AI Agents Are Challenging the Way Enterprises Are Built

In Brief

Box CEO Aaron Levie, Boston Consulting Group (BCG), and Microsoft CEO Satya Nadella delivered the same message in a single week: AI agents are exposing how unprepared most companies are to capture AI's value. 

Levie mapped the operational roadblocks, BCG's survey of nearly 12,000 employees found that strategic clarity separates the winners, and Nadella warned that companies are paying for intelligence twice.

What Happened

AI agents, AI that can take multiple steps on its own to complete a task with minimal human oversight, are forcing enterprises to confront how they are organized. Box CEO Aaron Levie published a widely shared rundown on X of what he is hearing from enterprise IT leaders putting agents into production. His biggest theme: Most companies are built in silos, but agents work best when tied to a process, and the most valuable processes cut across those silos. Underneath that sits data fragmentation, since agents struggle to give accurate answers when a company's data is scattered and non-standardized. In a world where everyone can tap roughly the most advanced AI models on the market, a company's proprietary context, its own data captured and formatted for agents to access, becomes the real competitive advantage, Levie argues.

He added that tokens, the chunks of text AI usage gets billed on, are the wrong metric, and that companies should manage to business outcomes such as revenue or shipped product instead.

BCG's fourth annual AI at Work survey of nearly 12,000 employees across a dozen global markets backs him up. Seventy-four percent of frontline workers are now regular AI users, up 23 points from last year, and 61% believe agents could do at least half their job within three years. The survey's core finding: strategic clarity, not access to tools, is what separates the organizations getting real value.

Then Nadella joined in, publishing a post on X called The Reverse Information Paradox that drew 7.5 million views in less than a day. Nobel Prize-winning economist Kenneth Arrow famously observed that a buyer cannot know the value of information until they have it, at which point they have effectively acquired it for free. Nadella flipped it: In the AI age, the buyer risks handing over their own knowledge simply to use the product they purchased. "You essentially pay for intelligence twice, once with money, and again with something even more valuable, the proprietary knowledge you must reveal to make that intelligence useful," Nadella wrote.

SmarterX founder and CEO Paul Roetzer discussed these ideas on Episode 225 of The Artificial Intelligence Show.

The Key Numbers 

74% - Frontline workers who are now regular AI users, up 23 points in a year [BCG survey]

61% - Workers who believe agents could do at least half their job within three years [BCG survey]

8 hours - Weekly time savings reported by 42% of AI users [BCG survey]

30% - Organizations with AI agents integrated into workflows, up from 13% last year [BCG survey]

7.5+ million - Views on Nadella's Reverse Information Paradox post in under a day

Why a Clear Strategy Beats Tool Access

The time savings are real, and then they evaporate. BCG found 42% of AI users save eight hours a week, a full workday, with even higher rates in marketing (60%), IT (53%), and HR (50%). But 66% receive little or no guidance on what to do with that time, and more than half are not reinvesting it in more strategic work. "Okay, great, we gave them tools, and maybe we even trained them how to use the tools, but we didn't train them what to do with the time they're saving, and so they're not redirecting that into strategic efforts and new campaigns and new ideas and things like that, which is a huge opportunity for companies," says Roetzer.

Leadership communication is the weak link. Only a third of frontline employees say leadership's communications about AI are clear, and only 28% see a strong connection between what leaders say and what the organization actually does. In his eight pillars of business AI transformation, Roetzer says, "The most important thing was clarity from leaders, that leaders understood the moment, and that they were clear in their communication." BCG's five CEO imperatives, all of which Roetzer endorsed, push the same direction: own strategic clarity personally, measure value instead of adoption, redesign work end-to-end rather than buying more tools, put people at the heart of the redesign, and govern AI as a moving target.

The redesign has started, and the talent must come from within. Organizations using AI to reshape workflows end-to-end nearly doubled to 42% from 22%, and 72% of respondents say expectations for the skills they need have shifted. That collides with Levie's warning that agent talent is so scarce most companies will have to train for it internally. "AI literacy is fundamental," says Roetzer. "The companies have to own reskilling and upskilling their own people. It's going to be the fastest path."

The biggest wins change the work itself. Levie's point that the best use cases fundamentally change the work being done, instead of just replacing an existing process, maps directly to the framing Roetzer featured in his MAICON 2025 keynote.

"Optimization is doing things better, faster, cheaper. That's 10% thinking. We should absolutely be doing it. But innovation is reimagining what's possible and creating entirely new forms of value. That's 10x thinking, and that's what we want to be doing."

— Paul Roetzer, founder and CEO of SmarterX, Episode 225 of The Artificial Intelligence Show

SmarterX Take

The through line from Levie to BCG to Nadella is ownership of the learning loop. Roetzer flagged how remarkable the messenger is: Microsoft is OpenAI's biggest investor, and Nadella is in essence taking the anti-OpenAI position, urging companies to build private evals, the internal tests that define what good AI output looks like for your organization, and to keep their orchestration layer, the routing system that decides which AI model handles which task, decoupled from any single provider. Roetzer added a buyer-beware warning about intermediaries that sit between you and the models, naming Perplexity as an example: "They are 100% taking the data that you're putting in, the prompts you're using, the things you're building, as training for their own businesses. That is the value to them."

His pragmatic bottom line: for most knowledge workers, this is the trade we have always made with business software, from Gmail to Microsoft. The calculus flips for genuinely sensitive work. Lawyers working on patents, in Roetzer's view, should be on an open source model, one a company can download and run entirely on its own systems, so proprietary knowledge never leaves the building.

What to Watch

Agent governance is the next pressure point. BCG found 84% of employees have heard of AI agents, 30% of organizations have integrated them into workflows, and another 50% have run pilots. Yet half of respondents say their companies lack clear governance for managing teams of people and AI. Awareness and adoption are outpacing the systems built to supervise them.

Multi-model enterprises are coming. Roetzer highlighted Levie's view that almost no enterprise will bet on a single model provider. Companies want routing layers that can shift workloads between models for cost and performance, plus an internal open source backup in case a provider goes down or gets too expensive. Nadella's "choice" prescription points the same way, so expect orchestration to become a standard line item in enterprise AI strategy.

Only 13% Have the Governance Foundations to Scale AI

Just 13% of organizations have all four governance foundations in place, meaning an AI council, an AI roadmap, generative AI policies, and an AI ethics policy, according to the 2026 State of AI for Business Report, and 32% have none of them. Roetzer noted on the show that BCG's findings reinforce that research: agent adoption is more than doubling while the structures to govern it barely exist.

The full report, based on more than 2,100 responses from professionals across roles, functions, and industries, maps where organizations actually stand on adoption, governance, training, and strategy. Read the full report →