How to Govern AI Agents in the Enterprise Without Slowing Innovation

Do you know what AI Agents are running across your estate?

If the answer isn't a confident YES, then AI governance shouldn't be an afterthought, it should be your next strategic priority.

In this episode of the Inspiring Tech Leaders podcast, I look at how to govern AI agents in the Enterprise, without slowing innovation.

This podcast episode covers: 

💡 You can't govern what you can't see. Implement consolidated discovery tools to scan across all clouds and create a single source of truth.

💡 Autonomous systems can't be legally or ethically accountable.  Every agent needs a designated human owner responsible for its performance and compliance.

💡 Stop paying for 3-4 different agents that do the same job. Use estate audits and visualisers to identify dormant or redundant assets and reinvest the savings into true innovation.

Stop guessing and start controlling your AI ecosystem. 

Available on: Apple Podcasts | Spotify | YouTube | All major podcast platforms

Start building your thought leadership portfolio today with INSPO.  Wherever you are in your professional journey, whether you're just starting out or well established, you have knowledge, experience, and perspectives worth sharing. Showcase your thinking, connect through ideas, and make your voice part of something bigger at INSPO - https://www.inspo.expert/

On the Balance Sheet®
Interviewing executives from community banks and credit unions about key economic issues.

Listen on: Apple Podcasts   Spotify

Chapters

• 0:00 Introduction
• 1:14 Setting the Scene
• 2:00 The Problem of Visibility
• 2:57 The Risk Side
• 3:20 [Ad] On the Balance Sheet®
• 3:55 (Cont.) The Risk Side
• 5:01 Towards Central Governance
• 5:51 Ownership and Accountability
• 6:40 Managing Cost and Redundancy
• 7:16 The Culture of Governance
• 7:59 Wrap Up

Transcript

Welcome to the Inspiring Tech Leaders podcast, with me Dave Roberts.  This is the podcast that talks with tech leaders from across the industry, exploring their insights, sharing their experiences, and offering valuable advice to technology professionals.  The podcast also explores technology innovations and the evolving tech landscape, providing listeners with actionable guidance and inspiration.

Today, I’m looking at a topic that’s suddenly very real for every large organisation embracing AI, which is how do you effectively manage AI agents at scale?

AI agents are autonomous or semi-autonomous pieces of software that carry out tasks on behalf of people.  These are no longer futuristic experiments, they’re proliferating across departments and clouds, sprouting up wherever teams find business value.  But that rapid growth brings its own challenges.  Without clear governance, Technology leaders are facing a kind of digital sprawl, with a tangled estate of AI agents that are costly to run, risky to monitor and extremely hard to control.

Today I will look at why this agent sprawl matters, how it affects risk, cost and innovation, and most importantly, what leaders can do to govern and turn complexity into competitive advantage.

So let’s start with the scale of the challenge. Analysts now forecast that the number of active AI agents deployed in enterprises could exceed one billion by 2029, that’s a forty-fold increase from what we see today.  These aren’t just bots sitting in some isolated sandbox.  They’re operating across your various departments and often without central visibility.

It’s reminiscent of the “shadow IT” of the cloud era, where teams bypassed central processes to get innovation out the door.  But with AI agents, they can execute logic, touch sensitive data and even make business decisions, the potential impact is considerable.  Visibility isn’t just an operational nicety, it’s a foundation for security, cost control, compliance and trust.

At the heart of effective management is visibility.  You can’t govern what you can’t see.  A potential scenario could be that a Marketing team builds chatbot agents on one cloud platform while a Procurement team builds data-processing agents on another.  Without consolidated discovery, central IT has no unified view.

Modern tools are trying to address this with automated discovery, scanning for agents running across major ecosystems such as Salesforce, AWS or Google, and pulling metadata about their capabilities, the models they use, and the data they access.  Turning that into a standardised profile allows governance tools to normalise, compare and manage assets regardless of vendor or platform.

This matters not only for security teams but for operations and finance.  Imagine trying to verify which loan-processing agent has permission to touch financial records when that information is scattered across multiple untracked deployments. Manual chasing of documentation simply doesn’t scale.

