Corporate AI Strategy, my ass
Author: Iris Meredith
Date published: 2026-11-19
If you've been on LinkedIn at all over the last few years, you will almost certainly have found it completely inundated with talk of AI. This is nothing new: a lot of us tech writers have written quite a lot about it. Most of what is said on LinkedIn and in many other spaces is, of course, nonsense. Still, if you observe carefully, you can see some patterns, as different people tend to produce different kinds of stuff about it. Tech people and the startup crowd get very enthusiastic about vibe coding and the kinds of agent that wind up draining your bank account buying two tonnes of cream corn. Other professionals tend to get very enthusiastic about research and document/image generation. The rationalist crowd who are expecting the singularity are always around, of course. And then there are the executives at large companies, who spend an awful lot of time writing LinkedIn posts, articles in the industry press, podcasts, conference presentations and any other media you can think of about what they call "AI strategy".
Reading these documents or watching these talks is an exercise in inflicting suffering on yourself. If you know anything about machine learning, AI or the industry that the executives purportedly work in, it becomes immediately obvious that the executives in question simply do not know what they're talking about. The writing or presenting style is uncomfortably friendly, as though they're trying to make superficial charm cover for a lack of substance. And, inasmuch as you can pick out a thread of argument it all, it's often disjointed, with a whole pile of individual points that don't add up to anything, almost like a Gish gallop that aims to give the impression that they have a coherent AI strategy.
While we are slowly building stability and building more consistent income, we do still struggle to make rent and cover bills some weeks, and this is unfortunately one of them. If you'd like to contribute to helping keep me writing, then, you can do so via Patreon, Liberapay or Stripe.
If you've had a bit of exposure to strategy in other fields (especially in fields outside of business), you might find this a little troubling beyond the simple fact of the executives being completely ignorant about AI technology and the industry that they're working in. In fact, you might wonder a bit about what the hell these executives are doing when they're doing strategy, and why their discussion of strategy is always so damned vague. These people are hired and paid large amounts of money for this, after all: shouldn't there be a bit more substance to it all?
The somewhat unfortunate answer is that executives, as a general rule, aren't doing strategy as it's understood in military and political fields. In fact, executives, having neither the education, nor the formation, nor the aptitude for genuine strategic thinking, are usually incapable of doing strategy at all.
That's a serious claim, to be sure. After all, the bulk of the average person's exposure to any kind of thing that gets called strategy is precisely this kind of strategy. This means that to back up the argument here, we need to discuss what strategy, in the fields where it actually matters, is. We need to demonstrate what a serious strategic analysis looks like, in fact, in order to show why what the executives are doing, about AI and everywhere else, is very much not it.
What is strategy?
In military theory, we divide our military analysis into a triplet of layers: tactics, operations and strategy. Tactics discusses the details of battles: how do we expect to actually win a fight between two groups of soldiers in the field? Operations is about logistical art and co-ordination: how do we get the correct supplies and soldiers to the correct place at the correct time in order to have and win the correct fights? And finally, strategy is about the overarching war goals: why are we fighting, what are we hoping to get by fighting, and why and when might we not fight or make peace?
These layer of analysis are nested in a hierarchy: tactics sit inside operations, and operations sits inside strategy. It's not difficult to see why this is the case: it doesn't matter if you win every battle you fight if the battles you fight are failing to advance operational goals, and it doesn't matter if you're advancing operational goals if those operations aren't getting you to where you want to be strategically. Pearl Harbour, for example, was a well-executed operation that was a disastrous mistake from the perspective of Japanese strategic ends, as Japan really could not afford to draw the USA into the war.
Strategy is the outer layer of this: it's where we reckon with how we use what we have, in the wide sense of the term, to attain our goals, and also where we think about what our ends are and what they should be. In the military context, this is the point at which the military makes contact with politics and economics: as Clausewitz puts it, "war is the continuation of politics through other means", and in any study of military strategy it becomes clear that the prosecution of a war or the running of a military is unavoidably political.
