Table of contents
The future of enterprise AI isn’t being shaped by how much organizations invest. It's being shaped by whether employees are actually able to use AI in meaningful, responsible ways.
Research shows that AI usage is rising far faster than organizational readiness. Three in four knowledge workers now use AI at work, yet AI literacy remains uneven, and many employees are learning without guidance or structure.
Employees are moving even faster than most organizations can keep up. According to EY, 85% of employees are now learning how to use AI outside of work, and 83% say the skills they have today are entirely self-taught.
This bottom-up fluency is accelerating the move toward a distributed, employee-driven AI enterprise, where experimentation often starts at the edges before formal programs take shape. And the impact is already visible.
Agentic AI (systems that can understand goals, orchestrate tasks across systems, and act within policy boundaries) is redistributing power across the organization, turning AI from a centralized initiative into a business-wide capability that frontline teams can design, deploy, and iterate on themselves.
Leaders are sensing this shift: 91% of executives say non-technical employees are already driving AI adoption.
The new organizational reality is clear: the role of technology and business leaders is no longer to build every agent or control every initiative. It’s to enable a decentralized ecosystem that allows safe, governed innovation to happen everywhere.
The data confirms it: AI influence is shifting to the frontline
For decades, IT held a near-monopoly over enterprise technology decisions. That era is ending, not because IT has become less important, but because AI innovation scales faster when it’s distributed.
The power vacuum: Influence is moving beyond IT
Only 38% of executives believe IT will remain the most influential function in AI agent development over the next three years. Who’s taking the lead instead?
- Operations (56%)
- HR (52%)
- Finance (47%)
These teams deeply understand the business workflows that agentic AI aims to optimize.
Bottom-up momentum is winning
Nearly 78% of executives believe that the best agentic AI projects should be allowed to originate anywhere in the company, not just from leadership or IT.
This is the new reality: employees closest to the problems are proving to be the best problem-solvers. Coca-Cola Consolidated saw this firsthand when non-technical business users helped to build working AI solutions during an internal hackathon, demonstrating that teams embedded in day-to-day operations often generate the most practical and high-value automation ideas. And it’s beginning to pay off.
Enterprises report that AI success is increasingly defined by outcomes like process reinvention, cited by 53% of executives, and the unlocking of entirely new capabilities, highlighted by 47%.
Together, these results show that the real impact of AI comes from empowering the people who understand the work best to redesign it.
IT’s new mandate from builder/owner to enabler/partner
This decentralization doesn’t diminish your importance. It helps elevate it.
Instead of being a guardian, you become the platform that powers enterprise-wide intelligence. Here’s what that evolution looks like.
Role 1: The secure platform enabler
Your new mandate: Ensure you have the infrastructure that empowers others to innovate safely. To securely use agentic AI, frontline builders need a foundation that includes:
- Govern and manage data access, so employees see only what they’re permitted to access.
- Secure integration points into systems and tools, ensuring employees can trigger actions safely.
- Identity and permissioning layers that enforce role-based access when employees build or use agents.
- Agent governance and monitoring, so employee-built automations operate within enterprise policies.
- Reasoning APIs and orchestration tools that allow teams to design and run AI-driven workflows without exposing sensitive systems.
And here’s a crucial mindset shift: 96% of executives prefer a useful AI tool over access to the latest model. Utility wins over novelty, and IT controls the utility.
When IT secures the foundation, non-technical teams can prototype, test, and deploy their own agents without compromising enterprise integrity.
Role 2: The governance and compliance partner
Decentralization only works if there are guardrails. IT owns the governance architecture that ensures:
- Transparency
- Ethics
- Security
- Regulatory alignment
- Explainability
- Auditability
According to the governance research:
- Safeguard data by balancing innovation with privacy and compliance.
- Embed fairness and transparency, ensuring outcomes remain equitable and explainable.
- Build cross-functional governance by aligning HR, IT, Legal, and Security.
- Build a culture of AI use by fostering literacy, trust, and responsible adoption in everyday work.
The cultural shift: Why control is the enemy of scale
Traditional IT models were built on control, but the decentralized AI enterprise demands co-creation.
The data is unambiguous: 78% of successful AI projects now originate from non-leaders, and 73% of executives believe their companies underestimate the cultural shift required to scale AI.
The message is clear: innovation can’t be forced from the center. It has to be supported at the edges.
For you as CIO, this means leading a cultural reboot. Your role evolves from being the “permission gatekeeper” to acting as a mentor, an enabler, a policy architect, an infrastructure provider, and ultimately, a leader of trust.
The winning model is fully collaborative: IT defines the secure foundation, business teams reimagine the workflows, and governance keeps everything aligned and safe.
The leadership action plan for decentralized AI governance
To thrive in the agentic era, organizations need a governance model that accelerates innovation rather than slows it down.
Traditional approaches — centralized approval, heavyweight committees, and long review cycles — weren’t built for a world where employees are experimenting with AI daily. The speed and distributed nature of adoption demand a more adaptive system.
A new model is emerging: lightweight, federated governance. It sets the standards for responsible AI use while giving teams across the business the freedom to build safely within shared boundaries.
This isn’t about loosening oversight. It’s about moving from “approving every idea” to creating the conditions where safe experimentation can flourish through guardrails, transparency, and continuous visibility.
Here’s how leaders can put this into practice.
Step 1: Rally your champions
Decentralized AI requires a different kind of change agent.
