At a large global insurance company, the chief information officer faced a frustrating (and common) problem with his device: PowerPoint kept crashing. A general-purpose AI assistant produced a long list of possible causes, from Microsoft Office and memory issues to a device problem or recent update.
The IT team reviewed the device and identified the conflict—a newly installed AI note-taking application interfering with PowerPoint—and resolved it quickly.
That incident captures a challenge many companies will confront over the next few years. A large language model can tell you dozens of reasons PowerPoint might crash. What matters operationally is knowing why this PowerPoint is crashing on this device right now.
When nearly every company can access the same powerful AI models, simply having AI will not create durable competitive advantage.
AI won’t be the advantage
We’re still relatively early in enterprise AI, yet access to sophisticated models is widespread. Companies can use multiple models, switch between them, and build similar AI capabilities on top of the same underlying technology.
The strategic question is “Which company can give AI a reliable view of its business and the authority to make better decisions?”
IT makes that distinction unusually visible. One organization I’m familiar with was struggling with virtual desktop login times of about four-and-a-half minutes. Better operational data revealed unnecessary applications and other sources of overhead, reduced its application footprint by roughly 30%, and ultimately brought login times down to about 45 seconds.
The advantage was understanding what was actually happening, not acting on assumptions.
Context changes the equation
I believe real-time operational context—data connected to the business’s current condition—will become one of the most valuable assets in the AI economy. Historical data can tell AI what usually happens. General knowledge can tell it what might be happening. Real-time data can tell it what is happening now. That distinction matters more as AI moves from answering questions to participating in decisions.
I’ve seen IT teams spend days investigating a spike in incidents because people had to manually export reports, compare systems, review changes, and correlate what was happening across thousands of devices. With AI revealing connections in that underlying operational data, that same analysis can be reduced to minutes.
AI turns fragmented signals into a usable operational picture. The model is making connections across information at a speed and scale that would be difficult for people to achieve manually.
Real value comes from trusted action
The first wave of enterprise AI largely focused on helping people find information, generate content, analyze data, and make recommendations. Those are useful applications, but I think the bigger economic opportunity comes when AI can act safely.
That’s also where things get much harder.
Consider IT operations. A system could detect deteriorating application performance, determine that a recent configuration change is the likely cause, and recommend a fix. Acting on that recommendation across 10,000 devices is an entirely different proposition.
The system needs to know whether the action is authorized, which devices should receive it, what dependencies exist, and how to determine whether the intervention worked. The system also needs clear boundaries for when it acts autonomously and when a human needs to decide.
In my experience, this is where conversations about AI sometimes outrun reality. Trustworthy action requires governance and confidence.
For years, companies have accumulated enormous quantities of data without necessarily extracting proportional value. AI changes the economics of that data, as the value lies in relevance, freshness, and a connection to an authorized action.
I’m familiar with one support organization that once spent roughly an hour diagnosing certain remote employee problems. Technicians depended heavily on employees to describe their devices, home networks, and connectivity. Once IT technicians could diagnose data directly from the employee’s device and its environment, the same process took only 20 minutes.
Now imagine AI sitting on top of that same live context, continuously looking for patterns across thousands of employees and identifying issues before anyone calls the help desk. The same principle applies to supply chains, pricing, fraud, and customer experience.
The road ahead
AI’s progression is straightforward: Real-time data enables better decisions, and better decisions enable safer automation. Automation creates more data about what actually works. That feedback loop becomes increasingly valuable over time.
Competition among AI models will continue. Some will be faster, cheaper, more specialized, or better suited to particular tasks. But for most companies, the primary source of lasting differentiation, and the harder thing to replicate, will be everything surrounding the model. That is, a proprietary, continuously updated operational picture—with the governance required to act safely—and feedback loops that show whether those actions produced the intended result.
Leaders should look for competitive advantage in the unique context AI brings to increasingly ubiquitous intelligence and how effectively it can turn that intelligence into action. Anyone can connect to a model. Far fewer organizations have an accurate understanding of what’s happening right now and enough confidence to safely act on it.
As intelligence becomes abundant, context and the ability to act on it responsibly will become scarce.
Jed Ayres is the CEO at ControlUp.