Annie Tsai

I was on a flight back from Southern California this past weekend, sitting next to someone I had recently met — a growth advisor whose work takes him inside mid-to-large companies on a regular basis. As we were chatting about our lives and work, and I started sharing some of what we’ve been building at Interact over the past year: the top-of-funnel agentic outbound systems, the interactive digital user-generated content moderation pipeline, ICP search and personalized profile enrichment. The kind of work that when you’re in the middle of just feels like solving one hard problem at a time in service to scaling a business.

He listened, and then he said something that I’ve been gnawing on ever since: most companies can’t and aren’t doing what we’ve built.

So I pushed back a little. I asked him what he was seeing across his clients — what was actually getting in the way. He talked about the gap between mandate and skill, the strategic disconnect, and the simple physical reality of how hard it is for a large organization to turn the Titanic fast enough to keep pace with a technology that is moving at a fundamentally different velocity than anything most of us have managed through before.

And that’s what got me. Because I’ve been watching something interesting happen in software companies for years — the shift from waterfall to agile to continuous development on the back end has been one of the great operational stories of the last two decades. And yet those same companies, externally, still have to bucket their changes into thematic releases, into manageable chapters, because their customers and their teams can only absorb change at a certain pace no matter how fast the actual business can move.

AI adoption inside organizations is running directly into that same wall. The technology may be on a continuous deployment cycle, but organizations are still running on waterfall. Well, because humans are humans and change management is the oldest and most important strategic management skill an operating team must master. And a mandate can never close that gap — it just makes it more visible.

The Numbers Tell You Where to Look

The gap between what leadership believes is happening and what’s actually happening on the ground is measurable.

Deloitte’s State of AI in the Enterprise 2026 found that access to AI tools has increased 50% year over year, with 60% of employees now having access. But fewer than 60% of those employees use the tools regularly, which means the investment in licensing and rollout is running well ahead of the actual behavior change it was meant to drive.

Here is where it gets interesting. A January 2026 survey from CIO and BlackFog of 2,000 workers at companies with more than 500 employees found that nearly half of workers admit to adopting AI tools without employer approval. And more alarmingly, 69% of C-suite members and 66% of senior VPs appear to be fine with this, prioritizing speed over policy. So the people issuing the mandates are themselves operating outside the guardrails. That tells you something important about the credibility of the mandate as a change mechanism.

Meanwhile, Larridin’s State of Enterprise AI 2026 found that 45.6% of organizations don’t even know their own workforce AI adoption rate. And for the organizations where adoption is happening, the gap between the people doing it most and the people doing it least is a 4x difference — same organization, same leadership, same budget, radically different intensity. The frontier workers — the top 5% by adoption — are sending 6x more AI messages than the median employee.

The story here is that adoption isn’t completely absent, but it is this: underground, uneven, and compounding in ways that most leadership teams don’t have visibility into yet.

Why AI Adoption Is Stalling — And It’s Not One Thing

My new friend the growth advisor was right that there’s no single clean explanation. What I’ve seen, both from building inside a company and from the research, is three forces operating at once — and most organizations are only equipped to address one of them at a time, if that.

The first is infrastructure and readiness. The tools are moving faster than the underlying systems need to absorb them. Deloitte’s 2026 report found that only 40% of organizations say their AI strategy is highly prepared, governance readiness sits at 30%, and talent readiness sits at just 20%. Companies feel significantly more confident at the strategy level than they actually are at the execution level — which means the mandate is often issued from a place of strategic conviction that the operational reality hasn’t yet caught up to.

The second force is the identity and psychology piece, and this is the one that leadership decks skip most often. When someone has spent years being the person in the room who knows how to do something — who holds the institutional knowledge, who is the go-to for a particular kind of judgment call — a tool that can approximate that knowledge in seconds doesn’t feel less like an upgrade and more an audit of their value. BCG’s January 2026 research is clear on this: companies realizing the most value from AI have the most ambitious upskilling programs. The inverse is equally true — organizations stalling out are doing so on people, not technology. And Informatica’s CDO Insights 2026 adds an interesting layer: 65% of employees believe the data behind AI is solid, yet 75% of data leaders say those same employees need serious upskilling in data literacy and 74% need AI literacy. Blind trust without fluency is its own form of resistance — it looks like adoption, but it isn’t.

The third force is change velocity, and this is where the waterfall parallel belongs. McKinsey’s Global Tech Agenda 2026 — based on a survey of more than 600 technology and business leaders — found that layering AI into the tech stack may actually be the easy part. Teaching people how to use it effectively, and inspiring them to embrace it in their daily workflows, is harder. Nearly a quarter of top-performing companies, compared with just 15% of others, cite change management as a core challenge to scaling AI. And General Assembly’s report from SHRM’s AI+HI Project 2026 — drawing on 500 HR leaders — found that two-thirds of HR professionals say their organization hasn’t been proactive in preparing employees to work alongside AI. The rollout is outpacing the readiness, and the readiness is what determines whether the rollout actually lands.

