Your Workflows Are the Problem, Not Your AI Tools

In this week’s episode of Voices of Search, we spoke with Stephen Wunker, managing director at New Markets Advisors and author of AI and the Octopus Organization, about why most companies are getting AI adoption exactly backwards.

Stephen has spent more than a decade helping Fortune 500 companies navigate AI transformation, advising organizations like Microsoft and Meta along the way. His argument is a direct challenge to how most leadership teams are currently operating: the technology isn’t the problem. The organization is. And until companies stop layering AI on top of the same old structures, approval chains, and incentives, they’re going to keep seeing the same disappointing results.

The conversation covered why 85% of AI value gets created by 15% of projects, what the octopus can teach us about organizational design, and why every manager in your company needs to become a change manager—whether they signed up for that or not.

Key Takeaways From This Episode:

• Bolting AI onto existing workflows produces marginal gains at best. The real value comes from redesigning the system of work, not just adding a new tool to it.
• At Johnson & Johnson, 15% of AI projects generated 85% of the total economic value—a ratio that held because most pilots never changed how work actually got done.
• The octopus organization model decentralizes intelligence so the right people get the right data at the right time, without constant coordination overhead at the center.
• Leaders need to define three buckets before any AI implementation: what humans won’t do, what they shouldn’t do, and what they simply can’t do. Each requires a different approach.
• Governance and data quality are prerequisites, not afterthoughts. Without them, most organizations aren’t ready to capture the full value of AI, regardless of budget or headcount.
• In three years, companies will regret treating AI as an IT rollout rather than a fundamental business process transformation that touches HR, operations, and go-to-market strategy.

The Assembly Line Problem

Stephen opened with a historical analogy that reframes the entire AI adoption conversation. When electricity arrived, factory owners didn’t redesign their operations around the new technology. They just swapped out steam-powered machines for electric ones and called it progress. It took 35 years before Henry Ford realized electricity made the assembly line possible—and that was the real unlock.

“With AI, we don’t have 35 years,” Stephen said. “But we need to be looking for those systematic changes.”

The Johnson & Johnson example makes this concrete: J&J tallied up their active AI pilots and counted at least 900. When they mapped those pilots to economic value, 15% of projects were responsible for roughly 85% of the total value created. The other 85% of projects—the ones that just sprinkled AI onto existing tasks without changing the underlying workflow—produced very little.

Stephen’s diagnosis: “If you’re not changing the system of work, you’re probably not creating a whole lot of value. You might be ironing out a couple of points of friction.”

The implication for marketing teams is direct. Using AI to write the same reports faster, produce the same content more efficiently, or speed up the same approval process doesn’t change what the report is for, whether the content strategy is right, or whether the approval process should exist at all.

What the Octopus Gets Right

Stephen’s book draws on a genuinely unusual piece of biology. An octopus has nine brains—one central brain and one for each arm—connected by a nerve ring that keeps every arm aware of what the others are doing. Each arm can sense, think, and act independently. The whole organism moves fluidly without needing central approval for every action.

It’s a useful model for how AI can restructure organizations. Intelligence gets pushed closer to where decisions are actually made. Coordination overhead drops. The center stops being a bottleneck.

Stephen illustrated this with a hospital marketing department that was constantly being pulled in different directions by internal stakeholders across service lines. Cardiology, obstetrics, and urology each wanted their own campaigns. The marketing team was spending enormous energy helping those stakeholders figure out what they even wanted, before a single piece of work got created.

The octopus solution: use AI to give those stakeholders self-service tools so they can draft their own materials, discover their own needs, and bring something workable to the marketing team rather than a blank-slate request. Marketing shifts from doing the 0-to-80 on low-priority requests to coaching from 80 to 100 on the work that actually matters strategically.

“That’s the octopus,” Stephen said. “You’re devolving authority and enabling people to know what’s going on without getting super involved in every project.”

The Fence Paradox

One of the more counterintuitive ideas in the conversation came from a landscape architecture study Stephen cited in the book. Architects designed a playground with a big slide in the center and equipment around the edges rather than a fence. Kids clustered in the middle. Then the architects added a fence around the perimeter. A new group of kids came in and immediately spread out to use the full space, right to the boundary.

The lesson: implied fences are almost always closer than real ones. People operate inside an imagined constraint that’s narrower than what’s actually permitted. To give teams real freedom to be agile, you have to define the fence explicitly, because without it, everyone defaults to a smaller, safer territory.

“You need to lay out the fence paradoxically to give people the freedom to explore all the territory that they have,” Stephen said. “Oftentimes, the area for agility is a whole lot bigger than people think it is.”

For marketing teams implementing AI, this translates directly. Rather than issuing thick manuals about what people are allowed to do with AI tools, define what they can’t do, and let everything else be fair game. 

Stephen’s analogy here was sharp: Air Force pilots get a thick manual of permissions. Navy pilots get a thin manual of prohibitions, because carrier landings demand in-the-moment judgment that no manual can fully anticipate.

Treat your teams like Navy pilots.

