Most boardrooms are discussing AI strategy now, and they are as clueless as a penguin trying to fly.

Every few years an unknown phenomenon shows up in the world and poses a challenge to well laid out plans. COVID was that phenomenon three years ago. Last year it looked like interest rates and recession may be the chatter in the boardrooms but AI climbed into a higher pecking order of boardroom discussions. Why ? Because, jaw dropping demos of products powered by AI are dropping into the world every week. Handful of these products are finding adoption at a pace that history has never witnessed before. 100 million users in two months! The fastest technology adoption prior to that took about two years. Old technology like the telephones took seventy five years to get to 100 million users.

A Mega platform shift is underway due to AI

In the last 7 months, 7 years worth of progress has happened in the world of artificial intelligence. The idea that machines can learn and there can be artificial intelligence that can mirror humans germinated in the 1950s. However, for the last several decades including the much heralded breakthrough of deep learning in 2012 it felt like the progress was at a glacial pace. It was like tinkerers were building home appliances, but they used a hacked up battery. During all this while, only the battery design evolved. Suddenly, it feels like electric power is available on the grid at the touch of a switch. This means that the kind of products that can be created, consumed, purchased and distributed changes radically. In building software, the source of power is moving from batteries to the electric grid through AI.

Such a significant change in the dominant technology or infrastructure that underlies an ecosystem is what is called a platform shift.

‘Software is eating the world’ was the call at arms for many builders. This is getting replaced by ‘AI is eating the world’. In large enterprises, Digital Transformation is starting to get replaced with AI Transformation projects. Lot of this is due to the fundamental shifts that have happened in the technical architecture of AI building blocks called large language models, its super set is called foundation models. The underlying transformer architecture as it is called has scaled beyond language translation into what is being heralded as Generative AI. Text to text, text to image, text to speech, video, text to action, text to code, text to app.

Platform shifts are a bit like a scene from the movie Ice age 4. There is chaos, every player big or small is running helter skelter. In some platform shifts, incumbents came out on top, while in others, startups won. Startups had a significant role in driving the growth of the web and internet, as well as in leading the way in social media. However, incumbents had greater success in the mobile platform shift. Similarly, startups took the lead in Web 3. In AI, building large platforms and products will require substantial initial investment, giving incumbents an advantage.

This does not mean that you should shy away from AI. It means that you pick something where incumbents don’t have an advantage in AI.

Demand shocks are following, upsetting ranks

New technology does not inherently make it interesting for business. While some may appear cool at first, they may never take off. History is rife with a myriad of such examples. It is the behaviour change that is triggered when new technologies come to life that makes them interesting. Multi tenant architecture and ASP i.e application service provider were around for a long time but it was after the emergence of Google Search that led to the rise of assisted buying, that the SaaS industry took off.

100 million sign ups for ChatGPT is unlike anything we have seen before. Some follow on effect of this has been that lesser students are using sites like Chegg for doing their homework. ChatGPT seems to have taken away a lot of that mindshare. When Chegg’s CEO shared this in his Q1 2023 earnings calls, the stock dropped value by 40%. Similarly, developers have been using StackOverflow much less and less. In the same quarter the traffic for StackOverflow went down by 14%

If you don’t act now, you will be wiped like Yahoo, Nokia

If you don’t make shifts during platform changes, you may follow in the steps of Jerry Yang who missed all the shifts at Yahoo. Snatched a loss despite victory being handed over to him at least half a dozen times.

Or worse, go down the Nokia path. Best case study in human history on how to destroy $100b in value. Yahoo and Nokia are canonical examples of those who did not keep up with platform shifts.

Strategy means deliberate choice on how to win

In high school mathematics, they could have used the word relationships instead of algebra. Calling it linear relationships would have not made it sound math like. Therefore they had to call it linear algebra and terrify a lot of people. Make them abandon learning it. Word strategy is like that. They could have called it ‘choices’, ‘deliberate choices’. But that would have not sounded business-y for which people can charge a lot of money. That’s why they picked a fancy sounding word- Strategy. To simplify, deliberate choices you make in a business to win is called strategy.

