Who cares enough about AI to make it a priority?

“Gah, why don’t they care?!”

The most mystifying phenomenon of the rising AI era is how much people differ in whether AI should be a current priority for them (at work) in 2026.

I sure puzzle why certain companies either do or don’t aggressively adopt AI wholesale, especially when a “right” answer seems so obvious to me. But people aren’t idiots. Understand the factors that matter, and you can predict (somewhat) how a company will choose.

In short, companies weren't waiting around for AI to show up. They have things to do and problems to solve. For AI to be a true priority for them beyond dabbling, it has to displace other people, budget, or time priorities.

Even deciding to start dabbling only happens once an innovation has come far enough. But that’s not the focus here. (If you’re curious, I recommend the book Loonshots by Safi Bahcall and Bahcall’s concept of the “Three Deaths” of innovation.)

Back here, we’ll assume that a company has already decided to “do AI.” The question now is why or why not they have made AI a top business priority.

Why have they (or have they not) embarked on an aggressive AI-era transformation of their business and operations in total, especially when the “obviously right” answer to you is the complete opposite of their actions?

The simple answer

It’s quite simple at a high level but gets trickier in detail. Again, let’s assume that the company’s leaders and teams aren’t bozos. In that case, all bets are off anyway.

In sane companies, the decision on whether to embark on a total transformation of their business comes down to whether its people think that (1) the value of adopting AI aggressively exceeds the (2) risks and delaying it for the future, or, if it does not, whether AI at least (3) creates critical, urgent optionality.

That’s because changing a business that works as-is is dangerous. (And I’m saying that as a corporate innovation and startup professional who has dedicated his career to doing just that). If you’ve ever tried to build a business, you know how absurdly hard it truly is. If your hard work, luck, and timing have gotten you the good fortune of a functioning business, don’t just rip it up! Coming up with something new that works may be quite as hard as what you did before. You’re not guaranteed success at all, in fact. And so, it comes back to the balance of:

Value + Optionality > Risks ?

Let’s look at each in turn.

Value

Essentially, people do things that are more of a danger today than all others, or a no-brainer opportunity with low risk and high value, all measured within executives' remaining time in role and incentive time horizon.

If the value of AI is low, uncertain, or will realize later than leaders care about, nobody will act on it. As mentioned, companies have things to do. Nobody has spare time to take on something that’s not a priority.

Even if they decide that “AI” creates value, that can happen in different parts of their business, in solution (product/ service) creation, in solution delivery, and in the administration that makes it all possible*.

*(For business design nerds, the ways I divide the business here loosely maps to the St. Gallen business model and the Doblin ten types of innovation, kinda, and in a simplified way, also to the classic value chain.)

And the balance in which these parts of the business matter can differ by company and by aspect of AI and by aspect of the business just described and across time.

Let's take an example as we go. Let’s say you are a manufacturer of machinery. Here’s how AI might impact your company in different ways and how you might judge its importance:

Solution (product/ service):

You can make AI part of the machines themselves, for example, through self-guided navigation or other ways to achieve sophisticated performance of whatever your machines’ tasks are. When AI is good enough to achieve that, it can have a massive impact for manufacturers. Our manufacturer will care greatly to pay close attention to AI that fits here.

Creation and delivery of solution/ value:

Next, they can use AI in the production of machines, i.e., in factories. In that case, we're talking factory robots, algorithms, and data science to analyze production data, refine what is happening in the production, avoid defects, and source and keep just the right number of components. As before with the product itself, manufacturers can have substantial impact with AI here, especially because these improvements can save costs. That impact is a lot easier to measure predict than revenues improvements.

