The previous post argued that the AI conversation inside most organizations is repeating the SaaS cycle we've watched play out for the last fifteen years. AI vendors are temporarily receptive to enterprise feedback because their market position isn't locked in yet, but the receptiveness is a phase, and three years from now the platforms that survive the consolidation will be the same opinionated products with the same hardened roadmaps as every other mature SaaS company. The organizations that committed in the receptive phase will be doing workarounds against decisions they didn't make.
If the cycle is going to repeat, the practical question for any organization right now is what to actually do about it. The answer starts with knowing where you currently are, because they may find that they aren't as far along as they think they are further along in the AI progression than they actually are, and that gap between self-perception and reality is itself one of the failure modes the conversation produces. In other words, giving everyone access to ChatGPT/Claude/Gemini/CoPilot, even with an appropriate governance layer, and seeing what spills out on the other side is insufficient.
The progression has five stages. Once you can see them as stages, you can see where most organizations actually sit versus where they think they sit.
The first stage is awareness. The organization has read about AI capabilities, understands at a general level what large language models do, and has formed an opinion about whether the technology matters. Most organizations are past this stage by 2026. The leadership conversations are happening. The board is asking questions. The function exists in the org's vocabulary.
The second stage is experimentation. People in the organization are using AI tools in their personal workflow, whether or not the organization formally sanctions it. They're using it to draft, summarize, code, brainstorm, research. The use is individual and ad hoc. The organization has visibility into some of it and not into other parts. Most mid-market organizations are at this stage right now, whether they realize it or not. The experimentation is happening regardless of what the organization has decided to do about it.
The third stage is operational connection. The organization starts identifying specific operational problems where AI can plausibly help, and pursues solutions for those specific problems. The pursuit may involve buying a vendor product, building something internal, or a combination. What distinguishes this stage from the previous two is that the AI use is no longer general. It's pointed at a specific operational outcome the organization has identified as worth pursuing. A minority of mid-market organizations are at this stage. The ones that are usually got here because someone in IT or operations took the time to map AI capabilities to actual organizational needs.
The fourth stage is systemic redesign. The organization stops treating AI as a tool to bolt onto existing processes and starts asking what processes would look like if they were designed from the floor up assuming AI capability as a foundational input. This is a different kind of conversation from the previous stages. It requires the organization to question the design of work itself, not just augment the design with new tools. Very few organizations are operating at this stage. The ones that are have usually spent years getting there, and the path through this stage is longer than most leadership teams initially expect.
The fifth stage is transformational. The organization has integrated AI into the actual texture of how work gets done, with humans and AI operating in genuine partnership toward outcomes neither could achieve alone. The work doesn't look like the work used to look. The roles don't look like the roles used to look. The outcomes are different in kind from what was possible before. This is the stage the AI conversation usually skips to in vendor pitches and thought leadership, and it's the stage almost no organization actually inhabits. The leap from stage four to stage five is the hardest part of the progression, and it's not a leap that tooling alone can produce.
If LinkedIn is any indicator, my impression is that organizations think they're operating at stage three or four. I have a sneaking suspicion though, based on my complete unscientific, anecdotal experience is that most are actually operating at stage two. The gap between self-perception and reality is itself one of the failures the conversation produces, because the vendor pitches let organizations talk like they're further along than they are, and the absence of operational evidence is easy to miss when everyone is using the same vocabulary.
This is where IT and HR, as organizational functions, are positioned in a way that almost no other functions inside an organization are positioned. Most organizations are running their AI conversations through neither of them, and that's the structural mistake the strongest AI programs have figured out how to avoid.
IT and HR are both fluent in the same kind of organizational change: the kind where a technological shift creates new categories of work, retires old ones, and reshapes how the rest of the work gets done. IT has watched this happen with cloud computing, with collaboration platforms, with the entire CRM, SIS, and LMS markets, and with the data tooling arc described in the previous post. The pattern of vendor receptiveness followed by hardening, of opinionated SaaS products that subtly reshape the workflow they were supposed to support, of organizations buying into a worldview without realizing they'd bought into a worldview, IT leaders who've been in the field for any length of time recognize the shape of what's happening with AI because they've lived versions of it before.
HR has watched the parallel cycle from the workforce side. The shift to remote and hybrid work. The rise of gig and contingent labor. The transformation of customer service from in-house to outsourced to AI-assisted. The recurring waves of automation that have reshaped what entry-level work looks like in every white-collar function over the last twenty years. HR leaders who've been in the field for any length of time have watched workforces absorb technological shifts, sometimes well and sometimes badly, and they know what the difference looks like. They've watched roles disappear and new ones emerge. They've redesigned job descriptions to match new realities. They know, in a way most other functions don't, what it actually takes to move a workforce through a transition rather than just announce one.
The familiarity matters in a more specific way too. Both functions have watched not only the vendor side of these cycles but the work side, where new categories of professional practice emerged out of nothing and accreted into careers that didn't previously exist. The data work that's now its own department in most organizations was, twenty years ago, an adjacency to a few DBA roles. The cloud engineering practice that's now a distinct discipline was, fifteen years ago, a subset of what server administrators did on the side. Each of these cycles produced both a vendor churn and a new profession, and the leaders who paid attention to the second part of that story positioned their teams for the new work. The ones who only paid attention to the vendor part ended up reactive, doing implementations on platforms that other people had figured out the shape of. The AI cycle is going to follow this pattern. New categories of work are about to emerge. The IT function is positioned to recognize the technological side of that emergence. The HR function is positioned to recognize the workforce side. Neither one alone produces a complete picture.
