Luboslava Uram is COO and CTO at Solvd Group, a subsidiary of Allianz Group. Transforming the claim management experience.
For decades, technology leaders have managed technical debt: outdated systems, accumulated exceptions, architectural shortcuts and temporary solutions that somehow became permanent.
Today, I believe artificial intelligence is exposing another kind of debt—one that is potentially harder to resolve because it does not sit in our technology architecture: organizational debt. It accumulates through unclear responsibilities, duplicated processes, conflicting performance objectives, inconsistent data and decisions that require coordination across multiple organizational layers. Most enterprises have learned to operate despite these inefficiencies. People compensate for them every day.
But what happens when AI agents must navigate processes that were never clearly defined in the first place?
The Invisible Work That Keeps Organizations Running
Throughout my career leading technology platforms and global operations, I have observed how remarkably good people are at making imperfect organizations work. For example:
• A process officially belongs to one department, but everyone knows that unusual cases require calling an experienced colleague on another team.
• Three systems contain different versions of the same information, but someone knows which number to trust.
• Two functions share responsibility for an outcome, yet experienced managers understand who actually makes the final decision.
These arrangements become part of how organizations operate. They rarely appear in process documentation or technology architecture. Imagine introducing an AI agent into that environment. Which information should it trust? Which process should it follow? Who can authorize an exception? Who is accountable?
For a human employee, years of experience and informal relationships often resolve these uncertainties. For an AI agent, they must become explicit. Research supports the significance of this challenge. McKinsey’s July 2026 analysis, “The Operating Model Advantage“ (registration required), reported that only 21% of surveyed companies had fundamentally redesigned their operating models around AI. We are introducing increasingly sophisticated technology into organizations that have barely changed how they operate.
Why AI Makes Organizational Debt More Expensive
I see three areas where accumulated organizational debt becomes particularly visible.
1. Data Quality Is Often An Ownership Problem
Almost every AI transformation eventually encounters the same obstacle: Data is incomplete, inconsistent or unreliable. The immediate reaction is usually technological. Build better pipelines, clean databases or introduce another data platform. But why did the problem arise?
Often, different functions maintain competing definitions, nobody owns data quality across the entire process and business knowledge remains disconnected from operational records. PwC’s “2026 Digital Trends in Operations Survey“ found that 87% of surveyed operations leaders reported poor data quality affecting the value achieved from digital initiatives. Technology can correct individual data problems. But unless accountability for business meaning, quality and maintenance changes, those problems may return. The question is not only whether data is AI-ready. It is also who is responsible for keeping it AI-ready.
2. AI Forces Clarity About Decision Rights
In complex organizations, one team analyzes information, another recommends an action, a third owns the process and someone else approves the decision. People navigate these structures through meetings, escalations and negotiations. AI agents introduce a different challenge. If an agent can recommend or execute an action, who has authorized that authority? Which decisions remain human? Who can override an automated action? Who owns its consequences?
IBM’s June 2026 research among technology executives highlighted this accountability gap: Approximately “two-thirds of surveyed CIOs and CTOs report being held accountable for AI systems they do not fully control.” As AI moves from assisting people toward executing operational decisions, unclear accountability becomes increasingly difficult to manage.
3. Conflicting KPIs Become A Bigger Problem
Consider a customer journey involving operations, finance and customer service. Operations targets faster processing. Finance prioritizes lower costs. Customer service focuses on satisfaction. Each objective makes sense individually. But optimizing one can damage another. Experienced managers constantly negotiate these trade-offs.
What happens when AI begins optimizing decisions at machine speed? Without clear priorities, an agent might reduce processing time while increasing rework or minimize immediate costs while creating more expensive consequences later. AI can optimize what we tell it to measure. Leadership must decide what is actually worth optimizing.
Use AI Transformation To Pay Down Organizational Debt
I do not believe organizations should spend years redesigning everything before introducing AI. That would simply create another transformation program with uncertain outcomes. Instead, I would use AI implementation as an opportunity to resolve organizational debt where it matters most. Start with one significant end-to-end process and take four practical steps.
1. Make ownership explicit. Identify one accountable owner for the overall business outcome, even when execution spans multiple functions.
2. Establish trusted business context. Agree on data definitions, authoritative sources and responsibility for keeping information reliable.
3. Clarify decision authority. Define which decisions AI can support, which actions it may execute and where human judgement or approval remains necessary.
4. Align performance measures. Evaluate the complete customer or operational outcome, not simply the productivity of individual activities.
Only then should the organization determine which work to automate and how to introduce AI agents. This approach does not eliminate every complexity. But it prevents organizations from automating problems they should have resolved first.
The Real Cost Of Doing Nothing
For years, enterprises have absorbed organizational complexity through additional coordination, management layers, manual interventions and experienced employees who understand how things really work. That approach has a cost, even when it remains invisible in technology budgets.
AI presents an opportunity to challenge it. The most valuable AI transformation may not be the one that automates the greatest number of tasks. It may be the one that finally eliminates unnecessary handovers, resolves conflicting responsibilities and makes better decisions possible across the business.
This is why I increasingly see AI transformation as an organizational leadership challenge, not simply a technology initiative.
We should stop asking only how to make AI work within our organizations and start asking what needs to change in our organizations for AI to work effectively. The debt has been accumulating for years. AI is simply making the bill harder to ignore.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?