Unfortunately, factory investments aren’t simple to implement, and they also often require substantial up-front investment costs, not to mention the time it takes to build or upgrade factories. So the decision to go all-in needs to be more thoughtful. The manufacturer in our example will care about the improvements that are possible. But they will check carefully that the value is sufficiently big and certain to be worth the upheaval inevitably caused. Also, Industry 4.0/ 5.0 improvements started long before generative AI. Factories have been improving for years. So AI needs to prove big value to be incrementally useful, over and above what the company may already have implemented. Overall then, AI matters here but only at later stages of maturity when stability and value are up and costs are down from the early days.

Next, we can use AI in the sale and distribution of the manufacturer’s machine. This is somewhat meaningful in a range of ways, from further supply chain optimization and inventory distribution at dealers to efficiency improvements for sales teams. But it is a mere evolution of prior efforts in that direction, not a total, massive change. So AI likely won’t warrant complete upheaval here for the manufacturer in our example.

Administration:

You can also use AI in your headquarters functions. But in manufacturing, to stick with that, headquarters cost is a relatively small share of expenses and its impact on revenue is only indirect (meaning true but hard to assign to a single cause). So AI here fundamentally won’t be a high priority for manufacturers. It will be adopted when there's something obviously useful, maybe as part of other updates. But it will not be something where the company will try to be world leader.

Note that that’s sensible even though a lot of advances occurred in 2026 that help in headquarters functions. They are simply not a big-enough deal in every industry. In addition to the low share of value that AI can generate here, unlocking that value also takes significant cross-functional coordination. So it’s both lower-value and hard. Not really an endorsement to focus here unless forced into it or in specific places with low-hanging fruits.

Risks

The overall decision of AI adoption versus non-adoption, is, as ever, a balance of two risks too, namely risk to adopt on the one side and risk not to adopt on the other. The former is an internally-created capability, and the latter is necessity forced on you by the outside world.

But let’s look at things in more detail. We can use the same categories as for value above and consider AI-related risks in the product, its value delivery, and the administration that enables it.

Solution (product/ service):

Solution obsolescence and evolution describe the risk of new technology, in this case AI, making the thing itself unnecessary. The metaphorical buggy whip that no longer is useful and buggies no longer work is the most prominent example. Conversely, on the positive side, new developments may make former niche products a bigger deal. How that applies to our machine manufacturer is a bit unclear unfortunately. But you can imagine the company’s leaders reviewing their product line one-by-one and deciding whether that product will stop being useful, become massively more useful, or stay about similar in the AI era.

Solution/ value delivery:

For the first time ever, many companies now face a risk of customers becoming competitors. While that had been a risk through vertical integration in the 1980s and its conglomerate thinking era, it's now something that can happen to software companies and those that deliver services of any kind, in particular. A machine manufacturer won’t worry about that risk from its end customers but may consider the risk of others in the value chain becoming competitors. That’s just classic corporate strategy though. It’s not readily apparent that AI will somehow change the game materially in this industry.

As in prior eras of significant change, there may also be new solution delivery channels that rise and previous ones that become less attractive. For example, in the .com era, e-commerce rose as a way to sell and buy basic products, while physical stores, for a while, were seen as outdated. For a manufacturer in the AI era, that’s likely not a big risk. Expensive, bulky, heavy machinery is not likely to be bought by bot anytime soon. Repair and other after-market services might be affected, but even then only at the margins, in things like bot-based order intake or AI-scheduled dispatch. Again, not massive.

Also consider marketing and sales risks through issues like the prisoner's dilemma when marketing and sales become so cheap for many business types that one can now produce it at way higher volumes than before. Good marketing simply no longer stands out. The very technology that makes things "better" also causes its own ineffectiveness. Just like email, at first, was lauded for its incremental speed and then became a problem because it allowed people to send so many emails that the value of each mail declined and people became drowned in communication. This certainly will be a problem in the AI era for most companies. But how much it affects our specific manufacturer depends on how much online marketing they do, vs. sales force-based or other in-person efforts like tradeshows. Overall, this risk also is more of a nuissance than a reason to rebuild one’s total playbook.

