The promise of the benefits of artificial intelligence has boards and C-level executives at companies from venture funded startups to Fortune 500s all asking their leaders the same thing: how is your team using AI? This pressure from the top, as well as the endless social media noise about the latest AI tools and possible efficiencies, is creating an atmosphere of anxiety inside of companies where the spend on large language models is going up, sometimes at the expense of additional headcount, however adoption is fragmented. Here’s a typical playbook: company buys licenses to one or more of the LLMs either Claude, ChatGPT, CoPilot, Gemini, or all of the above, they run a few lunch-and-learn sessions where people share what’s working and what’s not, and the result is a few super users who are trying to push the envelope of what’s possible with no direction, and a whole lot of skeptics who are asking, “how is this actually useful for me?” After talking to dozens of functional area leaders and individual contributors we have found that there are three real barriers to AI adoption, and to any measurable outcomes that executives are looking for: technical setup gaps, use-case education, and governance clarity. Technical Setup Gaps When teams want to go beyond prompting, they very quickly hit a gap when it comes to data access and quality, as well as basic integrations and permissions. Automation is only possible when your internal systems know how to talk to each other through connected APIs via your engineering team or no code workflow automation tools like Zapier and Make.com . Additionally, some teams we’ve talked to are going through the painful process of data clean-up in their CRMs and document management systems to make sure that whatever systems AI plugs into are reflecting business reality. Use-case Education One-off workshops and even weekly sessions are a good start for basic awareness and as a kick-off point for experimentation with AI tools across the company. But leaders are quickly seeing that without clear role-based directives they’re leaving the majority of employees more frustrated than anything else. This is also not something that HR or your talent development team can easily solve as real adoption requires domain expertise within each team to make it a reality. The next step is to have each team define what manual tasks should be automated first, then either develop internal expertise on a team-by-team basis with “AI experts” in sales, marketing, customer success, operations, etc. that develop solutions themselves then teach the rest of the team what to do, or bring in external functional area experts to teach what they’ve already done at other companies. Companies like School16 are starting to offer this type of role-based training. Governance Clarity While companies are offering licenses and basic adoption mandates, many lack clearly defined rules about what data can go into which tool, and where employees have autonomy and where they don’t. Often the safest move for employees is to use AI only for low-stakes tasks and avoid anything that touches sensitive data. Security is a real concern, but employees can’t feel like they’re at risk of getting fired anytime they use AI. One first step that companies can take is to create a data classification guide that outlines what data can go into public AI tools and what is allowed with an enterprise-grade license. This should be followed up with specific use-cases shown by VPs and other company leaders who can demonstrate how they use company data with AI tools for their own work to then empower others in the company to do so themselves. We’re a little over 3 years into the proliferation of large language models and in many cases the impact of AI on board room conversations has been greater than on true enterprise efficiency. That said, the rules of how companies operate are being rewritten in real time and from a cultural and structural perspective, these things take time. If we want to do right by our employees and smoothen what is likely to be a significant transition across every single discipline, we must go from a mandate-driven approach to a capability-informed approach where our people are empowered by real direction and easy access to information that gives them the confidence to transform our companies for the future.