It's not that hard to extend this definition to other organisations, and in particular to the kinds of companies and institutions that most of us work in. We might not be fighting a war exactly, but the principles are similar: we have our most granular layer of how we expect to perform the basic functions of what our institution does (manufacturing, teaching, healthcare, those kinds of things), then the operational and logistical art that goes into making sure that the basic functions are done in the right way, in the right place and at the right time. Finally, the strategic layer involves figuring out what the organisation wants to do in the long run, what needs to happen for that to be achieved and how that can be done using the means that the company or institution has. This means that "business strategy" is not an oxymoron: it's quite possible and reasonable for such a thing to exist, and we would, in fact, all be better off if more of it happened than is currently the case.
If business strategy can be a thing then, and if people can reasonably be expected to analyse it, we might fairly ask what skills and traits a person might need in order to be able to effectively perform strategic analysis, or to "do strategy". And from our definition, we can make some inferences.
First off, if strategy involves understanding and pursuing strategic ends, it's important that the person doing the analysis knows enough about what those strategic ends are. It doesn't do to do strategy if you don't understand where the organisation your strategising for wants to go, what it's trying to do, or what, given its positioning, it should be trying to do. Similarly, you probably need to know quite a lot about what means you have available to you, what means you could have available and what other people have. Finally, you need to be able to put together a whole lot of complicated, blurry ends and a lot of complicated, blurry means in such a way that you end up with a couple of important ends to pursue and choose means sensibly so that you wind up with a set of ends and means to meet them in a way that matches your criteria.
Given that this is ambiguous, often not something that you can really put a number on and difficult to put into words, strategic analysis (pace Clauzewitz) is much more an art than it is a science. In fact, aesthetic considerations of unity and intensity are very important in judging the merit of a strategy. Doing strategy well is thus as much a factor of aptitude and personality as it is knowledge and experience: the strategist has to have the kind of personality that demands that the composed strategy makes sense and is of high quality while being OK with the fact that there are no hard and fast rules for identifying when that's the case.
So, how do our AI strategists match up to these standards?
Ends
Not so well, as it happens. The average executive, to put it mildly, does not have a clear view of their strategic aims. We're going to demonstrate this by performing a basic strategic analysis for a country and demonstrating what goes into one, which will show just how much the strategic analyses being performed by your average business fall short. Both because it's logical and because it lets me write something amusing, we'll begin with our strategic ends.
We begin by discussing overarching goals and figure out what we should, as a state or a company, be trying to achieve if we want to attain our desiderata as a group. The first desideratum is almost always survival: if your company collapses or your state gets conquered, you're out of the game and stop being able to do strategy, so the first goal is always to ensure your survival. The neorealist school of international relations suggests that this is the only legitimate desideratum for a state, but in practice most groups of interest will have goals other than survival that can be ideological, religious, status-based or a myriad of other things. But survival's always in there.
Having determined or otherwise come to knowledge of the desiderata, we then look at our current situation as a state or whatever grouping of humans we're talking about. For a state, that might mean looking at where our borders are and how well they're secured. Are our neighbouring states allies or enemies? What's our financial situation like? What does the state of our military look like, and do we have the industry and expertise to change that? What do we need to do in order to maintain our survival and advance our other goals? What, in short, are our options for action that might advance those desiderata?
Having done an analysis of this kind, we can usually narrow down a wide range of hypothetical opportunities into a relatively smaller set of plausible actions that can be taken to get the state or the company closer to where it wants to be, or, if the organisation is happy with its current circumstances, maintain that happy state. The process for a country might look like something as follows, using my native Poland as an example.
We first ask "what does Poland want?" in the aggregate as a state. This could mean any number of things depending on the ideology of the people asking the question, of course: my personal takes are based on a broadly liberal socialist position. With that in mind, the first goal is the continued survival of Poland as an independent state, which is, for anyone who knows their history, consistently an issue. Secondary desiderata include maintaining the maximum possible sustainable standard of living for everyone in Poland, the protection and expansion of civil rights for all Polish citizens and the maintenance of our cultural practices.
