--- title: "Why AI Rollouts Stall, and How to Restart One | Clustr" description: "AI rollouts stall when access is treated as adoption. The common causes, how to diagnose a stalled rollout, a 90-day restart plan, and what to look for in an AI adoption partner." url: https://getclustr.ai/blog/why-ai-rollouts-stall source: https://getclustr.ai/blog/why-ai-rollouts-stall --- # Why Do AI Rollouts Stall, and How Do You Restart One? AI rollouts stall when access is treated as adoption: licenses go out, a launch happens, and nobody owns the change in how work gets done. To restart one, measure where people actually are, pick a few high-frequency workflows, train each team on its own work, name champions, and report adoption rather than attendance. Stalls are normal. Almost every organization that rolled out ChatGPT Enterprise, Microsoft Copilot, Claude, or Gemini has seen the early burst of curiosity fade into a small group of heavy users and a long tail of people who tried it once. The good news is that a stalled rollout is easier to restart than a new one is to launch. The licenses, the security review, and the early adopters are already in place. ## What does a stalled AI rollout look like? You are probably stalled if several of these are true: - Usage spiked after launch and has been flat or falling since. - A handful of enthusiasts account for most of the activity. - Teams use the tool for quick questions but no recurring work has changed. - Leadership is asking whether the spend is worth renewing. - Nobody can say who owns adoption now that the deployment project is closed. ## Why do AI rollouts stall? Mostly for people and process reasons, not technology ones. BCG's 2024 research on AI adoption found that [74% of companies struggle to achieve and scale value from AI](https://www.bcg.com/press/24october2024-ai-adoption-in-2024-74-of-companies-struggle-to-achieve-and-scale-value), and that around 70% of the challenges stem from people and process issues, with about 20% from technology and 10% from algorithms. RAND's 2024 study of [why AI projects fail](https://www.rand.org/pubs/research_reports/RRA2680-1.html), based on interviews with 65 data scientists and engineers, put misunderstanding what problem the AI should solve at the top of its list of root causes. In a workplace tool rollout, those findings show up as: - **No problem definition.** People were told the tool exists, not which of their tasks it should change. - **Generic training.** One demo for everyone, disconnected from each role's real work. - **Unclear rules.** Without guidance on what data is allowed, careful people opt out. - **No owner.** IT's job ended at deployment. Adoption is ongoing and needs someone accountable. - **No measurement.** Without a baseline, nobody can tell whether the problem is skill, confidence, policy, or relevance. ## How do you diagnose a stalled rollout? Before you retrain anyone, find out where the stall actually is. Three inputs are enough: - **Usage data by team** from your platform's admin console. It shows where usage is zero, where it is shallow, and where it is healthy. - **A literacy baseline** for everyone with a seat. We use SINA, our AI literacy platform, which scores fluency by role. The gap is often not where leadership assumes. - **Short conversations with team leads** about their most repetitive weekly tasks and what is holding their teams back. Put the three side by side and the pattern usually becomes obvious. A team with high literacy and low usage has a relevance or policy problem. A team with low literacy and low usage has a skill problem. They need different fixes, which is why one more company-wide training session rarely works. You can start with our free [AI readiness assessment](/assessment). ## How do you restart an AI rollout in 90 days? A restart is a short, sequenced program, not another launch: - **Weeks 1 to 2: Diagnose.** Usage data, literacy baseline, and team lead interviews. Name an owner for adoption. - **Weeks 3 to 6: Role-based training.** Live, hands-on sessions per function on real documents, inside the tool you already license. Identify champions in every team. - **Weeks 7 to 10: Workflow builds.** Rebuild three to five high-frequency workflows with the teams who own them, and publish clear data guidance alongside them. - **Weeks 11 to 13: Re-measure and hand off.** Re-run the baseline, compare usage, report to leadership, and give champions a playbook for the next quarter. The restart works because it changes the thing that caused the stall. It gives people a reason to use the tool on work they already own, the confidence that they are using it safely, and peers who can help when they get stuck. ## What kind of outside help should you look for? Different partners solve different parts of the problem, and the right choice depends on where your stall is: - **Change management methodology firms.** Prosci is the best-known example. Its [ADKAR Model](https://www.prosci.com/methodology/adkar) breaks individual change into Awareness, Desire, Knowledge, Ability, and Reinforcement, and many organizations use Prosci's methodology and certification to build internal change management capability. A strong fit if you want a proven change framework and practitioners trained to run it. - **Large consultancies.** Useful when AI adoption is one part of a broad transformation that also spans strategy, operating model, and technology. - **Your platform vendor.** Vendors publish adoption material, such as Microsoft's [Copilot adoption hub](https://adoption.microsoft.com/en-us/copilot/). It is free and worth using, though it is written for every customer rather than for your roles and documents. - **Hands-on AI enablement partners.** Firms that train your people on their own work inside the tool, rebuild workflows with them, and measure the result. This is where Clustr works. Whichever you choose, ask every partner the same questions. How will you measure where we start? Will training use our real documents and workflows? Which platforms do you train on hands-on? What will be different at day 90, and how will we know? What stays with us when you leave? ## How does Clustr restart stalled rollouts? We run the 90-day pattern above as an [AI adoption consulting](/ai-adoption-consulting) engagement, or as part of a broader [AI enablement](/ai-enablement) program when the organization also needs governance, champions, and an operating model for AI. Every engagement starts with a SINA baseline and ends with a re-measure, so leadership sees adoption data, not attendance sheets. It is the same approach behind our client work. At Forvis Mazars, regional, CPE-accredited workshops trained 300+ professionals on Microsoft Copilot. At a 150-person data consulting firm, we trained 150+ people on ChatGPT and Copilot and helped establish an AI Center of Excellence. At Gray Media, tailored microtrainings ran across 12 regions with weekly national office hours to keep momentum going. Details are on our [case studies](/case-studies) page. ## Where do we start this week? Pull last month's usage report, name one person accountable for adoption, and ask the leads of your two lowest-usage teams what their most repetitive weekly task is. Those answers are the start of your restart plan. If you want a second set of eyes, [book an AI briefing](/contact). We will look at your stack, your usage, and your teams, and tell you plainly where the stall is and what it would take to fix. Ready to enable your teams with AI? Book a call to discuss your AI adoption challenges. We use cookies We use cookies and Google Analytics to understand how visitors use our site. [Privacy Policy](/privacy)