The Functional Identity Crisis Beneath Enterprise AI

A few months ago, I wrote a three-part series on “Contribution”:

• Part 1: How We Create Value vs. How We Recognize It

• Part 2: What Counts as Contribution Now?

• Part 3: Making Contribution Legible

In part 2, I specifically argued that what counts as “contribution” was already blurry before AI.

Fast forward a few months to when I spent four days at Gartner’s marketing conference in Denver, where the dominant topic was exactly what you’d expect.

AI.

The conference floor had all the usual ingredients: tools, bundles, demos, maturity curves, roadmap language, capability maps, and plenty of confident talk about transformation arriving just after the next procurement cycle. (I like that line; it harkens back to my days as an Acquisitions Officer… but I digress).

Some of it was useful. Some of it was familiar. Some of it sounded, to my ear, like productivity with better packaging.

I do not mean that dismissively. Productivity & speed matter. Better access to information matters. I.e., the ability to generate, summarize, analyze, personalize, prototype, and experiment faster is not a small thing.

Unfortunately, one of the odd side effects of overblown AI language is that it can make the practical benefits sound smaller than they are. Still, the conversations that stuck with me were not really about the tools.

They were about the organization and the new connections that would need to be created in the system. Not in the vague “future of work” sense, where everything becomes important enough to become meaningless. I mean the practical operational questions that arise as soon as a large company tries to move from AI interest to AI integration:

1. Who owns the decisions?

2. Who pays for the experimentation, integration, capability building, and change management?

3. Where should AI support judgment, and where should human judgment remain more intentionally in the lead?

4. How do sales, marketing, technology, finance, data, customer experience, legal, HR, and performance management stop treating AI as an adjacent conversation?

And, underneath all of that, what does being in your “function” actually mean for contribution once AI enters the system?

That last question is the one I have been wrestling with for a while now.

Because AI does not just make work faster—it destabilizes functional identity.

And by functional identity, I mean the inherited assumptions about what a function uniquely owns, why it has authority, and how it proves its contribution to the business outcomes.

That may sound a bit dramatic, especially inside large companies where functions are not going away anytime soon. Marketing, Sales, and Finance will still exist. Technology, legal, HR, data, operations, supply chain, customer experience, and strategy will still exist.

I don’t think AI is a eulogy for the functional org chart. Then again…could it be?

Let’s dig in.

For a long time, functions have been able to define their value through proximity to certain kinds of work. Marketing owned the customer language, the brand system, the content engine, and the demand conversation. Sales owned the customer conversation, the account reality, the field signal, and the conversion of strategy into commercial action. Finance owned the economic model, the forecast, the investment discipline, and the performance story. Technology owned the architecture, systems, infrastructure, cybersecurity, and technical enablement. HR owned roles, capabilities, incentives, and talent systems. Legal and compliance owned risks, standards, obligations, and guardrails.

Those descriptions are incomplete, of course. Every function does more than what fits in the shorthand job descriptions. But the shorthand matters because organizations rely on it when deciding who gets invited to the meeting, who has decision rights, who receives funding, and who is presumed to have authority to interpret the problem.

AI now makes some of that shorthand less stable, and the lines considerably blurrier.

Meaning that when a marketer can generate a first-pass research synthesis in minutes, when a finance leader can interrogate customer behavior more directly, when a strategist can prototype a communication plan, when a technologist can draft a business case, when a commercial team can create content variants, and when almost anyone can produce a halfway decent first draft of almost anything, the old borders become harder to defend by assertion alone.

This is where some of the enterprise AI conversation becomes a tad too squishy for how my brain works.

We talk about augmentation, productivity, experimentation, and responsible adoption. Those are all useful words. But within the actual lived-in “messy” enterprise, AI also reveals an older organizational tension: functions are not merely collections of expertise. They are also territories of identity, influence, budget, status, and control.

That does not make them bad. It makes them human.

