wrong?
Those questions cannot be solved with a better product demo. They have to be addressed through the way the change is introduced.
One of the more useful ideas that came up at the roundtable was putting skeptics into the pilot group instead of filling it entirely with people who already love technology. Enthusiasts are usually willing to forgive problems because they want the experiment to succeed. Skeptics ask harder questions.
That can be valuable.
If the people who were initially resistant become advocates, their support often carries far more weight with the rest of the organization.
Start With Work People Already Want Off Their Plates
Another theme I heard was the importance of letting the business identify the problems worth solving.
One distributor had asked business unit leaders to identify their most tedious and time-consuming tasks, then ranked those opportunities before deciding where to apply AI.
That is very different from starting with a piece of technology and asking everyone where it might fit.
It also changes the adoption conversation. People are far more receptive when AI is being applied to work they already wish they did not have to do. The conversation becomes less about replacing jobs and more about removing repetitive tasks.
A simple question can make that distinction clear: Who here has enough time to do everything they need to do?
Usually, nobody raises their hand.
In distribution, there is almost always more valuable work available. Salespeople could spend more time with customers. Customer service teams could spend more time resolving complex issues. Operations leaders could spend more time improving processes instead of reacting to them.
The goal of AI should be to create capacity for that work.
That framing matters because employees rarely resist technology simply because they dislike technology. They resist when they do not understand what the change means for them.
Change Management Extends to the Customer
One thing I think is sometimes missed in conversations about AI adoption is that, in distribution, employees are only half of the change-management problem.
Your customers are part of the rollout too.
A distributor may be completely comfortable with a new workflow internally, but that does not mean customers will immediately trust it. If someone has placed orders the same way for years, forcing them into an unfamiliar experience can create unnecessary resistance.
We have found that choice works much better than force.
In one rollout, the new ordering experience initially handled after-hours activity. That created a relatively low-risk environment for customers to try it. Over time, confidence grew. Some customers eventually began choosing that option specifically because it was convenient.
That progression is important.
You do not always need customers to change their behavior on day one. Sometimes the best way to introduce new technology is to let them experience the value gradually.
Language matters too.
Terms that make perfect sense inside the technology industry may create unnecessary uncertainty for customers. “Automated order line” can feel much more familiar and understandable than “AI agent.”
The technology can be sophisticated behind the scenes while the customer experience remains simple.
Go-Live Is the Beginning of the Feedback Loop
Another mistake is treating deployment as the finish line.
The first few weeks of any meaningful operational rollout are a learning period. You discover edge cases. Employees identify workflows nobody considered during testing. Customers behave differently than expected. Small issues surface that seemed insignificant in a controlled pilot but matter when the system is handling real work.
That should be expected.
What matters is having a clear process for what happens next. Who owns the feedback? Who decides whether a requested change should be made? How quickly can the team respond? How do you make sure something fixed in one release does not break again in the next?
These questions sound tactical, but they are central to trust.
If employees report an issue three times and continue seeing it, they eventually stop reporting it. If customers encounter a problem and believe nothing changes, they revert to the old process. Once trust is lost, technical performance alone may not bring it back.
This is why AI deployment has to be treated as an ongoing operating partnership rather than a traditional software implementation.
Launch one workflow. Learn from it. Improve it. Then take on the next one.
Too Much Change Can Be as Bad as Too Little
One of the more interesting points raised at the roundtable was change fatigue.
A company may have dozens of good ideas for improving a workflow. Individually, each one seems small. But pushing 10 or 15 changes into the organization every week can make people feel as though the process is constantly moving underneath them.
Even changes employees originally requested can become difficult to absorb when they arrive too frequently.
Sometimes the better approach is to bundle improvements into fewer, more meaningful releases. Give people time to learn one version of the workflow before changing it again.
This is particularly important with AI because the technology can improve quickly. The temptation is to continuously push every improvement as soon as it becomes available.
But the organization has its own adoption speed.
Successful change management has to respect both.
Someone Still Has to Own the Change
This connects back to another theme that came up repeatedly: ownership.
Many distributors already have AI activity scattered across the organization. Someone is experimenting with Copilot. Another team has a data project. A business unit is testing a new workflow. Innovation is happening, but it may not have a clear direction.
Eventually, someone has to decide what is worth scaling and what is not.
That does not necessarily require a large AI department. It does require clear responsibility.
Someone should own the roadmap, evaluate which projects are realistically deployable, establish policies around approved tools and data, and make sure successful pilots have a path into production.
Otherwise, an AI initiative can easily become a good ideas department: lots of interesting activity without enough operational impact.
Final Thought
The technology behind AI will keep improving, and many capabilities that feel new today will become commonplace surprisingly quickly.
That may actually make change management more important, not less.
The distributors that get the most value from AI will not necessarily be the ones that find the most tools or launch the most pilots. They will be the ones that become good at introducing change without losing the trust of their employees or customers.
Start with a problem people actually want solved. Bring skeptics into the process early. Talk about tasks rather than jobs. Give customers choice. Set expectations that the first few weeks are a learning period. Create a clear feedback loop. And make sure someone owns what happens after launch.
Because the hardest part of AI in distribution is often not getting the technology to work.
It is getting the organization to change with it.
← Back to all articles