This is enough to get on with the next step, which is looking at Poland's situation (literally, how it's situated). Poland is a populous country that's becoming increasingly wealthy: its GDP currently sits between Switzerland and Taiwan, so we're talking about a broadly developed economy here. It also has a strong manufacturing sector, inherited from the Warsaw Pact years, and is an EU and Schengen Zone member, benefiting considerably from access to EU markets and institutional support.
Geographically, Poland has historically had issues thanks to the lack of natural barriers on the borders, with the exception of the Carpathian mountains to the south. Fortunately, most of the borders of Poland these days are with friendly nations (Germany to the West, Czechia and Slovakia to the south), so while political developments in, say, Germany with the AfD may be an issue, they're unlikely to become a direct military threat. To the north there's the coastline with the Baltic sea, a relatively small and shallow see largely also controlled by friendly nations. The only exception is the eastern border, where Poland borders the Russian enclave of Kaliningrad and its puppet state of Belarus.
Russia is a historically hostile power that sees Poland as a natural part of its empire or sphere of influence and has shown a willingness to pursue those claims via force, as we see in the current war in Ukraine. As Russia has distinctly overextended itself there, the risk of a direct invasion through Belarus or Kaliningrad is limited. However, the collapse of the Ukrainian state and its falling under Russian influence, or a Russian invasion of one or another of the Baltic states would significantly worsen Poland's overall security and is a much more plausible threat. Moreover, with the US, a major Atlantic ally of Poland, becoming increasingly unreliable as a security partner, American weapons systems cannot be relied on, and neither can the US nuclear umbrella.
Poland's economy and military have been recently undergoing rapid modernisation, with much of the older Soviet weapons systems being phased out in favour of Western ones. Poland has been focusing on two priorities: building up a sufficient mass of modern armour and artillery for the land forces, and modernising the air forces. The air forces mostly rely on American airframes, with Poland having 20 F-35s currently in service, and up to 44 more on order, while the land forces are leaning heavily on domestic production and have secured production licenses from multiple countries, in particular South Korea.
All this being in place, we've reduced the question of "what ends should we pursue?" from an extremely vague one to one that's a lot more tractable. The major threat to the survival of the Polish state at present comes from Russia, with subsidiary risks coming from the weakening of the EU and the increasing unreliability of the USA. The ability of Poland to defend itself and maintain a good quality of life for its people might also be threatened by economic weakness and a decline in its heavy industry, which is a strategically important sector.
A set of actions that Poland could take to improve their strategic position, then, might look like the following:
- Support Ukraine and the Baltic states in their defence as much as possible in order to maximally weaken Russia.
- Modernise and expand the land and air forces as much as possible, making invasion prohibitive.
- Deprioritise naval investment, as the Scandinavian nations, Baltic states and Germany outpower Russia's notoriously weak Baltic fleet several times over. Keep investment to a minimum credible contribution to security in the Baltic.
- Invest heavily in domestic heavy industries and military production: this minimises the risk of an unreliable ally such as the USA withdrawing support at a critical moment.
- In order to maintain economic growth and a high quality of life for citizens of Poland, as well as maintaining the domestic economy, export markets for Polish-manufactured military equipment should be developed wherever possible.
- The primary international relationship of importance is with the EU: this needs to be maintained wherever possible. Atlantic relations with the US and the UK should also be kept friendly where possible, if only to keep the F-35s in the air.
- Some of these relationships, however, are likely to be unreliable. Looking for new allies and trade partners is thus of importance, particularly in the Asia-Pacific. South Korea, Taiwan, Japan, Australia and New Zealand might be excellent candidates.
These goals can be prioritised one way or another, but this is a broad universe of "things Poland could do to make the things they want to happen happen".