Most do strategy by imitation, they call it following playbooks. They look at someone who has been successful and figure that if they could retrace the exact steps others followed then they would become successful too. They miss the context. When you ignore the context then it is like playing blind chess. Choices you make should take your surroundings into account. Usually big enterprises are dealing with complexity and volatility. Where startups play, especially when platform shifts happen the environment is uncertain. Startups make choices under uncertainty and ambiguity. One other reason startups don’t like the word strategy is because the strategy (choice making framework) books that are written are primarily for those dealing with the surrounding of complexity. Most choice making frameworks break down under uncertainty. One of the key lessons to keep in mind is that when situations change, choices change. When situations change, even the choice making framework changes. So playbooks are almost useless.

Another common approach to strategy is to make grandiose sounding statements. Something like we are an ‘AI first’ company now.

‘AI first’ is sloganeering of a fake CEO.

It’s someone who wants to appear knowledgeable without actually understanding what’s happening. Sloganeering may look cute, but in areas of extreme uncertainty, it can get you killed.

Don’t think like that. If anything, the mobile platform shift has shown us that ‘mobile first’ world does not mean a winning ‘mobile first’ strategy. Sure, the mobile first world meant applications like Uber, Pinterest emerged, but no enterprise won over the competition because they had a ‘global, mobile, social’ strategy. Mobile first world benefitted the platform companies that won Google and Apple. Not enterprises like Nokia or Blackberry. It is sloganeering like these that give strategy a bad name.

Anyone who has worked in a startup knows that every few months the North Star has to be defined again. However, by definition a North Star stands for an unchanging core set of beliefs. This creates a cognitive dissonance. The North Star must change when an earthquake happens. The trouble is that the North Star is a leaky metaphor. Take Open AI for instance, it started as a research company and non profit. Now it is staring in its mirror, a consumer company, a platform company and a research company. Due to this it has to shake off its shackle of a non profit. Many including Elon Musk are miffed at this because their narrative of the North Star is breaking. Direction should not be mistaken for destination. Those who anchor their identity too close to such language will find it hard to shake it off. This also becomes a big problem when category names get redefined. If the mission statement is to be world’s #1 CRM, then it is hard to make sense of what to say when the world of CRM does not matter.

It is funny how this works. To align a team, you need slogans. But if you don’t think deeply about the strategy and do blind sloganeering, it can take you in the wrong direction. Strategy is a must before cute language and sloganeering.

So then how do you do strategy? That is, make deliberate choices. Especially when everything around you is changing.

Like Boyd, the fighter pilot said, you observe what is going on, you adjust your shot thought, announce the trigger, and finally you take shot action. Do this and you can win against well resourced players. Boyd called this the OODA loop- Observe, Orient, Decide and Act. When you are a single player, you can go through this loop real quick in your head. When you lead an organisation, to get everyone to change their orientation would require giving them all an aligned map. Biggest trouble when working with a large team is that if you don’t give the same map to everyone, it will create an internal crisis. And changing the orientation becomes crazy hard.

Simon Wardley, an engineer who became CEO for Ubuntu had spent the more than15 years recognising fakery and imposter syndrome that CEOs and board members go through on the topic of strategy. Simon had experienced this himself and overcame that feeling. Then he took it upon himself to help anyone who is suffering from strategy faking syndrome. To make choices in a way that they feel certain about it. He picked ideas from John Boyd and well studied strategists like Sun Tzu; he devised a simple thinking tool, to discuss and decide these choices. This visual thinking tool is called a Wardley map.

In his own case using the visual tool had helped him make some deliberate choices that took Ubuntu from 3% market share to more than 70% market share in 18 months. Read his book published on medium on how to do this mapping. For a quick intro watch this talk

Use a visual tool like this to decide your “how” and thereby define your own playbook. When you build a visual map, you and your team can have an aligned conversation.

Every map needs an anchor point like the north direction in a physical location map. For business maps one of the best anchors is customers. Next define an X axis and Y axis for the map.