Instead of relying on a small centralized team, organizations need distributed champions embedded across HR, Finance, Operations, Customer Experience, and other frontline functions. These are the people who understand the work intimately, the workflows, the edge cases, the friction points, and are best positioned to identify where AI can make the greatest impact.
But champions don’t simply appear. Leaders need to intentionally identify and equip them, ensuring they have:
- A clear understanding of what “responsible” experimentation looks like
- Access to the foundational infrastructure and datasets
- A forum to share learnings and escalate risks early
- Support from IT and security teams when they encounter blockers
These early adopters become credibility builders, pilot owners, and cross-team connectors, demonstrating what “good” looks like and helping the rest of the organization gain confidence.
Bottom-up change doesn’t happen in isolation. It happens when people who are already close to the problems feel empowered, trusted, and supported to redesign how work gets done.
Step 2: Expand with intention
As champions deliver early proofs of concept, it’s tempting to scale immediately, but sustainable growth requires discipline. The organizations that succeed expand deliberately, ensuring each agent aligns with their strategic goals and risk profile.
Three principles guide this phase:
1. Apply guardrails gradually.
Start with essentials: data access controls, permissions, provenance, and explanation. Add more safeguards as adoption grows, avoiding friction that slows momentum.
2. Prioritize workflows with clear impact.
Teams will surface dozens of ideas. Focus on those tied to measurable outcomes — shortened cycle time, fewer handoffs, faster decisions, or new capabilities previously out of reach.
3. Ensure every agent connects to enterprise policy and data.
Fragmentation is the enemy of scale. Agents that can’t honor identity models, data boundaries, or privacy requirements shouldn’t move forward, regardless of their potential.
When done well, expansion feels organic: teams build what they need, within a shared fabric that keeps everything consistent and secure.
Step 3: Refine through real-time feedback
As usage grows, governance must evolve from static documents to living infrastructure. Leading organizations treat governance as a continuous, adaptive feedback system that surfaces insights and guides iteration.
This includes:
- Performance analytics that show how agents are being used
- Usage telemetry that reveals adoption patterns
- Drift detection that identifies when an agent’s behavior begins to diverge
- Decision logs that provide traceability and accountability
- Compliance signals that help maintain alignment with regulatory standards
These feedback loops are essential for maintaining trust. They ensure leaders can see what’s working, what’s not, and where additional training, adjustments, or guardrails may be needed.
In this model, governance becomes always on and always learning, providing clarity and structure without slowing the pace of innovation.
The agentic AI era is an opportunity for enterprise leaders
The shift toward decentralized, employee-driven AI is a redefinition of how organizations create value. What makes this moment uniquely important is the speed at which AI is moving from controlled pilots to everyday decision-making, often faster than organizations can redesign their operating models.
Multiple studies confirm this acceleration.
- McKinsey found that genAI adoption across organizations has doubled, jumping from 33% to 65% in a single year.
- And according to Gartner, by 2026 more than 80% of enterprises will have citizen developers building automations, up from just 20% in 2023.
What’s emerging is a new dynamic: AI has become a shared capability that every function expects to use and contribute to. This creates both a challenge and an opportunity:
The challenge
AI is spreading faster than the guardrails required to manage risk, data access, and consistency. Shadow AI, fragmented tools, and uneven literacy introduce governance gaps that leaders can’t ignore.
The opportunity
The people closest to the work are now positioned to redesign it. Employee-driven experimentation is generating new workflows, faster cycle times, and entirely new capabilities that would never emerge from a centralized model alone.
And leaders are already seeing the upside. A recent Accenture survey found that organizations empowering distributed AI builders achieved a 3.5× higher rate of innovation, along with significantly stronger trust and adoption. Meanwhile, 48% of executives report that their most successful AI initiatives in the past year originated outside traditional leadership channels.
The message is clear: AI leaders who shift from gatekeeping to enabling, who focus less on controlling every build and more on creating the secure conditions for innovation, will unlock far greater scale, creativity, and business impact.
This moment is unique because the stakes are high on both sides. What can be gained: accelerated transformation, faster innovation cycles, and a culture of empowered problem-solvers.
What can be lost: control of data, inconsistent outcomes, and a fragmented AI ecosystem lacking governance. Leaders who embrace this shift will turn AI from a departmental capability into an enterprise operating layer, one that fuels innovation everywhere, not just at the center.
How Moveworks helps CIOs enable decentralized, agentic AI
Moveworks helps accelerate this transformation by offering a platform designed with enterprise-grade security: one that empowers leaders to standardize trust while decentralizing innovation.
Moveworks provides:
- A secure enterprise AI platform that lets non-technical teams build and deploy agents safely
- Federated governance built into interactions
- A unified reasoning layer for orchestrated, multi-agent autonomy
- Pre-configured guardrails for permissions, privacy, and compliance
- Real-time analytics for adoption, performance, and ROI
- Cross-department orchestration across HR, IT, Finance, and Operations
The Moveworks AI Assistant helps give a foundation to standardize trust, empower business units, and scale innovation, without sacrificing control.
By decentralizing AI responsibly and shifting from AI builder to AI platform enabler, you can create organizations where everyone can innovate safely and Moveworks AI Assistant makes that future possible.
To go deeper into the employee-driven AI shift, including data on frontline experimentation, governance models, and how leaders are adapting, download the full report.