This is, in general, the kind of analysis of strategic ends that we should be looking for when a company talks about their "strategy". There are a couple of things worth noting. First of all, this kind of strategic analysis is specific to a particular country with a particular situation: this does not generalise easily to other countries, and in fact every country is going to have a strategic position that's specific to them. Companies will be similar. Secondly, this is pretty detailed and specific: while we're talking about broad-picture stuff here, even that broad-picture stuff is pretty granular, and this is just a blog post: doing it seriously would mean going into even more depth. Finally, the analysis takes considerable background knowledge and education: you need to know really quite a lot about history, geography, economics, culture and a whole lot of other things to be able to do an analysis of ends well. A serious scholar in the field, working on Poland, needs to have a working knowledge of at least four languages, for example, and even a relative amateur like me would find it difficult to work effectively without having at least a little German and Russian (I can read both languages with the aid of a dictionary, and even that slight capability has prevented me from making some pretty basic mistakes). In short, this isn't really something you can just jump into straight out of university with an MBA. This isn't just me saying it either: most high command in western militaries have advanced degrees and a lot of them have PhDs.
Measured against this metric, most of what passes for analysis of ends in the business world fails abysmally. The chief failure is vagueness: the vast majority of strategy documents and strategic plans that I've seen don't even go into the minimal depth that this post has, and even more seriously, they don't differentiate based on their organisation's actual situation. The end result is that most business ends analysis are some undifferentiated slop of "do less bad, do more good": a set of goals so vague as to be meaningless for actual decision-making. A large part of this is likely the fact that business culture has mostly internalised the Friedmanite dogma wherein the only valid end for a company is "maximise shareholder value", and when everyone has the same end, serious strategic thinking becomes impossible. The end result is that your average executive has no idea what they should be doing and in fact is largely oblivious to the idea that they could know what they're doing: their fallback when this happens seems to be mostly to do whatever will improve their wealth and status regardless of what it does to the company. I think the available evidence bears out this hypothesis.
Even if business leaders knew more about what analysis of ends should look like, moreover, many of them simply don't have the intellectual horsepower or knowledge to do it seriously. It's not at all uncommon for executives to jump from industry to industry leveraging their "management" expertise, for example: in such a situation, your average executive will know very little about a company's strategic position in a given industry, or really any of the particularities of a given company. That means that strategic analysis of any kind is almost definitionally impossible. Even more general knowledge of economics, the industry's regulatory environment or even such basic things as how economies react to changes in base interest rates are rampant among executives: while executives tend to be a bit more intelligent than the mean person, they learn very little and are very seldom rigorous or serious about anything. Is it any surprise that they tend to manage by muddling through and latching onto things that seem impressive?
Finally, we need to discuss what bearing this has on AI "strategy". When doing strategic analysis, we begin with ends: what do we want to get done, and what options do we have for actions that will do it? AI, being first and foremost a means rather than an end, doesn't have a huge amount of bearing on this, at least not in the sense that you would have an "AI strategy" in the sense that most executives mean it in. AI can come into the analysis of ends, but in the sense of a threat. For example, the widespread availability of cheap drones has the potential to make existing tactical systems in the military unworkable: you can see this in the Russo-Ukrainian war when the sheer level of visibility and battlefield awareness that drones allow make it very difficult to make advances or shift the front line and have led to the battlefield becoming largely static and positional in nature. An AI strategy at the level of ends should mean a strategy responding to AI: is there any way in which this technology poses a threat to our current position, and if so, what can we do to mitigate it? This might involve the use of AI, but it's impossible to say whether it does or not unless we've established the facts about what this technology is going to do in the first place. Companies are not, on the whole, actually doing this.
So, the state of analysis of ends in business is generally abysmal. Does business do any better at analysing means? Let's find out!
Means
What we mean by means when it comes to strategy is, in the wide sense, the set of capabilities we have, or the toolbox of things that we can do. This dictates, more or less, which of the strategic goals listed above we might be able to act on and how. It is, more or less, the analysis of the toolkit that we have for solving problems and making things happen. Tools and equipment are included, but so are things like education, institutional knowledge, public opinion and even morale.
