Lorraine Bardeen, who leads AI strategy for Microsoft’s commercial business, said at a recent Stanford Institute for Economic Policy Research conference that global enterprises are entering a full re-architecture cycle.
“What they are embarking on right now is rebuilding their entire company,” she said. However, the shift is not about reducing headcount, but expanding customer reach and increasing throughput by pairing human workers with agents. “Are they embarking on rebuilding their company to get rid of people? No, they are not.”
Bardeen said enterprises are settling into three modes: humans using assistants for routine tasks; teams embedding agents into daily workflows; and entire functions moving to autonomous operation.
One example already operates inside Microsoft. A sales agent responds to small-business leads without human intervention. “They will never hear from a human ever,” she said. That agent writes outreach emails, logs activity and books revenue.
Hard Part Is Rebuilding the Company Structure
Also participating in a panel discussion at the Stanford conference, Acuity CEO Neil Ashe said the biggest hurdle is not AI performance but organizational capacity. “The technology is not the hard part. It is the changing of the company part that is the hard part,” he said. Acuity manufactures lighting and building systems used in commercial spaces. Ashe said one of the company’s major engineering modernization programs once required up to ten years. With agentic tools, the company solved it in 30 days.
Advertisement: Scroll to Continue
The breakthrough created a new challenge. The rest of the workflow needed to be redesigned to unlock the benefit. Ashe said this dynamic will widen performance gaps across industries. “There will be haves and have nots,” he said. Firms that can effectively redesign operations at scale, while managing risks and change resistance, are better positioned to capture productivity gains and sustain AI-driven growth.
Ashe said AI also reshapes early-career work. Acuity continues to hire junior talent because judgment becomes more valuable when information becomes abundant. The pressure instead falls on managers who must rethink what early roles should be in an AI-augmented environment.
Data Bottlenecks Slow the Shift
Snowflake vice president of AI engineering and research Dwarak Rajagopal said fragmented data remains one of the biggest barriers to scaling AI. “The data is like everywhere within an enterprise,” he said. Pilots often succeed within single domains, but agents stall when information spans systems with inconsistent governance rules.
Snowflake built an internal enterprise agent that answers natural language questions about company data. Rajagopal said employees ask “almost like 12,500 questions per week” and save “about 15 minutes per question.” But productivity becomes harder to measure when worker behavior changes. Employees now ask follow-up questions they previously avoided because the manual effort was too high.
AI models will keep improving, but the economic payoff will depend on how many companies can reorganize fast enough to use them. Demonstrating clear ROI is crucial, reassuring the audience that strategic restructuring and value proof are essential for sustained AI investment and success.
CFOs are also tightening expectations as returns remain uneven. A new PYMNTS Intelligence survey showed that only 26.7% of CFOs plan to increase generative AI budgets over the next 12 months, down sharply from 53.3% a year earlier.
The pullback signals a shift from experimentation to disciplined, results-driven spending. The divide is clear. Half of the firms reporting very positive returns plan to expand their budgets, while only 16.7% of companies seeing negligible results intend to do the same. The numbers show that future AI investment is now tied directly to proof of financial or operational value.
For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.
