Your Innovation Pipeline Might be Stuck with a Hairball; Please Don't Throw AI at it.

In 1996, Gordon MacKenzie, a thirty-year Hallmark veteran, gave corporate life one of its most durable metaphors. In Orbiting the Giant Hairball¹ he described how rules, processes and approval chains tangle together over time into a mass with its own gravity, pulling every new idea toward the average. Thirty years of process optimization later, the hairball has not dissolved. It has been upgraded. AI did not cut through the tangle; it changed what the tangle is made of.

AI made opinions and output free

Anyone can now produce a deck, an analysis, a recommendation or a plausible point of view in one click. Work that once took a team a week takes an afternoon; work that took an afternoon takes a prompt. Most often it is not wrong. It is simply abundant. Every meeting has more inputs, every thread has more opinions, and every reviewer has the means to answer analysis with counter-analysis by the next morning. Every new person on the thread is another hair.

The false feeling of efficiency

All that output feels like progress towards something. In practice, it slows things down: more to read, more to react to, and more ammunition for reopening decisions that were nearly made. Ezra Klein observed in The New York Times that the people he knows with small armies of AI agents feel more productive, yet their work has not improved, and in some cases it has clearly declined. His conclusion: "the feeling of efficiency should be mistrusted."²

Research is beginning to quantify the effect. HBR gave the flood a name: workslop. Research from BetterUp Labs and Stanford found that 41% of workers have encountered AI-generated output that looks polished but lacks substance, costing nearly two hours of rework per instance.³ A June 2026 HBR follow-up by Matthias Holweg and Thomas Davenport calls the compounding effect "knowledge decay": as polished but hollow output spreads across teams, the organization's collective knowledge base deteriorates.⁴ The noise does not just waste time. It degrades the very information decisions rest on.

Underneath sits an asymmetry worth naming. A say costs nothing and carries no risk. Ownership costs plenty and carries all of it. AI made the first free and left the second expensive.

The scarce thing flipped

Alex Imas of the University of Chicago points to where this leads: as automation makes expertise abundant, what stays scarce is human. Attention, trust, relationship.⁵ Ideas and output are the commodity now. Scarce is the person who builds trust across departments, keeps the story straight through every review cycle, and puts their name on the outcome. 

AI made that piece of human work the bottleneck, and therefore the most valuable job in innovation. Everyone is arguing which model is smartest, or which workflow saves the most hours. The actual constraint of getting one person the standing to carry a decision through a corporation is the unsexy part nobody talks about, but we think it deserves attention.

The unsexy 80% of innovation

At Bundl, we've come to realize that often we're not required in the room to generate ideas anymore. What we've quietly become experts in is what happens after: navigating stakeholders, protecting owners, and making sure decisions actually get made. Call it decision engineering, momentum work, or simply getting things unstuck. It is the unsexy part of innovation, and it is what moves an idea from the boardroom to reality.

Consider one technology company that ended up deep inside that hairball. Its product should already have been in market: validated with paying pilot users, business case approved, board informed. Yet the launch date had slipped three times.

A closer look revealed a familiar pattern with a new ingredient. More than 40 stakeholders had a voice in the launch; none had the mandate. A single internal champion was making calls without formal sign-off from the departments he depended on, while three workstreams ran in parallel with no visibility into one another. And because everyone now had AI at hand, every review cycle produced fresh, polished counter-material: a risk analysis here, an alternative pricing model there, each one reopening questions that had been settled a month earlier. Without ever deciding to, the organization had made opinions abundant and ownership scarce.

Getting the product unstuck required no new analysis. It required sorting the 40 voices into five decision makers, with the rest formally consulted or informed. It required giving the champion a written mandate, signed by the department heads he had been working around. Every open question became a one-page memo: owner, options, recommendation, deadline. The launch itself got a commit-or-kill date. The committee session that had loomed for months took 45 minutes, and the product went to market eight weeks later.

The idea was never the bottleneck. Ownership was.

So before feeding the pipeline another idea, ask one question first: who owns the last one that stalled?

Sources

1. MacKenzie, Gordon. Orbiting the Giant Hairball: A Corporate Fool's Guide to Surviving with Grace. Viking, 1996.

2. Klein, Ezra. "Why the A.I. Job Apocalypse (Probably) Won't Happen." The New York Times, Opinion, May 3, 2026. https://www.nytimes.com/2026/05/03/opinion/ai-jobs-unemployment-silicon-valley.html

3. Niederhoffer, Kate, Hancock, Jeffrey T., et al. "AI-Generated 'Workslop' Is Destroying Productivity." Harvard Business Review, September 2025. https://hbr.org/2025/09/ai-generated-workslop-is-destroying-productivity

4. Holweg, Matthias, and Davenport, Thomas H. "Don't Let AI Slop Muck Up Your Company's Processes." Harvard Business Review, June 2026. https://hbr.org/2026/06/dont-let-ai-slop-muck-up-your-companys-processes

5. Imas, Alex. "What Will Be Scarce?" Ghosts of Electricity (Substack), 2026. https://aleximas.substack.com/p/what-will-be-scarce