Having established what strategic ends might be worth pursuing, we generally begin looking at what kinds of means might be able to bring those ends about. Different strategic ends will obviously rely on different needs to make them happen: to go back to our Polish example, we have a top-level goal of strengthening our army to make an attack by Russia prohibitively difficult. Assuming that that's a good end to pursue, there are multiple means by which we could pursue it. We could, for example, decide to focus on expanding the size of the military by recruiting more people: this might involve putting more funds into marketing the army and making joining a more attractive prospect by way of pay and other perks. We could alternatively implement conscription or selective service of some kind, compelling people to join the military by legal force. Alternatively, we could focus on improving the quality of our troops by way of more intensive and robust training, hoping to compete on troop quality. We might also elect to improve the quality of equipment our troops have, hoping that an advantage in metal will get better outcomes. In that case, you have to decide how and what: do we want to invest in our air force and our armoured troops, our navy, or do we want top-quality light infantry instead? How do we get the equipment: do we manufacture it ourselves or buy it? If so, from whom? Each of these decisions is going to have different overall impacts and different levels of effectiveness when it comes to meeting the strategic end that the means is being turned to.
In Poland's case, investing in land and air forces is obviously a better means for achieving the stated end than investing in the navy: the Russian navy is notoriously bad and the balance of power in the Baltic Sea in particular is overwhelmingly against Russia. Russian air and land forces, by contrast, are capable of being a real threat (even if they aren't really showing it at the moment). It makes sense, then, to expect a land incursion in force across one border or another, or an attack on an ally. Moreover, the kind of light infantry focused military developed by many nations that expect to be primarily supporting allies or projecting force overseas isn't appropriate here, so heavy investment in top-quality light infantry is not a great use of resources. Neither do we want a situation where we have a relatively small number of top-quality systems. What we want, rather, is a large mass of armour, artillery and aircraft that can effectively oppose a full-scale land offensive, with aircraft being chosen for intermediate-range strike capability and air superiority.
When it comes to the question of buying or making, Poland has elected to buy its aircraft but build much of its land equipment locally. This makes a considerable amount of sense: aircraft are extremely difficult to build, especially at the high-end (China, for example, has only been using domestically produced jet engines for a decade or two from memory), and while Poland has the skilled labour needed to learn, that takes time and effort that Poland may well not have. Buying aircraft can usually get you a better product than what you've made domestically and can also act to maintain and improve relations with important allies: another one of our strategic ends mentioned above. As mass is important, but so is quality, Poland has aimed for a high-low mix of top-end F-35s from the USA and the light strike variant of the KAI FA-50 from South Korea. The aim is for the FA-50s to take on lower-risk or less demanding missions, allowing the F-35s to be used where quality really matters. In the case of land warfare, by contrast, Poland easily has the institutional and process knowledge, workforce and industry to produce tanks and artillery in real volume. Moreover, production volumes in other places are often limited: Poland placed an order for 500 HIMARS systems from the US a few years ago that the US manufacturing facilities were simply unable to fulfil. Buy buying technology from South Korea along with production licenses and coupling that with an expansion of their already existing production lines, Poland is thus able to produce all the equipment they need locally and generate (hopefully) a meaningful export market in the process.
The point of going through all of this is that you have to have a really good understanding of the means available to you in quite a lot of depth in order to be able to make sound strategic decisions, and an awful lot of details need to be known and taken into consideration: I'm giving a version of all of this that's simplified to the point of uselessness, and this is still going into far more depth than most strategic documents that businesses produce. This isn't a hypothetical: states quite often make mistakes based on failure to correlate means and ends effectively and it often ends messily. Putin's invasion of Ukraine, for example, was in large part this kind of mistake. The kind of decapitation strike that they intended to launch on Ukraine requires highly effective light infantry, an air force capable of highly effective precision strikes, rapid insertion and suppression of enemy air defences, and a military culture trained for and accepting of a considerable level of command autonomy. Instead, Russia has tended to focus on its nuclear deterrent, rocket forces, and it's always dedicated a considerable amount of money to its internal forces for suppressing political dissent. Consequently, the capabilities that they'd need to actually pull that off were underinvested in, and now here we are. From what I've seen, your average executive is operating largely on Putin's level.