Unmanaged agents expose organisations to real risk.  From a security perspective, they can become blind spots.  Agents that aren’t centrally tracked might access sensitive data long after they’ve outlived their usefulness, creating compliance challenges or even regulatory exposure.  Teams may not know who built an agent, what data it touches, or whether it still behaves as intended.

From a risk management lens, let’s flip to the cost side. With multiple regional teams independently building or buying similar tools, redundancy becomes extremely common.  It’s not unusual for large enterprises to be paying for three or four different summarisation agents that effectively do the same job just because no one had a consolidated view.

Leveraging a visualiser to filter and compare the estate by job type can highlight these redundant assets. Technology leaders can then consolidate, rationalise spend, and reinvest savings into new capabilities or innovation.

Which brings us to governance.  Governing AI agents isn’t a box-ticking exercise. It’s a continuous commitment to visibility, compliance, accountability and optimisation.  It begins with automated scanning and moves into establishing a baseline of truth, to create an inventory that includes every AI agent whether it’s bought, built or inherited through acquisition.

A governance policy should mandate that every agent exposes its purpose, capabilities and data privileges in standard formats that systems can parse.  This prevents the situation where teams deploy tools without oversight and makes it easier to monitor activity across multi-cloud landscapes.

But governance also means setting the rules of engagement.  This means looking at who owns each agent, who is accountable for its behaviour, and how exceptions or incidents are handled.

A key part of effective governance is assigning human ownership to every AI agent. Autonomous systems may perform tasks, but they can’t be accountable in any legal or ethical sense.  Only humans can.  Every agent needs a designated owner responsible for its performance and compliance. This person becomes the point of contact when issues arise, and the steward of outcomes that matter to both your business and regulators.

Towards this end, governance isn’t just about central IT, it’s about cross-functional coordination.  Be it security, compliance, audit, risk, or HR, all stakeholders have roles in shaping how agents behave in context.  Expanding your governance councils to include heads of data, legal, and ethics functions, creates a space where policy becomes practical and enforceable.

When we talk about cost management, it’s not just about license fees.  Agents that sit idle are still consuming compute, API quota allocations, memory footprints and potentially expensive model calls.  Regular estate audits help identify dormant agents that should be decommissioned or redeployed.

Admins should also build dashboards that help track performance metrics: which agents drive business value, which are redundant, and which are underperforming. A good governance framework has a feedback loop, it doesn’t just enforce policy, it measures adoption, impact and opportunity cost.

Turning governance into a strength, rather than a drag on innovation, means treating it as a culture and capability, not just a set of rules.  Good governance accelerates innovation by providing guardrails that enable teams to experiment without exposing the organisation to outsized risk.

Think of AI governance the way you would safety protocols in a factory.  Without them you might build something quickly, but you risk injury, liabilities and unforeseen friction as you scale.

This also means thinking about how humans and agents collaborate. Clear escalation paths, human-in-the-loop checkpoints, and daily oversight aren’t bureaucratic red tape.  They’re practices that make complex multi-agent systems trustworthy and reliable.

As organisations move from pilots to broad deployment of AI agents, the biggest differentiator won’t be how many agents you deploy, it will be how coherently they operate as an ecosystem.  Coherence comes from visibility, governance, accountability, and continuous lifecycle management. 

Effective governance should give your organisation confidence that AI agents are doing the right work, with the right data, under the right oversight, every time. It should help eliminate surprises, manage cost, support compliance and unlock innovation in a disciplined but expansive manner.

So, if you’re a Technology leader in your business, ask yourself this, do you know what AI agents are running across your estate?  Do you know who owns them, what data they access, and whether they truly deliver business value? If the answer isn’t yes, then governance shouldn’t be an afterthought, it should be your next strategic priority.

Well, that is all for today. Thanks for tuning in to the Inspiring Tech Leaders podcast. If you enjoyed this episode, don’t forget to subscribe, leave a review, and share it with your network.  You can find more insights, show notes, and resources at www.inspiringtechleaders.com

Head over to the social media channels, you can find Inspiring Tech Leaders on X, Instagram, INSPO and TikTok.  And let me know your thoughts on the governance of AI Agents.

Thanks for listening, and until next time, stay curious, stay connected, and keep pushing the boundaries of what is possible in tech.