As companies rush to deploy artificial intelligence, the real bottleneck is not software. It is whether people actually use it, trust it, and change how they work. That is the argument innovation expert Soren Kaplan has been making as organizations discover that buying licenses is not the same as adoption.
Kaplan, a Wall Street Journal author and founder of InnovationPoint, has studied AI rollout across large enterprises.
His conclusion is blunt: most firms are earlier in the journey than their dashboards suggest.
Counting activated accounts or completed training modules gives a false sense of progress.
Real adoption shows up only when tools are embedded in daily workflows, informed by useful data, and supported by new habits.
Few organizations have reached that point. A handful of standouts exist, but they remain exceptions.
The distance between leaders and everyone else is growing. The pattern is consistent.
Teams often score well on access and attitude. Employees can open the tools and many want to learn.
The weak spots are processes, skills, and the ability to feed AI the information it needs.
Without documented workflows and a shared understanding of how work actually gets done, even capable systems stay at the edges of the job.
Kaplan and communications strategist Melanie Barna, who have compared notes across Fortune 500 and Fortune 1000 companies, describe this as an organizational problem rather than a technical one.
That distinction matters for how leaders manage the change.
Treating AI like a traditional project, with a fixed destination and a one-time training burst, misses the nature of the technology.
Capabilities keep shifting. What people need to learn keeps shifting with them.
Kaplan argues that progress depends on a continuous cycle of experimentation and support, not a rollout that ends when the slide deck is delivered.
Leadership behavior is the lever.
Kaplan points to companies where executives stop presenting polished success stories and instead describe what they personally tried, where the tools failed, and what they deleted before anything useful appeared.
That kind of candor reduces fear and makes experimentation legitimate.
Communications teams, he says, are often better placed than they realize to frame the work around real business problems instead of leaving the narrative to IT or HR.
When they do, AI stops being something that happens to employees and becomes something they can shape.
The practical advice is equally unglamorous. Stop waiting for a perfect enterprise strategy.
Map one process from start to finish. Make audiences and workflows explicit enough that a model can use them.
Then connect those efforts to outcomes leaders already care about, such as faster, more relevant internal messages.
In one example discussed alongside Kaplan’s research, tailored manager communications lifted click-through rates from 20 percent to 80 percent in a day.
Results like that do more to build confidence than another generic demo.
The larger claim is simple. AI will not deliver value on its own.
Value appears when people stay engaged: when they feel safe enough to try, when leaders model imperfect learning, and when the organization treats adoption as culture and operations rather than procurement.
Firms that keep the human side at the center are the ones most likely to close the gap between what the technology can do and what actually happens at work.