This comes up a lot with the software engineer's perennial gripe about managers not knowing anything about tech. I don't believe that software engineers genuinely think that managers need to know exactly how to write software in order manage it well: in fact, a lot of software engineers make downright terrible managers. What we're complaining about, I think, is that our leaders have an extremely limited grasp of software as a means. They don't understand what aims software can help you achieve or how, and they don't understand how software translates to actually achieving that aim. All they see is "squiggles on a screen that turn into status".
Nowhere is this more obvious than with our current discussion of LLM technology. What I've stressed above is that means are only really a thing in relation to ends: a means has to be, at least theoretically, a way in which someone could possibly achieve some kind of strategic goal. AI as currently constituted in the business world is simply too vague to be talked about as a means: it's everything and nothing all at once. Because of this, it seems like it can be a means for any strategic aim, but is in fact a means for none of them, and because it's far too vague as a term to say anything strategically useful about, we wind up with our current situation where companies spend immense amounts of money on AI, hoping for something, anything, that it'll be useful and revolutionary for. We're paying, in short, for a putative means which we're hoping will action some strategic end eventually.
Even when there is some obvious strategic end that an LLM-based technology could be a means for (for example, coding agents), the lack of understanding of how the means achieves the end means that things still don't quite work. The idea that coding agents can replace software engineers, for example, comes largely from the fact that managers don't understand how software solves problems, and thus often default to the idea that the chief factor is basically volume: you throw code at the problem until it goes away. Given that coding agents produce lots of code, then, they think that problems solved by coding are now solved by the LLM. However, this doesn't work, for much the same reason as there's no amount of infantry that will replace an F-35.
So, the analysis of ends is abysmal in the business world, so is the analysis of means, and both of these go double for stuff related to our current AI moment. Which leaves the correlation of the two, which your average corporate leader also sucks at.
Correlations
As mentioned earlier in this article, a popular definition of strategy is given as the "correlation of ends and means". Having discussed what strategic ends and strategic means respectively look like, you can probably guess by now what correlating the two looks like: given these ends and these means, which ends do we pursue, which do we prioritise and which means should we use to accomplish each end?
Correlating ends and means is an iterative process, and not one that goes "find your ends, then find the best means for each end". For instance, if you happen to have the means to achieve one of your strategic ends extremely easily (say you have an extremely highly educated population and you want to maintain an edge in scientific research as a result of it) it can be worth throwing the resources at that end even if the end isn't of primary importance to you. Similarly, there can be ends that you find quite important but that you don't have the ends to achieve except at extreme cost: in that case it might be worth setting that immediate end aside. New Zealand, for example, would have extreme difficulty enriching uranium in any real volume, so unless it was quite literally a case of "develop a nuclear weapon now or you die", it's probably not worth New Zealand investing a great deal in a nuclear program.
The same means can also aid in the achievement of multiple ends, and we have in fact touched on this already when discussing military procurement. All else being equal, it might be better or easier for a country to manufacture their own military equipment, but buying someone else's can, as mentioned, be good for reinforcing important alliances and otherwise engaging in diplomacy. Or similar diplomatic plays. Conversely, if you're interested in developing an export industry for economic reasons, putting the effort and resources into developing your own industry can be worth it even if buying the weapons would normally be easier.
