You’ve invested in the tools. You’ve bought the licenses. You may have even hired a consultant to handle some light training or coaching. And yet, six months later, your AI initiative is stalled, underused, or quietly declared a loss. Or maybe you don’t even know if it’s working, though you do know you’ve already blown your IT budget for the year.
Don’t worry – you’re not alone! Though do worry a bit, since it’s not the tool’s fault…
Catie Maillard, Global Head of People at Data Society Group, has watched this pattern play out across organizations of every size. Her diagnosis is consistent: most AI projects fail because of the humans around the technology, not the technology itself.
“Not having people champion change, not having executive guidance or executive sponsorship, forgoing guardrails – that’s where you see a lot of these efforts fall short. And the biggest issue is when folks are creating change in a silo.”
Understanding where AI projects go wrong, and why, is the first step toward building initiatives that actually stick.
NO INTERNAL CHAMPION
Every successful change management initiative has people behind it who are responsible for making sure it doesn’t die. This is not necessarily the CEO or CTO, or even the most senior leader on the team. It’s often a long tenured coworker, a team lead, or an operational leader within a team acting as a champion for change. Without that champion, AI efforts quietly fade into the background.
“That champion doesn’t need to be an executive. That just needs to be someone internally who is a leader, who people listen to, who adopts tools, who knows the workflows and can really figure things out.”
Identifying and empowering an internal champion before you launch is not optional. It is the single most practical thing you can do to increase the odds that an AI initiative survives contact with the real world.
MISSING EXECUTIVE SUPPORT
Champions build momentum from the ground up, and this needs to be coupled with sponsorship from the top down. Executive sponsorship unlocks budget, tool access, cross-team alignment, and the organizational credibility that signals this initiative actually matters.
“If an employee needs a tool, they need their executive to approve it. They need approval for the ability to connect this tool to other systems and approval for development or consulting resources to make it work. And Executives are responsible for making sure employees experiment and implement safely, cost-effectively, and at-scale”
Executive sponsorship is not just about budget. It is a visible signal to the rest of the organization that this work is serious. For more on what responsible, outcomes-driven AI adoption looks like in practice, see Data Society’s AI Advisory Services overview:
https://datasociety.com/data-society-launches-ai-advisory-services-to-support-responsible-outcomes-driven-ai-adoption/
SILO-BASED ADOPTION
An AI pilot that succeeds in one team and goes nowhere else is not a transformation. It is an experiment. Real organizational change requires cross-functional coordination, shared learnings, and visibility that does not happen when teams are working in isolation.
“The biggest issue is when folks are creating change in a silo, which is what often happens in areas where there’s not a lot of psychological safety and there’s generally a lot of fear.”
The antidote to silo-based adoption is a deliberate cross-functional structure. Understanding the relationship between AI upskilling and AI implementation is a useful starting point for organizations trying to break out of the pilot trap:
https://datasociety.com/ai-upskilling-vs-ai-implementation/
FEAR AND PSYCHOLOGICAL SAFETY DEFICITS
The news cycle around AI and job displacement has done real damage to employee trust. Workers who believe they are training their replacement will not experiment openly with AI tools.
“I’m seeing mandates that are: ‘get this done or else you’re fired.’ Sure, we’ll see some short-term gains because people are scared and they have bills to pay. But long-term, usually you find that people develop in a silo and they’re not sharing information because they want to keep their jobs.”
Psychological safety is not a soft concept. It is a precondition for innovation.
NO FOLLOW-THROUGH AFTER THE PILOT
Successful AI pilots happen every day. The harder question is what comes next. When a pilot succeeds but no change management infrastructure exists to scale it, the initiative plateaus.
Organizations that scale AI successfully treat the pilot as the beginning of a process, not the end of a project. The 2025 AI Readiness Report from Data Society offers a detailed look at why most AI initiatives stall before they scale:
https://datasociety.com/the-2025-ai-readiness-report-insights-to-build-your-2026-strategy/
HOW TO SET YOUR AI PROJECT UP FOR SUCCESS
The organizations succeeding with AI are not necessarily the ones with the most sophisticated tools. For a practical framework on prioritizing where AI actually belongs in your organization before you launch, see AI Isn’t the Problem. Prioritization Is.:
https://datasociety.com/ai-isnt-the-problem-prioritization-is-heres-where-most-organizations-get-it-wrong/
THE BOTTOM LINE
If your AI initiative has stalled, the first place to look is not your tech stack. Look at your people structures: who is championing the work, whether leadership is visibly invested, and whether your teams feel safe enough to try and fail.
“Psychological safety needs to be at the heart of all of these transformations. If companies are looking to be around for a while and really see how they can use AI to transition to the workplace of the future, they need to focus on their people.”
READY TO BUILD AN AI INITIATIVE THAT ACTUALLY STICKS?
Watching an AI initiative stall inside your organization? Data Society works with HR and people leaders to diagnose what is getting in the way and build the people-centered foundation your AI strategy needs. Get in touch with our team to start the conversation:
https://datasociety.com/contact/
RELATED RESOURCES FROM DATA SOCIETY
– The 2025 AI Readiness Report: https://datasociety.com/the-2025-ai-readiness-report-insights-to-build-your-2026-strategy/
– AI Upskilling vs. AI Implementation: Why Enterprises Need Both: https://datasociety.com/ai-upskilling-vs-ai-implementation/
– AI Isn’t the Problem. Prioritization Is.: https://datasociety.com/ai-isnt-the-problem-prioritization-is-heres-where-most-organizations-get-it-wrong/
– Data Society AI Advisory Services: https://datasociety.com/data-society-launches-ai-advisory-services-to-support-responsible-outcomes-driven-ai-adoption/
Frequently Asked Questions
Most AI projects fail due to people and organizational issues rather than technology problems. The most common causes include a lack of internal champions, missing executive sponsorship, low psychological safety that prevents open experimentation, and adoption efforts that stay siloed within one team.
The single most common root cause of AI project failure is the absence of a dedicated internal champion. When no one owns the initiative day-to-day and keeps it connected to both executive leadership and frontline employees, even well-funded AI projects lose momentum.
Yes, poor change management is one of the leading causes of AI project failure. AI transformation requires ongoing communication, workflow redesign, updated performance expectations, and a culture that supports experimentation.
The most effective prevention strategies are human-centered. Identify and empower an internal champion before launch. Secure visible executive sponsorship. Build psychological safety so employees feel comfortable experimenting and sharing failures. Create cross-functional structures that allow learnings to move across teams. And treat the pilot as the beginning of a continuous change management process, not a one-time rollout.
HR is often the canary in the coal mine when AI projects are struggling. Signs like increased burnout, disengagement, or fear-driven behavior are often visible to HR before they surface in productivity metrics.
