What follows isn't one person's opinion. It's a cross-section of experts and practitioners independently making some version of the same argument.
“A Systematic Approach to Experimenting with Gen AI”
Harvard Business Review — January–February 2026
https://hbr.org/2026/01/a-systematic-approach-to-experimenting-with-gen-ai
Documents Siemens, P&G, and Google using disciplined, staged experimentation — targeted pilots, evidence-based scaling decisions — to navigate the gen AI “productivity J-curve” rather than rushing to full deployment.
“When AI Joins the Product Team, Will Leadership Still Drive Innovation?”
California Management Review, Shivam Srivastava — February 27, 2026
https://cmr.berkeley.edu/2026/02/when-ai-joins-the-product-team-will-leadership-still-drives-innovation/
Argues that as AI speeds up ideation and testing, leadership judgment — exercised at defined go/no-go gates — becomes more valuable, not less.
“Two Frameworks for Balancing AI Innovation and Risk”
Harvard Business Review, Faisal Hoque — March 6, 2025
https://hbr.org/2025/03/two-frameworks-for-balancing-ai-innovation-and-risk
Cites data showing only 26% of companies produce a working AI product and just 4% see meaningful ROI — and argues structured, framework-based decision-making is what closes that gap.
Global Consulting Firms
“Accelerating the AI Transformation”Bain & Company — 2026
https://www.bain.com/insights/accelerating-the-ai-transformation/Describes the move from isolated AI wins to institutional transformation as a matter of ruthlessly selecting which experiments graduate to scaled programs — rejecting a single universal decision point in favor of risk-differentiated approval paths and gates.
Independent and Boutique Advisory Practices
Peer-Reviewed Academic Research
“Driving Business Value: A Strategic Framework for AI Adoption and Deployment Success”Dr. Robert G. Cooper (2026). IEEE Engineering Management Review — early access, March 2026
http://www.bobcooper.ca/articles/artificial-intelligence-in-npdOutlines RAPID, a gated system for managing AI projects that helps managers “do the right projects” and “do them right.” Four stages, each separated by decision gates that confirm feasibility, user acceptance, and business value before the next round of investment — including how to build a business case when the economic value of AI benefits is still uncertain. (Available from the author's website, Article #7.)
“Alternative Approaches to Innovation Project Portfolio Governance”Siiri Tuominen & Miia Martinsuo, Project Management Journal — April 13, 2024
https://journals.sagepub.com/doi/10.1177/87569728241242429A study of four companies finding that formal governance structures — stage-gate and gatekeeping practices among them — persist alongside agile execution, with one portfolio leader describing a deliberate shift toward “more structured governance mechanisms and processes.”
The Discipline's Own Originators, Responding to the Same Signal
“What Gatekeeping REALLY Is: A Management Perspective”Michelle Jones, Stage-Gate International
https://www.stage-gate.com/blog/what-gatekeeping-really-is-a-management-perspective/Defines the gatekeeper's role across the four functions leaders keep describing as missing — in-stage learning velocity, timely business-value decisions and prioritization, resource reallocation, and unified cross-functional alignment — and explains Stage-Gate's staying power alongside Agile: the two jobs are complementary, not competing.
What's Consistent Across All of It
Read side by side, these sources — written independently, for different audiences — converge on four things that gated governance restores to a portfolio.
- In-stage learning with discipline. Teams spend their time on the work that actually de-risks a project, not on paperwork. Evidence builds gate by gate.
- Business value, checked on a schedule. Go/kill decisions happen at defined points in a project's life. That's what allows weak projects to be stopped, and their resources redirected, before sunk cost takes over.
- Resource allocation tied to evidence. Funding releases at each gate, sized to demonstrated results rather than to whichever project has simply been running longest.
- Leadership back in the room. This is the thread that shows up most: nearly every source frames the gap as a leadership-engagement problem. Ungoverned agile execution pushed senior leaders out of the decision. Gated checkpoints bring them back — not as one executive with a veto, but as a cross-functional group making the call together.
What This Means for Leaders
To be clear, none of this argues against agility. Nearly every source above pairs gates with agile execution rather than replacing it. Agile didn't fail. It was never a portfolio governance model on its own, never designed to make Go/No Go decisions on projects. Organizations that treated Agile as a portfolio system lost track of their projects, what they were building, what it was worth, and whether the business case still held up.
AI has simply raised the stakes. It often entered the portfolio through a vendor relationship rather than a business case, and it carries risks — model drift, compliance exposure, unproven ROI — that are hard to manage without a defined gate to catch them.
Continuing the Conversation
This analysis raises the obvious next question: if gated governance is coming back as the trusted model, what does doing it well actually look like in practice, especially with AI built into the process? That's the focus of PDMA's Ignite Innovation Summit, October 8–9, 2026, in Atlanta, marking PDMA's 50th anniversary.
Dr. Robert Cooper opens the Summit with “Winning at New Products in the AI Era,” on how AI is changing each stage of the new product process without replacing the structure itself. Michelle Jones and Penny Wise (3M) lead a live, real-time benchmarking workshop, “Key Characteristics of the Next Generation of New Product Governance Systems,” where participants assess their own governance model against peers in the room. A third session, a private Leaders' Roundtable of business and product leaders, will be announced in September.
Registration: https://www.pdma.org/events/register.aspx?id=2053691
This piece reflects an independent scan of publicly available research and commentary current as of August 2026. Full citations and links for every source appear inline within each section above.