Can’t, Won’t, Shouldn’t

Stephen’s three-question diagnostic framework is one of the most practical things to come out of the conversation, and it’s worth slowing down on.

Before any AI implementation, ask: what won’t humans do, what shouldn’t they do, and what can’t they do?

→Won’t covers tasks that technically get done but inconsistently and poorly — like summarizing a meeting and distributing notes to everyone who attended. Nobody actually does this well. AI handles it without thinking.

→Shouldn’t covers tasks humans are performing that aren’t a good use of their skills. Stehen’s example was physicians spending half a patient consultation staring at a keyboard. Doctors didn’t go to medical school to become typists. AI scribes that capture and synthesize notes in real time don’t eliminate doctors — they let doctors practice medicine. Patient satisfaction improves. Physician job satisfaction improves. Notes get better. Everyone wins.

→Can’t is where Stehen sees the most underinvestment. This is where genuinely transformative change lives. He gave the example of market segmentation: most organizations cap out at four or five customer segments because human teams can’t operationalize more than that. A sophisticated gaming company might run 50 or 100 segments—not because they have more people, but because machines act on the segments. AI makes this level of personalization accessible to organizations that would never have attempted it before.

“If that’s all you’re focused on,” Stehen said of the won’t and shouldn’t categories, “you’re thinking about making the world as it is a little bit more efficient. You’re not thinking about really transformative types of change.”

Organizational Antibodies

Every organization has an immune system, and it doesn’t discriminate between threats and improvements. Stehen calls these organizational antibodies, as in the insidious forces that quietly reject change even when leadership is nominally committed to it.

Some are obvious: poor data quality, absent governance structures, and no clear measurement of AI outputs. 

Others are invisible: work getting quietly shifted to another department instead of genuinely changing, tool adoption without any change to the underlying process, or the gap between what an AI implementation was supposed to do and what it actually changed day-to-day.

The antidote is to anchor every AI initiative in a specific business objective—not “implement AI” but “reduce our campaign cycle time by 50%” or “increase customization for enterprise customers.” 

When the objective is clear, AI becomes a means to an end that people actually want to achieve, rather than something imposed from above.

Stehen was direct about the job-cutting narrative many companies have used to frame AI adoption: “We’ve already seen some high-profile announcements get walked back.” Salesforce, Duolingo—organizations that announced sweeping AI-driven headcount reductions and then quietly revised the story. 

The more durable approach, in Stehen’s view, is deploying people on higher-value work, not eliminating them to hit a quarterly number.

The Lightning Round

Our breakout lightning round pulled out some of Stehen’s sharpest thinking on where organizations actually break down—and what leaders should be doing differently right now:

→Biggest mistake leaders make when rolling out AI? Treating it as a technology implementation rather than a system-of-work change. “Put tech at the table, but look at it as the whole system you need to change.”

→One sign an organization isn’t ready for AI transformation, regardless of budget? No data governance and no guardrails. “If you don’t have good data, you don’t have good governance, then you probably want to attend to some of that first.”

→Where do marketing teams underestimate the human side? They focus on tool adoption and miss everything downstream—how work flows change, how decision rights shift, how visibility into each other’s data changes team dynamics. “You need to think big before you zero in on how particular tools get used.”

→What belief about productivity is AI breaking? The one AI itself created—that it would eliminate massive numbers of jobs. “I’ve got a word processing capability, and I’m still gainfully employed. Let’s think strategically about it and not narrowly in terms of how we can get rid of 50% of jobs.”

→What will companies regret not changing sooner? Treating AI as an IT rollout. “It is not just an IT change, it’s a business process change. AI can transform what your value proposition is. It can transform your channels and how you go to market. Many companies have not yet been taking those steps.”

AI Is Ready. Most Organizations Aren’t. That’s the Real Problem.

Technology isn’t what’s holding companies back. The tools are good, they’re getting better every year, and access isn’t the issue. What’s lagging is the organizational will and structural readiness to actually change how work gets done.

Stehen’s framework gives leaders a place to start. Define clear aspirations that actually help people make choices. Build real mechanisms for change, not just tool access. Cultivate the culture and operations that make new ways of working stick over time. And make every manager a change manager—because this can’t be handed off to IT or a center of excellence and considered done.

The organizations that close the gap between adoption and transformation won’t just be more efficient. They’ll be doing things their competitors structurally can’t, all because they intentionally built the organization to use them.

Voices of Search is a daily SEO and content marketing podcast hosted by Jordan Keone and Tyson Stockton. The show delivers actionable strategies and data-driven insights to help marketers navigate the ever-evolving world of search engine optimization and content marketing. New episodes air weekly, covering everything from technical SEO to AI discovery, featuring industry leaders and practitioners sharing real-world frameworks and proven tactics.

Subscribe to Voices of Search on Apple Podcasts, Spotify, or your favorite podcast platform. Follow Previsible on LinkedIn for updates and subscribe to the VOS YouTube channel for video episodes and clips. You can also visit the official VOS site to explore the full episode archive and submit your SEO questions for future episodes.