And then there's built capability. It happens often enough that, while identifying strategic ends to pursue, you find yourself with an end that needs to be prioritised but that you don't have an appropriate means for. However, you have the capacity, given time and input, to develop the means that you need. In this case, strategy becomes a game of anticipating what means you'll need to have available and proactively developing so that when the end that you had anticipated earlier needs to be acted on, you have the necessary means to hand. Strategic analysis then becomes a question of thinking about, given your strategic situation, how your situation is likely to evolve going forward and what you should do to prepare for it.
If you don't have a strong understanding of either ends or means, this is obviously going to be a difficult or impossible process, and it's thus no surprise that a lot of executives struggle with it. Moreover, the whole process, as Clausewitz pointed out, isn't easily reducible to a set of algorithms: the space of all things that could happen is too great and many of the most important factors are vague to the extent that you can't even put a numeric value on them. Being good at strategic reasoning needs a lot of knowledge, not only tolerance of ambiguity but an ability to effectively game out ambiguous situations and not a little courage to make a decision when a decision is needed but not before. In short, this is all really hard and takes a lot of training, quite a lot of specialist knowledge, a deep understanding of the situation you're in and probably some natural aptitude as well.
Your average corporate leader, by contrast, did not get their position by going through training, exhibiting knowledge and understanding and demonstrating their natural aptitude for leadership. Rather, they probably got it by privilege getting them through the door in the first instance, followed by an ability to play office politics, fail upwards and take credit for successes, whether earned or not. Thinking too hard about whether what a company is doing is sensible or not is a disadvantage: they key is to not fail in a noticeable way while kissing ass. In short, when choosing corporate leaders, we're selecting for cowards who, while they aren't stupid as such, are very comfortable not thinking when thinking would lead them somewhere inconvenient. We then teach them that the only really valid end for a company to have is maximising shareholder value and shuffle them between companies and positions in completely unrelated sectors every few years. Anyone experiencing this is going to be, to use the technical term, completely ruined for any kind of serious strategic thinking. No wonder things are such a mess.
This is glaringly obvious when it comes to AI strategy. Now, by this point we have what we need to formulate what an AI strategy or at least the process of formulating one, if we were being serious, should look like:
- Examine the technology in depth and make, inasmuch as that's possible, an objective assessment of what the tool is or is not capable of.
- Establish whether the tool, either as presently constituted or in the near future, is likely to become a threat to any of the pre-existing strategic ends or desiderata that the company is pursuing.
- If it isn't, keep the technology under observation but don't necessarily act immediately.
- If the technology does pose an immediate threat in this way, mitigating the threat becomes a first order strategic aim
- The goal then becomes finding appropriate means to mitigate this threat effectively. One of those means might involve adopting the technology yourself: consider whether or not this will help, whether or not it's the best option and whether or not your company can, in fact, effectively adopt the technology: do we in fact have the built capability to effectively use this means?
- Only at this point do we consider how we might use this new means: where do we put it, where in our company does this belong, what oversight do we expect to be needed for it?
(There exists an alternative pathway in which the technology is an obviously applicable means for a pre-existing end, but I've not seen that happen yet).
Most executives are incapable of any of these steps. They don't know the industry or their situation well enough to formulate sensible strategic ends, nor do they understand what the means they have to hand are capable of. They generally lack the courage to prioritise effectively or to make and then own a decision in an uncertain situation, and thanks to the notoriously short tenures of most of our executives, they don't really have the ability or inclination to think about built capability or long-term strategy in a serious way: their thinking is limited to next quarter's results. And for the most part, they haven't been exposed to any serious intellectual thinking about strategy (the MBA emphatically does not count). Even a relatively smart person (which most executives are) is going to struggle to do any real strategic thinking in that situation. And, of course, a lot of executives are actually just dumb: they're a minority, to be sure, but they're very much there.
The end result is that, when faced with something like AI, they panic. They get the impression that the technology should be a threat regardless of whether it is or not and, unable to reason about an effective response, they default to status-preserving behaviour, which in this case is jumping on the bandwagon and hyping the thing up, regardless of actual evidence on the ground.
