The corporate instinct is almost universal. A CPG manufacturer, a hotel group, a insurance provider, a retailer with forty years of category data decides it needs to get serious about AI. So it looks inward. Finds the person who knows the business cold, who's intelligent, who's ambitious, who's survived the politics, and who's got the credibility with leadership. Gives them a title. Stands up a steering committee. Calls it AI transformation.
What actually gets built is an unconsious and expensive defense mechanism against real change.
The bottom line is that years of institutional knowledge isn't neutral cargo. It tells you what got approved in the last budget cycle. It tells you what the legacy system can and can't do. It tells you which EVP will kill a proposal before it reaches the third slide. That knowledge, all of it accumulated through genuine hard work, actively prevents the real question from being raised: what does this technology make possible that we've never been able to do before?
Insiders can't ask that question with any real conviction because their career capital is invested in the current systems and answers surviving.
Agentforce, Copilot, and the Cost of Knowing Too Much
Before any CPG, hospitality, insurance, or retail executive decides this is really a technology industry problem, consider what happened to three of the most technically sophisticated companies when they let insiders architect their AI strategy.
Salesforce's Agentforce was built by people who understood the company's existing contracts, pricing logic, and customer relationships better than anyone alive. And then enterprise buyers found that identical customer scenarios triggered different execution paths depending on how the model read intent that day, a pattern Salesforce's own engineers reportedly described internally as a "doom-prompting" cycle.
One buyer, interviewed as part of Redpoint Venture's 2026 CIO research, described the Salesforce AI solution with the sort of precision that doesn't survive a committee review: they expected a miracle and it wasn't happening. The AI layer they shipped was optimized for the business Salesforce already had, at the exact moment that their business needed to be interrogated from the outside.
Microsoft's Copilot story is quieter and in some ways more damning. Their AI product has been available for over two years to a base of 440 million paid Microsoft 365 subscribers. As of early 2026, only 3% were willing to pay for CoPilot. Microsoft's model increased per-seat costs by 40 to 60 percent because it was designed by people who understood Microsoft's enterprise monetization framework almost too well. What they didn't have was any working real understanding for what customers would actually pay for AI-native value rather than AI bolted onto a familiar invoice. You can't price your way out of a preference problem, and you can't see the preference problem clearly when your entire frame of reference is protecting the existing licensing tier.
ServiceNow's situation is different in kind but identical in cause. The company spent years building switching costs so formidable that customers felt genuinely trapped. That moat is now an showing issues of leaking after they rolled out a less than stellar and overpriced AI solution. Access to Now Assist, Servive Now's generative AI layer, requires an expensive upgrade to higher tiers, which has pushed some IT decision-makers toward evaluating in-house alternatives rather than paying for the premium. For a company whose entire value proposition was built on the certainty that customers couldn't leave, that's not a data point they can now ignore.
If this can happen at Salesforce, Microsoft, and ServiceNow with their engineering depth and AI budgets, it can happen in your category management team. It can happen in your revenue management function. And it almost certainly will, if the person driving the effort grew up inside the building.
The Refounding Path
Redpoint's 2026 Market Update made a pointed observation about how traditional software organizations build executive teams versus how AI-native ones do. Traditional organizations hire "been there done that." AI-native ones hire first-principles thinkers, because no playbook exists and speed matters. The implication and stakes for non-technology companies are the same.
The path forward is what Yale's Program on Stakeholder Innovation & Management term as "refounding." Refounding, the kind of architectural rethink that AI actually demands, requires someone willing to walk into a room full of people who built the existing model and tell them, with specificity and credibility, why the whole thing needs to be rethought. It's a disinctly undiplomatic act with a bias against collaborative exercises, requiring someone whose career capital is not invested in the current architecture surviving.
Nobody inside the building can do this because the human brain doesn't voluntarily dismantle the frameworks it spent a career constructing. The sunk cost is cognitive. It's professional identity. The insider who built the integrated business planning system doesn't wake up one morning and conclude, with genuine conviction, that integrated business planning as a discipline needs to be turned upside down. They optimize and automate it instead. They make it more efficient. They apply AI to make the existing process slightly faster and slightly cheaper. And they call it transformation.