Production risks can also emerge in that the new ways of making and delivering things (in this case, for example, machine delivery by autonomous truck) may simply not have been proven out as fully as previous technologies yet. And prior technologies like additive production (3-D printing) and industrial robots already had an impact here for manufacturers. This, too, is likely a low risk specifically in manufacturing, or at least not an urgent one. Leaning into unproven AI-based methods here may instead expose a manufacturer to producing lower quality, which hurts their brand, or even expose them to legal and compliance risks.

Administration:

Interestingly, a lot of legal exposure and crime-defence risks concentrate on the administrative side, or at least enter here. This is the realm of cyber attacks, spam, scammers, market bubbles, and irrational pessimism, accounting problems, and on and on. It matters somewhat less what kind of company we use as our example here. If yours is a company with modern computer systems connected to the internet and a large staff, you need to care about the impact of AI in Finance, Legal, HR, Safety & Security, and so on.

Optionality

Technically, there is a third angle, separate from value generation and risk minimization, for deciding whether a company should go all-in on an AI-era business and operations transformation: That is the nerdy topic of “optionality.” In other words, does AI open up business-level options that matter to have and that we didn’t have before?

But while true, it’s harder to generalize the impact of optionality for example purposes. That’s because hard-to-measure human factors determine much of how a company will act on things they might but, by definition, need not do. Most obviously, the impact of the optionality must also be urgent and critical to matter, and that’s not at all a certainty. So while it’s real, you may need to do custom sleuthing to find out the impact of optionality from AI on a business.

Still, let’s take a moment to consider the some basic types of such options. The list mercifully is simple and short, even if hard to apply in real life.

First, there is the possibility to get permanently ahead, or conversely, the inability ever to catch up, in the face of winner-takes-all dynamics (such as accelerating, reinforcing effects that can visibly have runaway results or market capture effects like natural monopolies or single-homing requirements). In other words, by getting in early, a company may create the option to roll up an entire market. That is the most apparent and critical optionality topic. But it does not always apply. Whether it does so for our manufacturer depends on their geography and the specific kinds of machines they sell.

Second, there is the more hazy impact of unlocking new capabilities and activations, either operationally or strategically. This is a very open-ended option with near-endless possibilities, and so it depends on the company's fundamental interest and aptitude in exploring new spaces and even just thinking about them. The impact could be very real, from changing what work is done inside the company (vs. contracted), to new markets, products, or customer groups, to enabling totally new ways of doing work or completely new types of solutions or businesses. In many ways, that is the domain of corporate innovation groups, which, in turn, come in many types, likestartup partnerships, M&A groups, business design labs, accelerators, and more. So it’s niche-y. Let’s leave it aside for now.

Third, and most abstract, there is the possibility to take a fundamentally proactive, aggressive, opportunistic or defensive, skeptical stance to everything you do. This is a function of the company's overall culture, opportunism, and dynamism. It's hard to describe specifically how this looks because it can take so many forms. For example, in the early 2000s, U.S. retailer Target tried entirely new things that truly had not been done before, such as vertical fashion shows down the sides of buildings. It was not clear how much of an impact such efforts might have. It was the sign of a company simply being proactive and creative and trying things. In that case, it paid off via pop culture relevance, such as being featured on Oprah Winfrey’s show, itself a key tastemaker of the times. That said, considering how this will play out at your company specifically for AI is quite qualitative, i.e., tricky to assess.

Tempering forces

Across all three of these topics - value, risks, and options - limitations and benefits that are inherent to companies and to AI technology moderate or accelerate impact.

You may think of other factors, but here are some of the obvious ones:

First, there's the company's exposure to the physical world. While AI researchers are well aware of the limitations of using AI in the real world, there are also many efforts to overcome those limits (from MCP-type solutions to create interfaces for device control through entire ecosystems like self-driving cars, drones, and robots). But directionally, if you, like the manufacturer in our example, largely operate in the physical world, AI will have more modest or delayed impact.