For the Y axis with the customer as the anchor, list the important jobs to be done and break it down to components that make it up. In the above example if you want to make a cup of tea, you need tea leaf, hot water, which itself needs water and a mechanism for heating. This elaboration is also called the value chain.

For the X axis follow how the evolution of tools (& its components) used for the job to be done has changed over a period of time. An example of X axis is how compute evolved from being like Magic from Charles Babbage days to being Custom built in the days of the mainframe in the early 1920s. It then moved to Product in the 1980s when Bill Gates decided to put a PC in every home. Finally with the invention of virtual machines in 2009, computers moved to the cloud and were available as a Commodity. Available at the fetch of an API. This change happened first over 100s of years and then over decades and eventually over a few years. When such phase transition happens, industries shake out. Winners and losers trade places.

Watch the evolution flow of tools

Theory of disruption is the study of how incumbents failed to watch the evolution of tools, its components, the respective jobs to be done. Clay Christensen’s insight was that if you don’t pay attention to how new tools are replacing existing jobs to be done then you get uprooted. Nokia board members were so confident about their manufacturing ruggedness at the time of the iPhone’s release, that they laughed it off completely. Every Nokia phone was tested by running a road roller over it. They were confident that a touch screen based phone will never pass a rugged test like that. The iPhone proved that ruggedness is not what customers care the most about. Andy Grove did not make that mistake when shifting from the microprocessor to the memory business at Intel.

AI started 70 years ago, in 1950 but in the last few months it moved from custom to product and arguably commodity stage. This shift happened recently. What virtual machine (VM) was in 2009, foundation models (FM) is in 2023. That is what changes a lot of things. In the face of commoditization all those who made custom designs will become useless. Except in special needs most appliances will run on electricity consumed from the grid. No one buys or sells washing machines that are powered by a battery. Every enterprise must draw their own wardley map and ask what has changed for them. What were they building (expensively) with batteries and now with cheap electricity available what should they redesign.

Let’s draw a generic example and see what the changes are for an enterprise.

The killer app in AI for enterprise that has garnered the most attention is a co-pilot.

If you draw a map for the copilot and its job to be done then it may look something like this.

The key job to be done in this is 100X Enterprise Productivity and a Unified Customer View.

As this map is sketched out some of the following shifts can be noticed

• Natural language is the user interface for end users to all software
• Natural language is the programming language for all software
• No need for data models or structured data to build an application
• Unit of compute is a neural network model, not a von neumann machine
• Unit of storage moving away from symbol storage to vector storage

Lot of things are changing at the same time across the entire value stack.

Based on your current position & movement, change your choice

For those that are asking whether you should build an LLMs and what should be your generative AI strategy, a far better question to ask is where should you play in the above map ?

Should you play here at the top closest to the services. Something like what Turing is doing

Or much lower in the stack, where the incumbents who can make investments in the computer and storage primitives have advantage. For example, there will be a rise of startups in vector databases. They will face stiff competition from incumbents. Technical architecture, developer adoption, and eventually a robust business model will be key in winning.

While there are gaps on the infra side and new categories will emerge. There is also the rise of new middleware that will plug the new interfaces and the old software systems. Is that where you should play ?

What this platform shift enables is powerful software systems that end users will be able to query through natural language. Many of the new systems will be built using a language closer to a human language. However, the AI systems may not provide a certain kind of reliability, coverage or privacy. Those problems still have to be solved. Is that what you should solve ?

Every Services company will be an AI company, with better margins. What will be grossly misunderstood though is the difference between AI services and AI products. Any founder that has built SaaS in India knows the pain of dealing with services. It kills the margins in the business. Build once, sell multiple times nature of the product business is what is more desired. Moreover scaling a services company means increasing headcount therefore services business, while guarantees revenue, does not provide margins and hence gets lesser valuation.

As the cost of building an application will go down, with AI copilot a lesser number of engineers will be needed to build sophisticated applications. Margins will improve and scaling of the services business would necessarily not be correlated to headcount. It is quite possible that a Infosys 3.0 will likely emerge. A big services company like Accenture employs about 100,000 to generate a billion dollars in revenue. With AI it may generate the same revenue with 20,000 or even 2000 people on their staff.