And if AI changes who can do the work, how quickly the work can be done, and how much of the work can be generated before a traditional functional expert gets involved, then each function has to answer a more demanding question:

“What do we make better in a way that would be meaningfully weaker without us?”

• For marketing, the answer cannot only be more content, faster campaigns, or greater-scale personalization. It should also be sharper customer understanding, stronger judgment about what should and should not be personalized, clearer demand choices, better brand standards, and a deeper ability to connect market behavior with business strategy.

• For finance, it can’t simply be tighter control over spending or another layer of approval around AI investment. It should enable a more honest economic model for value creation, better allocation discipline, clearer productivity assumptions, and the ability to distinguish measurable enterprise value.

• For sales, the answer should not be only more automated outreach, faster account summaries, or an AI-generated CRM. It should be sharper customer sensing, better prioritization, stronger account strategy, cleaner feedback loops from the field, and a more disciplined connection between customer conversations and enterprise choices.

• For technology, it should not be limited to platform ownership or vendor governance. It should be the architecture, reliability, security, interoperability, and technical judgment required to make AI usable beyond isolated experimentation.

• For HR, it should not be another training curriculum. It should be the redesign of roles, incentives, learning systems, talent pathways, and managerial expectations so people are not asked to “adopt AI” while still being measured against yesterday’s job design.

• Legal and compliance have their own version of the same test. Their contribution is not merely saying no with better documentation. It should help the organization better understand risk so it can move with confidence, rather than either freezing or pretending the uncertainty does not exist.

Every function has a legitimate claim. But this is when it gets more problematic…

One common enterprise instinct is to solve ambiguity by assigning ownership. When something matters, we want a name, a box, a leader, a governance body, a dashboard, and a standing meeting. Taken at face value, that usually helps. A company without clear ownership can drift forever in the pleasant fog of collective interest.

But AI is awkward because it cuts across the enterprise too thoroughly to belong cleanly to one function.

• The CIO has a claim because architecture, security, infrastructure, data access, vendor standards, and technical reliability matter.

• The CFO has a claim because investment discipline, funding models, productivity assumptions, value capture, and performance measurement matter.

• The CMO has a claim because customer understanding, content systems, demand generation, personalization, and experience design matter.

• The Chief Commercial Officer has a claim because customer relationships, field execution, channel behavior, pipeline quality, account prioritization, and commercial adoption matter.

• The CHRO has a claim because capability, adoption, role design, leadership behavior, and workforce transition matter.

• Legal, compliance, privacy, and risk leaders have a claim because the costs of getting certain things wrong are not theoretical.

• The COO (or all operations) has a claim because AI that never reaches the workflow is often just an expensive demonstration of possibility.

The problem is that all of them are right.

But this is where I also become skeptical of the reflex to hire a Chief AI Officer or create another executive title to sit above the mess. There may be contexts where that makes sense. Highly technical businesses, regulated environments, fragmented governance, or organizations with no natural integrator may need a more explicit AI leadership role.

But in many companies, a new title can become a clean-looking answer to a problem that is anything but. It may even make the problem worse if the title becomes a substitute for the harder integration work.

If the real challenge is cross-functional integration, then creating a new function focused on AI can become another boundary to manage. It can give the organization the comfort of saying “someone owns AI” while leaving the more uncomfortable questions untouched.

• Who changes the workflow?

• Who gives up budget?

• Who absorbs the transition cost?

• Who owns and manages the notoriously irregular operating costs?

• Who changes incentives?

• Who decides whether the productivity gain becomes margin, reinvestment, faster learning, better customer experience, or simply more work?

A title cannot resolve those questions on its own. In fact, the title can become a hiding place, or worse, a scapegoat.

The better question may not be who owns AI. It may be who integrates the competing claims when every function sees a different part of the truth.

This is where the Chief Strategy Officer, or the strategy function more broadly, may have a more important role than many organizations initially assume.

