As AI becomes more deeply embedded in enterprise operations, IT leaders are facing a bigger question than which technology to adopt: How do you decide where to invest, what to govern and who ultimately owns the outcome?

In this episode of TechEDGE, Abby White and Theresa Houck connect three challenges increasingly shaping enterprise IT strategy. They explore why modernization doesn't mean eliminating every piece of technical debt, what happens when AI agents begin influencing decisions across the business, and why some organizations are moving AI into production while others remain stuck in pilot mode.

The conversation ultimately comes down to making more deliberate technology decisions — and building the structure necessary to turn those decisions into measurable business value.

What You’ll Learn

• How IT leaders can determine which technical debt actually deserves investment
• Why the rise of AI agents creates new questions around ownership and accountability
• What leaders should consider as AI systems gain access to more business processes and decisions
• Why the number of AI deployments may be the wrong way to measure AI maturity
• What separates organizations generating AI value from those stuck experimenting
• Why governance and decision-making architecture are becoming critical parts of enterprise AI strategy
  
  Why It Matters
  IT leaders aren't just managing technology anymore. They're increasingly managing where and how decisions get made.
  
  Technical debt can constrain what's possible. AI agents can introduce new layers of autonomy and risk. And disconnected AI pilots can consume resources without creating lasting business value.
  
  For CIOs and technology leaders, the challenge is building enough visibility, ownership and structure to know what deserves investment — and creating a repeatable path from technology experimentation to business outcomes.

Prefer to read? Here's an excerpt of the podcast transcript:

Abby White: I really like how this month's three most-read stories build on each other. Theresa, the first article of yours is called "How IT Leaders Can Manage Technical Debt Without Modernizing Everything." Modernization doesn't just mean ripping everything out that's old, because some older systems may be doing their jobs perfectly well. How should leaders determine what actually needs attention?

Theresa Houck: It's really expensive to replace all your legacy hardware and software. By technical debt, we mean the future costs and extra work that pile up when software and hardware teams choose fast, short-term shortcuts over clean, robust design. Taking a shortcut helps you move faster right now, but later you're going to have to put in additional effort.

Rather than just replacing old hardware and software first, it's important to focus on what creates the greatest business risk, cost or strategic constraint. Legacy products might not be creating any of these problems. AI can speed up modernization when it's needed, but it can't decide what to modernize. Business value and risk still need to drive those calls, and humans make those decisions.

Abby White: You cited some interesting McKinsey research that says companies can spend an additional 10 to 20% on top of project costs dealing with technical debt. That's something that will get the attention of your C-suite. How should CIOs make the business case for spending money on what needs to be replaced?

Theresa Houck: It means translating what they want to do in IT into financial metrics like revenue growth, risk reduction and operational efficiency. Use financial terms, talking about risk, cash flow and profit, or show what inaction would cost the company, like how old systems slow down feature releases and increase infrastructure costs. Tie reductions in technical debt to business goals, like showing how modern systems support upcoming company goals, like implementing AI or speeding up new product launches.

Abby White: It's really pointing out what it's going to cost the company if you don't do it, right?

Theresa Houck: Right. In the article, there's an executive decision matrix that shows you when to modernize and when to wait, based on whether something is high-risk or low-risk.

Abby White: I'm glad you brought up the point that this is not a decision AI can make. AI is starting to help with some of that modernization work, analyzing legacy code or mapping dependencies and speeding up refactoring. But the technology itself still can't decide which systems deserve the actual investment, right?

Theresa Houck: That's right. It needs human judgment.

Abby White: That question of judgment gets more complicated once AI moves beyond helping IT teams and starts influencing the decisions itself. That brings us to your next article, Theresa, "The New Automation Stack. Who Governs AI Decisions?" One phrase from this story that really jumped out at me was agent sprawl. IT leaders have already lived through application sprawl, software-as-a-service sprawl, and cloud sprawl. Now they have dozens, maybe hundreds, of AI agents operating across the company. Why does that become a different kind of problem?