A mandate addresses none of these three forces. It is a direction without a map, issued to teams that are managing an infrastructure problem, an identity problem, and a change velocity problem simultaneously.

The Missing Bridge

Here is the thing that doesn’t show up anywhere in the research, because it’s structural and most organizations haven’t named it yet.

What I was describing on that plane — the agentic outbound system, the moderation pipeline, the enrichment work — none of that happened because someone issued a mandate and hoped for the best. It happened because there was someone in the room who could simultaneously see the business outcome that was needed, understand what the technology was actually capable of doing, and translate between the two clearly enough to ship something real.

That role — whether you call it a technical operator, an AI systems builder, an in-house architect — is what’s missing inside most organizations that are struggling with adoption. And a mandate can’t create it. A licensing agreement can’t create it. A strategy deck can’t create it. Deloitte’s 2026 report found that insufficient worker skills are the biggest barrier to integrating AI into existing workflows, and that the number one talent strategy response is educating the broader workforce to raise AI fluency. That’s the right instinct, but far fewer organizations are re-architecting roles, workflows, and career paths — which is where the translation layer between strategy and execution actually has to live.

Larridin’s AI Hiring Pulse from February 2026, tracking 428 companies across 43,422 job postings, found that Product, Customer Success, and Engineering lead in AI hiring and adoption, while Finance and Legal lag behind — a 4x gap between the top and bottom functions within the same organizations, with the same leadership and the same budgets. The functions moving fastest are the ones that have people who can connect a business problem to a technical solution. The ones moving slowest are the ones that don’t.

Before you issue another mandate, the more useful question is whether you have someone in the room who can actually bridge the gap between the problem you’re trying to solve and what the technology can do. If the answer is no, that is where to start. You can mandate the destination, but you would be very lucky to mandate or snap your fingers into the bridge builder.

What Actually Creates Momentum

The organizations that are seeing real adoption got there because someone made the first win feel safe, small, and worth repeating. Here’s what that actually looks like in practice:

• Start with the most complained-about workflow, not the most strategic one. From a product standpoint, this is table stakes — you don’t launch a new product into your most complex use case. You find the fastest path to a felt win, you make it visible, and you let word of mouth do the internal work. General Assembly’s findings from SHRM’s AI+HI Project 2026 are worth sitting with here: AI isn’t an ordinary technology training program. It is a full-scale culture and change adoption process, and the hardest skills to build aren’t technical — they’re human. Start where the pain is most felt, and let the early win do the culture work for you.
• Find and elevate your internal champions. Peer-to-peer carries more weight than top-down, every time. A colleague showing a colleague what’s possible in their actual workflow beats an IT rollout deck by a wide margin. The question to ask yourself right now: who on your team is already using AI without being told to? That person is your first champion. BCG’s January 2026 research found that the companies realizing the most value from AI have the most ambitious upskilling programs and the resources committed to support them — the common thread across high performers is deliberate human investment running alongside the tool investment, not following it.
• Stop measuring license activation. License activation is a vanity metric. It tells you the tool is on. It tells you nothing about whether anyone is using it to do real work, or where they’re getting stuck, or what they’ve abandoned. Larridin’s 2026 measurement framework identifies the top barriers to useful AI adoption measurement as unclear responsibility (30.5%), fragmented ownership across teams (27.7%), and no correlation drawn between usage and actual outcomes (24.4%). These are organizational problems, not technical ones. If you don’t know what you’re measuring, you can’t improve what you’re building.
• Reframe the message from transformation to personal time savings. The word “transformation” has been worn down to almost nothing at this point. Every team that has sat through an AI strategy presentation has heard it. What actually moves people is a concrete, personal, low-stakes first experience — the answer to the question that every person on your team is already carrying: what is in this for me, this week, in the work I am actually doing? Informatica’s CDO Insights 2026 captures the paradox well: nearly 7 in 10 organizations have adopted GenAI, and almost half have moved into agentic AI, but confidence isn’t keeping pace with adoption. The trust gap is a communication and experience problem, and the fastest way through it is a win someone can feel in their own hands.

Your Takeaway This Week

Make one move this week to empower the people on your team to build real-world, high-value skills. Give them the license to play with it, ask questions, and run tests.

Help identify the mundane, low-stakes workflows your team complains about most — the two-hour weekly report nobody wants to write, the meeting recap that falls through the cracks every Thursday, the first draft of the thing that sits on someone’s to-do list for three days before anyone touches it. That is your starting point.

Show one person on your team, in twenty minutes, what changes when they let AI take the first pass. You don’t need a training program, tool announcement, or top-down mandate to drive change. A visible, felt win in a workflow that was already costing them time they didn’t have can be enough to trigger the tsunami.

That is how mandates become momentum.

If you found this issue useful, share it with a friend. I wrote this newsletter for us — the people who are building the future at work.

– Annie