I say that with some self-awareness. I work in strategy, so there is an obvious risk of sounding as if I have conveniently discovered that the answer is close to my desk.

That is not the argument. The argument is not that strategy should own AI. Nor is it that every company should avoid appointing a Head of AI or Chief AI Officer under all circumstances. Organizational context matters too much to make any universal claim.

The argument is that enterprise AI integration needs a mechanism for coherence.

Someone has to help the organization connect use cases to strategy, strategy to resource allocation, resource allocation to operating model implications, operating model implications to actual behavior, and ensure that behavior delivers incremental value. Someone has to make the trade-offs visible. Someone has to ask whether the organization is using AI to improve contributions or simply to preserve the inherited shape of the function, albeit with arguably better tools.

At its best, that is strategic integration work. In some companies, the formal strategy function may be well-positioned to facilitate it. In others, the same work may sit elsewhere.

The important thing is not the departmental label; the question is whether the integration work is actually being done, measured, and adaptable.

Strategy as integration is different from strategy as a deck factory. It is different from the annual planning theater. It is different from a clever narrative superimposed on decisions already made.

It is the discipline of choosing where the company is headed, what must change to get there, which capabilities matter most, which trade-offs are unavoidable, and how the organization learns when reality refuses to follow the plan. (See also: “Where to Play and How to Win”).

That kind of strategy work does not replace the CIO, CFO, CMO, CHRO, legal, data, or operations leaders. It needs them. It should respect the depth of their domains.

But it can help prevent AI from becoming a set of functional or fractured subplots, each operating in siloes:

• Marketing builds an AI content engine.

• Sales builds an AI-enabled account planning process.

• Technology builds a platform roadmap.

• Finance builds an ROI model.

• HR builds a training curriculum.

• Legal builds a governance process.

• Data builds standards.

• Operations runs pilots.

Each activity may be reasonable. Each may even be well managed. And still, the enterprise may not have changed very much—the work becomes faster without becoming better.

The company generates more ideas without improving its choices, teams create more content without becoming more relevant, and leaders review more dashboards without making stronger decisions. And perhaps most importantly, employees can save time without anyone agreeing on what the time saved is for.

A great deal of AI enthusiasm rests on productivity. But time saved is not automatically value created. It depends on where the time goes, which decisions improve, which work stops, which work becomes possible, and whether the organization has the discipline to capture the benefit rather than simply filling the space with more activity.

In my mind, this is one reason AI creates such an interesting contribution problem.

When production gets easier, judgment becomes more exposed.

If an AI system can produce ten decent options, the scarce skill is not producing the eleventh. It is knowing which option deserves attention, which should be discarded, which is merely plausible, which is on-strategy, which is off-brand, which solves the wrong problem elegantly, and which introduces risk that the organization does not yet understand.

The phrase “human in the loop” is often used in AI conversations and is directionally helpful. But I sometimes wonder whether it lets us off too easily. A human in the loop can mean careful judgment, accountability, interpretation, and ethical responsibility. It can also mean a person checking a box in a workflow that was never redesigned in the first place.

The point here is not to keep humans symbolically present. The point is to understand where human judgment changes the quality of the outcome and creates incremental value.

That requires functions to examine themselves with more precision. Where are we adding judgment, context, standards, and trust? Where are we improving decisions? Where are we deepening our understanding of customers—are we improving our customers’ experiences? Where are we helping the organization allocate resources more wisely? And where, if we are honest, are we mostly protecting our historical proximity to the work?

Those are not comfortable questions. They are also not indictments. Every function has accumulated work that made sense at one point and now persists because the system knows how to keep doing it. AI did not create that problem.

This is why I believe that the conversation about ownership and funding matters so much:

• Ownership determines who gets to shape the work.

• Funding determines what the organization is actually willing to change.

• Contribution determines whether the work deserves to exist in its current form.

