In PM World Journal: When AI Handles the Work

PM World Journal published my article When AI Handles the Work: How Program Managers Stay Valuable as an advisory article in its September 2026 issue – Vol. XV, Issue IX. PM World Journal is a monthly global publication for program and project management (ISSN 2330-4480), published by PM World Inc. with advisors, authors and correspondents around the world.

The argument is short. AI has made coordination cheap. Summarising a meeting, drafting an update, comparing two plans, organising a risk list – work that used to take an afternoon now takes minutes. That is usually framed as a threat to program managers. I do not read it that way. It is a forcing function. It separates the work that kept us busy from the work that made us valuable.

Nobody on my programs reports to me. I cannot offer a raise, grant a promotion, or lean on a reporting line to make ten teams move together. What actually holds a program together is that the program manager quietly carries a small piece of many other people's responsibilities and connects views that are each incomplete. The executive has the authority but not the time. The architect has the technical direction but not the roadmap. AI is very good at moving information between those views. It is not accountable for judgment – for noticing that the polite agreement in the minutes was never a real commitment.

The article works through four things that survive when the administrative layer gets automated:

• Create the conditions for a decision, not a report that one is late. Two weeks before a launch, two principal engineers deadlocked on architecture. The familiar move is a follow-up email and a yellow status. Instead I worked backwards: what exactly has to be decided, who is accountable for the call, which two options are viable, which criteria matter for this launch, and when does the delay start doing damage. I did not own the technical answer. I owned the decision-making system.
• Know enough about the domain to ask the load-bearing question. You do not need to be a structural engineer to renovate a house, but you do need to know enough to ask whether the wall is load-bearing. AI will happily explain an unfamiliar architecture, which removes the last excuse for staying shallow. What it cannot tell you is which question is load-bearing for this particular program.
• Bring product and business judgment, not just delivery. My team once spent a quarter building a monitoring dashboard exactly as requested. It shipped on schedule. A month later it had been viewed nine times. The users did not need another dashboard to visit – they needed the one alert that required action to appear inside the tool they already used. A schedule can tell you when something ships. It cannot tell you whether anyone will use it.
• Build the trust that makes people tell you the real objection. When two teams disagree, the visible argument is rarely the whole problem. Someone is protecting an operational burden the other side cannot see, or remembers a commitment that was not honoured. People only say that out loud to someone they trust to represent them accurately.

The line I keep coming back to: AI can sit at the table, but it does not get the vote. When the status report writes itself, the value moves to knowing what the report means. When the risk list is generated automatically, the value moves to recognising which risk deserves attention. The most valuable program manager in the AI era will be the person the organisation trusts to turn incomplete information into a responsible decision – not the person who produces the most information.

The article develops ideas I first presented as The AI-Era PM: Becoming a Strategic Technology Leader to the PMI San Francisco Bay Area Chapter in July 2026.

Read the article in PM World Journal →
Download the PDF →

Gupta, A. (2026). When AI Handles the Work: How Program Managers Stay Valuable. PM World Journal, Vol. XV, Issue IX, September. ISSN 2330-4480.