Feature
White Paper
The Future of Leadership Development: AI-Driven Manager Enablement
For years, enterprise app strategy was a cross-departmental decision: which tools do our people need and which can we cut. IT and procurement leaders ran the numbers, line of business owners surveyed employees and the CIO typically made the call.
AI is changing that. As AI agents and copilots become embedded into workplace platforms, they bring their own dependencies: integrations, data sources and app connections they need to function.
Increasingly, the apps that survive aren't the ones employees prefer, but the ones AI uses.
Not every operator agrees this is how it should work. But several builders and executives say it's already happening, whether IT leaders have signed off on it or not.
App Adoption vs. AI Integration
User adoption is being displaced as the metric that determines whether software stays or goes, according to Andrey Kudievskiy, chief executive of Distillery, whose company builds production AI systems for enterprise clients.
"Buying decisions are being led with 'what does our AI need' before 'what do our people need,’” Kudievskiy said.
Applications that cannot plug into AI workflows get deprioritized regardless of how popular they are with staff, Kudievskiy added. Microsoft's positioning of Copilot as the connective layer across 365 apps, and Google's similar push with Gemini in Workspace, indicate that platform vendors are using AI integration to deepen lock-in, he said.
Developers also see this change. The first question on a build used to be which tool the team already liked, said Tom Crawshaw, founder of the AI Architects. Now, he said, "it's the first thing I check: does this thing have a clean API my agent can actually call?" A popular tool that is locked down loses out to a duller one an agent can drive.
The decision is sliding away from procurement teams and toward whoever ends up wiring the automations, with everyone else inheriting that person's choices by default.
Not everyone reads the change the same way. "If the AI can use an app and your people won't, you haven't fulfilled what they were engaging you for, and the workflow is still broken," said Eric Vaughan, chief executive of IgniteTech. "Adoption is the proof that it actually works for a human being."
Betting an entire stack on one vendor's AI being the layer that holds everything together is a bet on a race nobody can currently call, Vaughan added. The perceived leader among large language models has changed multiple times recently, which is why IT leaders should be wary of lock-in.
The Bigger Shift: From Apps to Orchestration
The adoption-integration debate may itself be the wrong question, according to Edwin Miranda, founder of evolution consultancy Konsultora. "We're entering a period where applications themselves become less important than the intelligence layer sitting between them," he said. "The more important question is whether applications remain the center of enterprise architecture at all."
There's also an assumption that AI will simplify the app estate that's worth questioning, Miranda said. In many organizations, AI is adding a new layer of complexity on top of existing complexity, not removing it. The companies that use AI to orchestrate and simplify what they already have, rather than stacking on more tools, will win, he said.
The real fight is not which metric decides an app's survival, but whether apps will matter at all, Miranda said. Over the next several years, the center of gravity is likely to shift toward context, orchestration and intelligence, with individual apps becoming less load-bearing, regardless of which side of the adoption debate wins, he predicted.
Moreover, while Kudievskiy and Crawshaw describe a stack reorganizing around what AI can reach, with human preference becoming a secondary input, Vaughan calls that a design failure waiting to happen.
Workflows should route to AI first, and to a person only when the AI cannot resolve something, but the human override has to remain real and accessible, Vaughan said. Build the stack the other way around, and the person ends up serving the machine, which inverts what made the technology useful in the first place, he said.
How Lock-In Happens Without Anyone Approving It
Vendors should be asked to demonstrate measurable business outcomes from their AI integrations, not just connectivity, to separate capability from retention strategy, Kudievskiy said. Can the model be swapped without rebuilding the stack, are the APIs open, where is the data stored and how does a human override the AI when it gets something wrong?
If a vendor cannot answer those questions cleanly, that is itself the answer, Vaughan agreed.
Crawshaw is less concerned about vendor behavior than about internal accountability. Lock-in being created now was not signed off by anyone, because once an agent depends on a tool's API, removing that tool breaks the workflow even though no contract was negotiated and no IT or procurement process ever approved the dependency, he said.
Engaged, But Not Convinced: What's Eroding Beneath Your Strongest Engagement Numbers
Watch Now
Retention Is a Communication Problem: New Data from 6,000+ Employees
Watch Now
Who Will Own the Stack in 3 Years?
Who will control enterprise app strategy in three years? That’s up to leadership, not with vendors or models, Vaughan said. If the CEO and CIO own the architecture and refuse to delegate it, they keep control. If they do not, the vendor or the model controls it by default.
But if focus changes from applications to orchestration and context, as Miranda predicts, the app-by-app decisions IT leaders have made for years may no longer be the right step.
Either way, the stack is being rationalized. The question is who is making it.
Editor's Note: Our workplaces, our people, our tech stacks are all in a state of flux. More on the topic:
- AI Agent Sprawl Is a New Twist on an Old Digital Workplace Problem — As AI agents proliferate without governance, we risk damaging digital employee experience through AI sprawl.
- Cisco, ServiceNow and Microsoft Want to Manage Your AI Agents. They Only Do Part of the Job — Cisco, ServiceNow and Microsoft are racing to manage enterprise AI agents. All three are building useful infrastructure — but none can answer who's accountable.
- Redesigning Workflows for an Agentic World: Travel Spend — Automating a broken workflow just scales the mess. See how redesigning a common one — travel spend — sets the stage for AI that actually works.