Efficiency is not transformation. Read that again for impact. A CPG company that spends three years streamlining innovation modeling while a direct-to-consumer insurgent rebuilds category entry points from scratch has made itself more efficient on the way to becoming irrelevant. The newspaper industry's earnings held up for years while digital was quietly dismantling the structural story. Efficiency gains can look like progress right up until they don't.
What the Outsider Actually Brings
Three things. Almost none of which exist inside the building.
Freedom from institutional history. The outsider doesn't know what customers and internal functions asked for eighteen months ago. Doesn't know what got killed in the last roadmap review. That ignorance is the most operationally valuable thing they bring through the door, because it lets them ask questions that stopped being asked years ago because the answers were institutionally inconvenient.
Comfort with possibility-led thinking versus customer-led thinking. Every experienced product or commercial leader's instinct is to talk to customers and build from there. But customers can only tell you what they already understand. The hospitality customer in 2021 couldn't tell you they wanted individualized real-time pricing on rooms, amenities, and service elements combined because they had no framework for imagining it existed. The outsider builds toward the ceiling of what's technically possible, while the insider builds toward the floor of what's been requested.
Willingness to break things that currently generate revenue. This is the hardest organizational ask and the one most impossible to fulfill from inside. The refounder has to look at a process that works, that generates margin, that has stakeholders who depend on it, and tell those stakeholders it needs to go. An insider's professional identity is partly constructed from the things they're being asked to dismantle. Without exception, without apology, and without anyone in the building having the standing to stop them, the insider will always find reasons to preserve what they built.
The Governance Trap
Non-technology companies carry one additional complication that makes the insider problem worse. They route AI initiatives through IT governance structures built to evaluate and manage risk in stable systems. Unfortunately, probabilistic outputs, sixty-day test cycles, and a willingness to kill things without a lessons-learned document don't survive a twelve-gate intake process. Most governance structures were built for a different time and it almost always outlasts the insurgent trying to work around it.
This is where the CIO question becomes critical in ways most boards don't fully reckon with. The CIO who spends energy ensuring the process is followed and the stakeholders are aligned is genuinely not the right profile for a refounding moment. The organization needs someone willing to compress a decision that used to require twelve people into a room with three, take the accountability for whether it works, and move before the process catches up. That person almost never emerges from inside the building, and they certainly don't emerge from a steering committee.
This is not a cultural problem. The practical requirement is structural. The refounder needs a direct line to the CEO, not the CIO or CDO or CFO or CMO. A planning horizon of six to twelve months, not three years. Explicit permission to kill things that currently generate revenue, stated clearly, with cover when the stakeholders whose processes are being eliminated push back, because they will. And separation from existing IT governance, at least long enough for the first real signal to come back from a real human making a real decision.
Those of us who lived through the early days of eCommere growth and transformation have seen this exact dynamic play out. The eCommerce leaders who actually moved things were the ones detached from sales and marketing and reporting directly to the CEO, protected from the institutional antibodies that would otherwise have quietly neutralized them. The ones who weren't didn't last.
The Acquisition Parallel
Unilever's recent acquisition of Grüns is the cleanest available analogy for what structural protection looks like. Unilever didn't buy Grüns to make it more like Unilever. They bought it to access a growth engine built on assumptions their existing organization couldn't generate internally, and the smart play is to leave it structurally separate. The things that made Grüns valuable, its speed, its voice, its willingness to operate outside the conventions of the category, survive only if they're protected from the gravitational pull of the parent company's processes.
The AI refounder needs the same structural protection. The moment the refounding mandate gets routed through existing digital or technology governance, it becomes an IT or marketing project. The insider logic reasserts itself. The steering committee starts meeting. The intake form gets opened. And three years later, the organization has impressive governance documentation and a handicapped, half-baked AI program it can't quite explain.
The Real Question
Most organizations that read this will nod along and then promote someone internal anyway. They'll dress the role up with a new title and a mandate that sounds bold and then measure the person against last year's metrics. The governance reflex will win. The quarterly earnings conversation will reassert its authority. The refounder, if they're any good, will leave inside eighteen months.
The organizations that get this right will look back on their refounding hire the way cloud-era winners look back on whoever pushed them off on-premise infrastructure. Uncomfortable and counterintuitive at the time. Obviously correct in retrospect.
The question worth sitting with is whether the person you're considering actually has the standing to break the things that need breaking. If the answer requires you to think for more than three seconds, you already know what it is.