There's a slightly ironic layer to all of this on the IT side. The IT category isn't only the function that's lived through the SaaS cycles. It's also the category that's been doing AI in production, at scale, for years before the current LLM moment. CrowdStrike built a business on ML-driven endpoint detection in an era when most of the industry was still wedded to signature-based antivirus. Abnormal Security demonstrated that behavioral modeling could catch business email compromise that rule-based filters could not. Juniper's Mist platform showed that AI-driven network operations could meaningfully reduce the volume of work landing on the IT service desk. These weren't pilots or vendor demos. They were production systems generating production outcomes years before the AI conversation reached its current pitch.
The reason these solutions found early traction in IT isn't that IT leaders were uniquely visionary. It's that IT leaders were uniquely exhausted. The alert fatigue, the ticket volume, the manual triage burnout, the constant gap between what we needed to monitor and what we had headcount to monitor, all of it created a class of buyer that was unusually willing to actually adopt a tool that promised to do the work for them. There's a line, often misattributed to Bill Gates but actually traceable to a Chrysler executive named Clarence Bleicher in 1947 Senate testimony, that gets at the dynamic: the lazy man will find an easy way to do it. IT leaders weren't lazy, but we were tired enough to behave like Bleicher's lazy man. We bought the AI-driven tools because we couldn't keep doing the work the old way.
The leaders who run the functions where AI has actually been proving itself in production for the better part of a decade are now sitting through the same "AI will transform your business" pitches everyone else is sitting through, as if their teams haven't been operating AI systems all along. The institutional memory of what early AI adoption actually looked like, what worked, what didn't, what the integration costs were, what the vendor lock-in patterns turned out to be, is sitting in IT leadership, mostly unused by the broader AI conversation happening one floor up at most companies.
Both functions, when they're operating well, do translation work. IT translates between business problems and technical solutions: take a vague organizational need and turn it into a tractable specification, take a vendor's pitch and turn it into a clear-eyed assessment of what the product will and won't do for your specific context, take an executive's strategic intent and turn it into operational requirements that can be evaluated against actual implementation options. HR translates between organizational strategy and the people who actually do the work: take a strategic direction and figure out what roles need to change, what skills need to develop, what the workforce implications actually are, and how the change gets absorbed without breaking the organization's ability to do the existing work while the transition happens. Both translation functions are exactly what's missing from most organizational AI conversations right now. The conversations are happening at the leadership level and at the vendor level, with very little of the translation work on either axis that turns abstract intent into specific operational and workforce decisions.
Both functions also sit at the intersection of strategy and operations in a way few other functions do. The CIO is in the rooms where strategy gets discussed and in the rooms where operational reality gets exposed. The CHRO, when the role is functioning correctly, is in the same rooms but looking at the workforce side of both. Together they see both sides of the gap the AI conversation is failing to bridge: what the organization is trying to accomplish, what the operational substrate can support, and what the workforce can actually absorb. That triangulated visibility is what's needed to move an organization through the maturity progression, because the progression requires technological capability, operational understanding, and workforce readiness at the same time. The people who have one but not the others will produce conversations that flail in predictable ways.
I believe that the strongest AI programs will have figured this pairing out deliberately. The IT leader brings the cycle memory, the translation function on the technology side, and the operational read on what's actually tractable. The HR leader brings the same cycle memory pointed at the workforce, the translation function between strategy and the people doing the work, and the read on what the organization can actually absorb without breaking. The pairing produces the kind of conversation that the leadership-level AI discussion is supposed to be having but mostly isn't. It also produces a different relationship to the vendor pitches, because both functions are bringing skepticism informed by having lived through equivalent cycles before, and the skepticism is harder to dismiss when it's coming from two complementary angles instead of one.
This is the most important point most organizations are missing about their own AI moment. The AI conversation isn't an IT conversation that needs an HR implementation arm, and it isn't an HR conversation that needs IT support. It's a conversation that requires both functions in partnership, working at the strategy level rather than being treated as downstream executors. I believe that organizations that get this right end up with AI programs that are technologically sound and workforce-coherent. Organizations that don't end up with one without the other, and the absence of the missing piece eventually shows up as expensive failure.
I'm going to spend the next several posts on this site working through specific cases where I've watched this play out, what worked, what didn't, and what I've learned about how an IT function can actually move an organization through the maturity progression in a way that doesn't just produce another expensive vendor cycle.
The cases are not theoretical. I've been doing this work, in different forms, for a while now. Some of what I've learned is hard-won and would have saved me real time if someone had told me a few years ago. Some of it I'm still figuring out. The point of writing it up isn't to claim mastery. It's to share the work in a form that might be useful to other people doing the same job, and to be honest about the parts where I'm still learning.
The technology will keep moving. The vendors will keep pitching. The cycle will keep cycling. The IT leaders who do the actual translation work between organizational problems and AI capability will be the ones who produce real value over the next several years, and the ones who watch for the new categories of work that are about to emerge from this moment will be the ones whose teams come out of it with more leverage rather than less. The rest will be running pilots, and wondering why the field moved on without them.
— Chris