Forbes: 31% treat AI as employees

AI agents are starting to show up in a weird new place — not just in software stacks, but in the language companies use to describe their workforce. That sounds cosmetic, but it isn’t. Once a business starts calling an agent a teammate, a coworker, or even an employee, it changes how people think about ownership, review, and blame. The May 9 Forbes piece is really about that shift — the move from “tool” to “worker” — and why it creates management problems faster than it solves them. ### Where did the 31% number come from? The number appears to come from the Wharton School and GBK Collective’s enterprise AI adoption research, which has been tracking large U.S. companies over multiple years. That same body of research shows AI use among enterprise leaders has become routine — more than 80% report weekly use, and many say early deployments are already producing returns. So the 31% figure lands in a broader pattern: AI is no longer a pilot project, and leaders are now deciding how to structure it inside the company. (forbes.com) ### Why call an AI an employee at all? Basically, because “tool” undersells what agents are starting to do. A spreadsheet doesn’t draft outreach, route approvals, summarize policy, update records, and hand work to another system. Agents can. That makes executives reach for workforce language — names, roles, managers, even org-chart placement — because the software is acting less like a feature and more like a semi-autonomous operator inside a workflow. (hbr.org) ### So is this just branding? Not really. The catch is that language becomes governance. If an agent is framed as a worker, people start making worker-shaped assumptions — that the agent “owns” a task, that escalation can be pushed to it, or that a human reviewer is somehow less central. HBR’s new research on this is blunt: anthropomorphizing AI can shift accountability away from people, increase escalation, reduce review quality, and weaken trust and professional identity. (hbr.org) In other words, the naming convention can quietly change behavior. ### Why does finance and control care so much? Because finance teams live on sign-off, traceability, and evidence. If an AI agent drafts a forecast note, classifies spend, or prepares a board-ready summary, someone still has to own the inputs, the judgment, and the final approval. A human controller can’t defend an audit trail by saying the “AI employee” handled it. That is the real operational issue under the Forbes story — not whether the metaphor is cute, but whether companies are building review paths that survive scrutiny. (hbr.org) ### How far along are companies, really? Far enough that this is not fringe, but not so far that the operating model is settled. PwC’s 2025 survey found 79% of companies were already adopting AI agents in some form, and 66% of adopters said they were getting measurable productivity value. But PwC also makes the bigger point — many firms are still using agents as isolated helpers, not as deeply connected systems that transform work across functions. (forbes.com) ### What’s the human side of this? HR is already treating AI as a workforce issue, not just an IT purchase. SHRM’s 2026 reporting shows 92% of CHROs expect AI to be more integrated into the workforce this year, and 87% expect greater AI adoption in HR processes. That means org design, policy, compliance, and employee trust are now part of the same conversation as model capability. (pwc.com) ### Does treating AI as an employee help adoption? Turns out, maybe not much. The HBR experiment suggests the employee framing does not meaningfully increase people’s intent to adopt AI in workflows — which is important, because many leaders seem to assume the human-like framing will make the systems easier to accept. If that benefit is weak, but the accountability downside is real, the metaphor starts to look expensive. (shrm.org) ### Bottom line? The story is not that AI literally became labor. It’s that companies are crossing a line from using AI to organizing around it. And once that happens, the hard problem stops being capability and becomes control — who checked the work, who owns the decision, and who answers when the agent gets it wrong. (hbr.org)