So, in general, executives can't analyse their ends, don't know about their means and can't meaningfully correlate the two. What the hell are they doing?
Executives aren't doing strategy, and they certainly aren't doing it about AI
At this point, one comes up against the objection that executives talk about strategy all the time and have a lot of different tools for doing it. To which I respond that talking a lot about strategy doesn't mean that you're doing it. There's a whole genre of Real-Time Strategy games that contain no strategy but happen entirely on the tactical layer, after all.
Rather, what I think executives are doing a lot of the time is a kind of performance of strategic thinking. They know that strategy's important: everyone's told them so, after all. They also know kind of what it looks like. They don't, however, have the skills and knowledge to actually essay a strategic analysis. So they pretend, and create something that has the form of strategy without any of the substance.
Just think about the SWOT charts so beloved of management consultants. It has the feel of strategy: we're thinking about strengths, weaknesses, opportunities and threats, after all! In reality, though, this is just another variant of the quadrant plots so beloved of the management consulting class, and in practice the contents of the things is vague to the point of uselessness. How are we judging strength and weakness? Relative to whom? How do we know that these opportunities are real and not simply phantoms? None of these questions ever seem to get answered or even thought about, and no real synthesis ever seems to happen. And yet this is what passes for strategic thinking in so many companies the world over.
And if the average executive isn't doing strategy generally, they certainly aren't doing it about AI. That's easy enough to deduce statistically: all the talk of AI adoption, AI strategy and all that is remarkably homogenous: people are all saying the same thing in roughly the same ways. Yet if AI adoption was considered strategy, we would expect different companies to be adopting it in radically different ways and many of them to not be adopting the tool at all, varying based on the radically different strategic situations that everyone finds themselves in. Our observed pattern therefore tells us that strategy around AI is not happening, in line with our earlier observation that business leader generally don't do strategy.
In the end, there's no getting around the fact that to really do strategy, you have to actually know things, and quite a lot of things. Sure, you might not need to know about the exact minutiae of how to do a job, but practically speaking, you absolutely need to know what role the job has to play, what means it's a part of and how that means correlates to an end. The people whom we select as executives these days simply don't know any of what they need to in order to do any kind of strategic reasoning: the education and formation that they go through covers precisely none of what they need, and there's really nothing in our culture that would impose the discipline of needing to learn on them.
So we wind up where we are now, drowning in documents that purport to be strategic when they are, in fact, meaningless slop. And while it's fashionable to be somewhat cynical about this or to say that maybe we should just shut up about strategy and get on with our jobs, I'd argue that this is, in fact, a real problem that needs to be treated seriously and with quite a bit of gravity. Without a clear strategic analysis, decision-makers often default to the strategic sins: emotive strategy, or doing what feels good to them, treating operational concerns as though they were strategy, or sticking to an old strategy when the conditions that make that strategy workable no longer exist and haven't in a very long while. These things can have disastrous effects on organisations and the people who rely on them, and in the extreme they can get people killed.
And modern AI technology very much does change the strategic landscape. Being in a situation where it becomes extremely cheap to create large volumes of test and code that looks somewhat useful is a problem for a lot of business models, exacerbated by the fact that LLM technology makes distinguishing quality that much more difficult: it goes a long way towards making everything a market for lemons. And the effects on politics of having massive volumes of LLM astroturfing bots running around cannot be overstated. This demands a strategic response.
The fact that, in a situation where careful thought is required, most business leaders are completely ill-equipped to think strategically or do strategy in their positions bodes ill for us. And the worst thing is that, when our leaders are comprehensively failing to meet the moment, absolutely not doing any kind of "AI strategy", and in fact not doing strategy at all, we can't avoid all of the stuff about "strategy" that they insist on producing. Jesus fucking Christ.
While Arca doesn't teach technical strategy, it does teach the knowledge that you need to be able to do strategy in the first place. We're in early access until the end of the year, so if you'd like to pick up a high-quality and engaging learning resource, you can do so below!