If AI is treated only as a technology investment, the organization may underweight adoption, behavior, incentives, and workflow redesign. If it is treated only as a finance exercise, the organization may over-focus on cost takeout and under-invest in learning, customer relevance, or capability building. If it is treated only as a marketing opportunity, the organization may generate more demand activity without resolving data, governance, measurement, or economics. If it is treated only as a talent initiative, the organization may over-index on training and underweight decision rights, systems, and the actual redesign of work.

None of those functions is wrong—each is simply…incomplete.

Net, I’d argue that the enterprise question is how to hold the incompleteness long enough to design something better.

The best conversations I had in Denver were not with people pretending they had the whole thing figured out. They were with colleagues and practitioners trying to make sense of what they were hearing in real time. Comparing notes. Challenging assumptions. Asking what would actually work within a company with a history, incentives, systems, constraints, personalities, governance, and customers who do not care how elegant the internal operating model looks.

Most large organizations are not blank sheets of paper waiting for the future. They are living systems with inherited structures, embedded routines, and functions that have learned how to survive, deliver, defend, translate, and matter.

AI enters that system… and the system reacts to it like any other symbiotic organism: what. is. this.

So the work is not just to ask, “How do we use AI?” The work is to ask what AI reveals about the company. Where are the boundaries useful? Where are they performative? Where does functional expertise improve the outcome? Where does functional ownership slow the learning? Where are we funding ambition without funding the conditions that make ambition real? Where are we using the language of transformation to avoid the discomfort of reallocation?

And where are we asking people to adopt new tools while leaving the old measures of contribution untouched?

That may be the part leaders need to sit with before demanding “more AI” or making declarative statements about being “AI-first”.

I’d argue that the more AI compresses production, the weaker production alone becomes as a basis for functional authority. That does not make expertise less important. It makes it easier to test the quality of that expertise.

It may also reveal that some functions were adding more value than the rest of the enterprise understood. It may reveal that others were relying too heavily on ownership of activity as a substitute for contribution.

Most likely, it will do several of those things at once.

That is why I keep coming back to the functional identity question: functions may need to become more explicit about why they remain essential… but perhaps in a new form. A new operating model.

And finally, I don’t think that every company needs to make Strategy the owner of AI. Nor do I think that every company should rush to hire a Chief AI Officer or Head of AI and assume the integration problem has been solved.

In some cases, that role may help. In others, it may create another boundary around work that already suffers from too many boundaries.

The whole point is that AI integration will likely expose whether the enterprise has any real mechanism for resolving cross-functional tension when everyone is partly right.

That mechanism might sit with strategy. It might sit with a transformation office. It might sit with an enterprise AI council that has real decision rights rather than ceremonial alignment. It might sit with a CEO willing to demand clarity. It might take different forms depending on the company.

But it has to exist.

Because as ideas become more abundant and execution gets compressed, functions become less able to justify themselves by proximity to work alone.

They have to explain what they make better, what they are willing to fund, what they should own, what they should enable, what they should stop protecting, and where they are prepared to let the work change.

All in service to better outcomes.

Simple, not easy.

Additional references that informed this piece:

Gartner’s 2026 CMO Spend Survey supports the “investment ahead of readiness” point: CMOs are allocating an average of 15.3% of marketing budgets to AI, while only 30% report mature or fully developed AI readiness; Gartner also notes the risk of investing in tools faster than building data foundations, processes, governance, and talent.

Gartner’s Chief AI Officer guidance supports the skepticism toward rushing into a new C-suite title. It argues AI needs leadership, orchestration, and multidisciplinary governance, but does not necessarily justify a dedicated C-suite position.

McKinsey’s 2025 State of AI survey supports the operating-model argument: organizations are redesigning workflows, elevating governance, and assigning senior leaders to AI governance, but many still have not seen enterprise-wide bottom-line impact.

Deloitte’s CSO framing supports the strategy-as-integrator angle, particularly its view that CSOs connect AI activity to strategic objectives, cost/ROI, collective business value, and CxO alignment.

Share Big Things F@$t™