What happens when employees learn to build with AI faster than their organization learns to approve the work?
At Baptist Health South Florida, that question became a concern for Christine Kotler, Chief Marketing and Communications Officer. Her marketing and communications team spent the first year of its AI effort building literacy and confidence. Employees had a learning curriculum, time to practice, and peer support from colleagues known as “AI Sherpas.”
That investment helped people move beyond learning about AI to solving problems with it. In one example, the team built a dashboard that reduced roughly 10 hours of weekly work checking online listings to about two minutes of review.
But building useful tools brought a harder challenge: connecting them. As the team moved into optimizing and integrating what it had built, security and data concerns brought layers of approvals and governance reviews. Employees who were ready to advance their work found themselves waiting.
“We built this stuff, we’re ready,” Kotler says, describing their frustration.
In Episode 242 of The Artificial Intelligence Show, part of the AI Transformations series presented by Google Cloud, Kotler explains why she is raising this issue with colleagues in both IT and HR. She understands why a health system needs careful review. She also worries about what prolonged delays could do to the people who have become its most enthusiastic AI users.
Her warning gives leaders a different way to think about AI approvals. The cost of a slow decision may extend beyond a delayed project. It may affect whether the people capable of building the next one want to stay.
Getting employees to use AI is only part of the job. If training works, people will start asking for things the training program itself cannot provide: access to systems, permission to connect tools, and decisions about which uses can proceed.
That is a different problem from helping someone learn to prompt. An employee can be confident, skilled, and motivated while still being unable to move a project forward. More encouragement or another workshop won’t resolve a pending security decision.
Baptist Health’s experience makes that transition clear. Kotler describes the first year as a period of literacy, learning, and competency. The harder second-year work involves integrating tools the team has already built. The obstacle has changed, so the support employees need must change with it.
For leaders, this means looking beyond participation in training or enthusiasm for experimentation. Ask what employees are now trying to do, what requires approval, and where those decisions are getting stuck. A team can look successful on adoption measures while its most capable users are struggling to put their skills to work.
It also means being clear about what an invitation to experiment actually allows. Building something useful does not automatically make it safe to connect to other systems. But employees should be able to understand what has to happen next, who can make that decision, and what concerns they need to address.
Without that clarity, an organization risks asking people to take initiative without giving them a workable path to act on it.
The answer is not to remove safeguards because employees are impatient. Kotler explicitly recognizes the security and data concerns behind Baptist Health’s reviews. Her argument is that different kinds of work need different approval speeds.
She is advocating for a distinction between clinical AI applications that warrant a slower path and business or corporate support uses that could move more quickly.
“We have got to find a way to create lanes that are appropriately slow and appropriately fast,” she says.
That is a useful principle beyond healthcare. “AI” describes too broad a category to tell a leader how much scrutiny a particular project needs. The review should account for what the application does, what information it uses, what it connects to, and what could happen if it is wrong.
A marketing use case is not automatically low risk. Nor does a legitimate concern about one use justify treating every other use as equally risky. Business leaders need to work with their technical and security colleagues to make those distinctions explicit.
The goal should be a decision employees can act on. Approval lets them proceed within defined limits. A clear rejection tells them what cannot go forward. A request for changes gives them something to fix. An unresolved review leaves them waiting without knowing whether more work will help.
Kotler is pushing for faster routes where appropriate, not describing a problem Baptist Health has already solved. The leadership lesson lies in recognizing that the approval process itself needs attention as employees become more capable.
The employees who embrace AI may be doing exactly what their leaders asked: experimenting, improving work, and helping colleagues. If their efforts repeatedly stall, leaders need to understand how that experience is affecting them.
“The talent that we want to keep won’t sit here and wait forever,” Kotler says.
She is warning about the possibility of losing people if the organization cannot find ways to move faster. That concern makes approval delays relevant to HR, not just the teams responsible for reviewing technology.
For an employee, a stalled project can raise a larger question: Is this a place where I can use what I have learned? Leaders do not need to promise approval for every idea to take that question seriously. They do need to distinguish between a necessary limit and a process that leaves capable people unable to contribute.
Start with one stalled AI project. Ask the person building it what they are trying to accomplish. Identify the unresolved risk, who owns the decision, and what would allow the work to proceed or reach a clear stopping point. Then ask how the delay is affecting the employee’s willingness to keep trying.
An organization that invests in AI literacy should expect its people to want to use it. Before asking them to build more, make sure someone can decide what happens to the work they have already built.
Did you enjoy this transformation story? Go deeper on how AI is reshaping work and business with The Artificial Intelligence Show. Each week, we break down what matters in AI and what leaders should do about it.