Second, the complexity and judgment involved in doing your work matters. Not all human knowledge and work are encoded in writing, let alone in deterministic and detailed math or databases. The more your work depends on craft and judgment rather than being fundamentally simple or at least deterministic, the longer it will take for AI to have an impact. At the extreme end of this are types of work that have no predetermined endpoint and where the execution of work helps to shape the goals. For example, in art, it may be fundamentally impossible to advance the concept of what art is with AI because it is fundamentally a dialogue and an emergent goal. Even in the era where AI seems to be able to create “good art” and other judgment-heavy work, there are limitations.

For example, when OpenAI's models help to solve complex mathematical challenges, one reason why the mathematical community did not embrace the findings is because the process of reaching these results unearths new questions that are worth answering themselves. The AI-generated solution does not create such new questions, and so impoverishes the whole discipline. Also consider topics like fashion trends and politics, which involce elements of randomness, complex interactions with messy reinforcement mechanism, and a mix of other factors. They are hard to get “right” with AI.

Third, some work is fundamentally human. For example, even though there has been AI-generated content for religious services and there are AI-based toys that can tell goodnight stories to kids, there exist significant emotional, cultural, and moral barriers to making such essentially-human work automated. That said, there are significant variations in people’s take on this topic. For example, in Japan, robots are relatively common in elder care and support because, simply, there are not enough humans to do the work. On the other hand, even unrelated to AI, in the U.S., Starbucks slowed down their baristas' speed of making drinks because the speed gave customers the impression of Starbucks no longer being a human craft experience but a mere factory.

Fourth, societal attitudes affect overall awareness, embrace, or pushback of AI. Even when something works from a technical standpoint, it can cause rational or irrational pushback that limits to what degree rapid adoption is advisable to companies. The current variations in people’s reactions to data centers in their community represent one such example.

Fifth, there is the topic of gate keepers and others in the value chain. For example, until regulators so decide, crypto companies cannot join the more mainstream finance world. And unless car companies want to build electric cards, makers of electric drive trains have no market for their goods. In AI, similar factors apply, for example, for the sale and export of military-grade AI guidance systems for various weapons, which are subject to strict and capricious political oversight. And various AI-based office tools depend on other companies to be ready to adopt them. Without customer readiness, tool sophistication doesn’t matter.

Sixth, a subtle but critical factor is what companies consider a relevant timeframe for their investment and priority considerations. Privately-held companies often think in terms of entire working lives and what current leaders can pass on to the next generation. Similarly, some tech companies, notably Amazon under Jeff Bezos, consciously choose to compete in efforts that take long to generate returns, expecting simply to win by out-waiting their competitors. By contrast, public companies almost always prioritize the next 1 - 3 years, unusual cases like those dependent on on long-term R&D aside. This is doubly true if individual leaders optimize returns based on their own expected time in position, which can be even shorter. The most extremely and rigidly short-term focused group of companies often includes those owned by private equity investors, who have specific “value generation” plans and timelines that must be met because they, in turn, need to return specific levels of money to their own investors then. (Venture capitalists can vary in their future orientation based on the current market outlook.)

In AI, such discussions may seem irrelevant at first, given the speed at which changes are happening. But consider that we may just be in the very early stages of the AI-era. How patient you can be in waiting for a payoff makes a big difference on when you will adopt how much.

As mentioned, you may think of other factors that affect the overall pace of AI-era business and ops transformation in your industry. But these six already give you very different starting points for assessing what matters specifically for you.

Overall, expect rapid transformation to result once companies see urgent must-dos

In the end, industries will pace their AI-era transformations very differently. The only thing that is in common is that, until it becomes a must-have priority, it is not.

And once you consider the factors that affect what value and options AI unlocks and what risks it exposes companies to, it becomes simpler to understand why some companies evolve their company aggressively for the AI-era … and others don’t yet.

You may still not agree that a given company is acting in its best interest. But at least, you’ll have empathy for why they may act the way they do.

T.I.S.C.