Gartner: Change Management Doubles Your AI Tech Costs

How Deepo Seth Sees Enterprise AI Implementation Challenges

Deepo Seth is a Gartner director analyst who has conducted over 650 client interactions on AI implementation in the past year alone. His perspective cuts through the hype with a practitioner’s clarity: organizations aren’t struggling with AI technology—they’re struggling with change management, governance, and the human side of adoption. This ThinkCast episode delivers actionable frameworks for CIOs navigating the gap between AI promise and AI reality.

On the real cost of AI: “Rolling out AI means as much as the technology cost of rolling out AI are the change related aspects—the change management—which could be 100 to 200% of the cost of the technology part.” This reframes AI budgeting entirely. If you’re planning a $1M AI deployment, plan for $2-3M total including change management, training, and governance.

On the wrapper problem: “You take a core AI capability—AI is very great at summarizing content. But how Sally in recruitment will use it is different from how Tom in procurement would use it versus how Sarah in marketing would use it.” Generic AI tools fail because they lack context. The real work is building “wrappers”—people, process, or technology layers that make AI contextually useful for specific roles.

On the 10-70-20 adoption reality: “10% of employees, you roll out anything new, they will figure out what to do with it. 70 to 80%, the bulk of your organization—I can’t figure out new things, tell me what to do. And then there are 10 or 20% whom you can give whatever and they will keep on doing whatever.” This distribution explains why “give everyone ChatGPT” strategies fail. Most employees need AI embedded in their workflows, not handed to them as a tool.

On agents as anthropomorphization: “What AI agents did—they anthropomorphized AI. Suddenly it became humanlike. People started thinking of AI in terms of human roles… But AI agents are not performing roles, they are performing tasks within roles.” This clarifies the agent hype: the concept helps adoption by making AI relatable, but confuses implementation by suggesting role replacement rather than task automation.

On the post-agentic future: “Agents are the current manifestation of the way AI is being rolled out. But down the road, I think there is definitely a post-agentic world. The intelligence may be built into the models, maybe built into the network, maybe built into the systems themselves.” Agents are a waypoint, not a destination—plan for a 2-3 year window where agentic architecture matters, then expect the paradigm to shift again.

6 Insights From Gartner on AI Adoption Reality

• Change management exceeds tech costs - Budget 100-200% on top of technology spend for training, literacy, governance, and adoption programs
• Defend/Extend/Upend framework - Defend (generic tools, employee satisfaction), Extend (workflow integration, measurable KPIs), Upend (transformative bets, hard to quantify)
• Context wrappers are the real product - Raw AI capabilities need role-specific and process-specific wrappers to deliver value
• Agents help adoption, confuse implementation - Anthropomorphizing AI makes it accessible but risks conflating task automation with role replacement
• Post-agentic world in 3+ years - Intelligence will move from agent architectures into models, networks, and systems directly
• ACI over AGI - The future isn’t artificial general intelligence alone, but Augmented Collective Intelligence—humans and AI working together

What This Means for CIOs Planning AI Rollouts

Seth’s most important insight is reframing AI implementation from a technology problem to a change management problem. Organizations aren’t failing because models aren’t capable—they’re failing because they’re deploying AI without the wrappers that make it contextually useful. For organizations building or adopting AI agents, this means the agent itself is perhaps 30-50% of the work; the rest is training, governance, workflow integration, and helping the 70% of employees who need AI embedded in their jobs rather than handed to them as a separate tool.