Theresa Houck: Because it creates an unmanaged, invisible network of autonomous tools that expose companies to cybersecurity risks, compliance and privacy failures, and duplicated efforts and wasted costs. Sprawl happens when different teams build and launch agents independently without telling IT.

Abby White: I loved how you call AI agents "digital employees," which makes the governance issue really easy to understand. Companies wouldn't hire hundreds of employees without knowing who they report to or what they're allowed to do. Yet organizations could end up deploying agents across different departments without having that same level of visibility.

Theresa Houck: In the article, we cite a report from Deloitte that says the AI agents may eventually become what the report calls a silicon-based workforce that complements human workers. That's a powerful shift because it turns a technology discussion into a management discussion.

Most organizations would never want hundreds of employees across departments without the reporting structures. They need accountability, performance metrics and governance policies. But many are deploying AI agents without the equivalent controls. Procurement might have supplier evaluation agents, while logistics has shipment monitoring agents. Every deployment might make sense independently, but collectively, they create a huge governance problem. Businesses need visibility into what agents exist and which systems they can access. Every AI agent needs a clearly defined business owner. Governance needs to scale along with agent adoption, because an abandoned agent could continue to influence business activity.

Abby White: Having a thousand agents doesn't necessarily make a company more advanced than one with a hundred, right? It depends what they're doing. Before another business unit spins up another agent, someone should be able to answer: Who owns it? What can it access? What decisions can it influence? And how will we know whether it's actually creating value?

Theresa Houck: That's why it's important to know what you have, in order to have effective governance for them. You need to measure them and focus on improving business outcomes such as decision cycle time, risk reduction, operating margins, customer resolution and time-to-market.

Abby White: That leads us directly to your final story, Theresa. Your third article is called "The Hidden AI Competitive Advantage: Decision-Making Architecture." This article asks why some organizations are generating measurable results from AI while so many others remain stuck in that experimentation phase. A lot of companies have access to comparable models now, but where are the organizations getting real value and pulling ahead?

Theresa Houck: We found that the differentiator is execution. Specifically, creating what is called the effective decision-making architecture. Decision-making architectures combine AI, workflow designs, governance and human oversight. As AI agents scale across the enterprise, governance becomes a critical control layer for managing automated decisions, risk and accountability. This story talks about using decision-making architectures to turn AI pilot programs into implemented programs that produce measurable business results.

Abby White: Companies generating the greatest returns treat AI like an enterprise pipeline that moves ideas from experimentation to govern repeatable operations. You cited a McKinsey statistic here that nearly two-thirds of organizations remain in the experimentation or pilot phase, despite all the money and attention going into AI. What keeps companies stuck in that AI pilot purgatory?

Theresa Houck: McKinsey's research argues that trust and governance have become foundational to get value from AI. In this article, we provide a list of five signs your company is stuck in AI pilot purgatory;:

• You have more pilots in production than deployments.
• There's no executive owner for business outcomes.
• Success is measured by usage rather than value.
• Governance reviews take longer than implementation.
• And workforce training lags technology deployment.

Governance is becoming the control layer for an enterprise decision-making architecture. As AI agent adoption grows, companies are basically creating a new digital workforce. This workforce requires permissions, oversight, auditing, escalation paths and even performance management.

Abby White: Maybe the question CIOs should be asking about AI projects is what happens after the pilot? Along with that, who owns the outcome? What workflow changes? How will we measure value, and how does this become something the organization can repeat? 

Looking across these three stories, there's a really useful thread here for IT leaders. With technical debt, you need enough visibility to know where investment matters. As AI agents spread, you need to know what they're doing and who owns the decisions they influence. And if AI is going to create meaningful business value, the organization needs a repeatable way to move from experimentation into operations.

Theresa Houck: That's exactly right. All three articles also carried themes emphasizing the human decision-making process that is not taken out of using AI, and it's still needed. And everything is about achieving business outcome goals.

*Transcript lightly edited for clarity and brevity