You cannot intellectualize your way through this

Some management experts believe that in strategy ‘thinking is most supreme’ and all problems can be solved through superior thinking. Initial prudent thinking and deliberate choices gives a good starting point but not the full answers. In startups you can only see what one iteration allows you to see. You may know where you want to reach, but the road is only visible to the extent that the headlight of your car shines on.

Build a skunk team within your organization. With full autonomy and no burden of history and alignment needed to existing corporate strategy, let them reimagine what solution can be built that is 100X better. Let them ask what has changed, what will. Billion dollar sized companies follow the two pizza rule for their skunk teams, yours should not be any bigger. Giving them the autonomy and initial map to bring back the lessons and that will get the strategy that actually works. Speed of the lessons matters most. Boyd’s core teaching of OODA loop is that.

Remove Sundar, bring in Satya. Pun intended

Coming up with a better strategy is one part, executing is a different ball game. Especially when they are radical.

In his book Hard things about hard things Ben Horowitz talks about the difference between peacetime CEO and war time CEO. During platform shifts you must bring war time leaders and phase out the peacetime leader. You must remove the Sundar (beauty) in your org and bring Satya (truth) in the organization. Pun is intended here. Both Google and Microsoft have Indian CEOs and that is where the comparison ends. Sundar is a peace time CEO, Satya is a war time CEO. Google did the research that created ‘transformer’ but it is startups (OpenAI) from outside that built the killer app for AI i.e ChatGPT. Google was waiting and watching, still is. Microsoft has spent more dollars in research in the last 20 years through Microsoft Research but that did not stop Satya from partnering with OpenAI to bring GPT to market.

In uncertain times, when platform shifts are happening you need driven folks who can make swift and tough decisions.

If decisive actions are not taken on deliberate choices then it becomes merely an academic exercise.

Draw your map together, but hand reins to a wartime leader

Remember, your strategy will depend on your context.

First staff a two pizza team of best entrepreneurial tech, marketing & sales and design folks from within your enterprise.

Sit them along with your management team and list down critical questions to ask. After you have listed 5–7 questions, ask what is the most important question to ask.

Ask:

“What is changing for you and your customers?”

“What has moved from custom to product or commodity? “

“What are your competitors doing? “

Another great place to anchor choices is to focus on:

“What is not changing ?”

Human motivations don’t change, everyone wants a promotion, everyone wants to grow. Humans don’t like to feel stupid in front of colleagues, they like it when egos are massaged. Behaviours also change rarely, a certain way of doing a workflow is less likely to change unless and until a 10X improvement comes up in front of them.

Put all of this in the visual wardley map and come up with a first draft of what you should do. Where should you play and how that gives you an edge over others that are playing.

Let the skunks team execute and bring back weekly lessons.

Ensure that a strong leader is given the reins of the organisation that he or she can make tough decisions and drive these changes.

As Howard Schultz, former CEO of Starbucks says “In times of war, you need a war time CEO. It’s about making tough decisions, mobilising resources, and leading with clarity and purpose.”

PS 1 — Do you need this complex sounding tool called wardley map to come up with your strategy? Probably not! Many first rate thinkers, for example Peter Thiel are able to think strategy in their head without a visual map.

My experience of having this conversation with over 100 startups of all sizes has taught me that when you have a large group of people who need to align, a visual map makes the process of alignment extremely effective. Also, a visual map is the most useful when aligning co-founders who are on different tangents. As is this case with any tool, use it where it is relevant. Once you reach your outcome, move beyond the tool. Recognising the change and making your deliberate choice is more important. Don’t get stuck in the tool.

PS 2 — This blog post is based on over 50+ conversations with all kinds of audience — investors, public company board members and various sizes of B2B startups. If you would like to access the powerpoint used, here is the link https://bit.ly/IndianB2BAIStrategy

published initially at https://medium.com/@mtrajan/every-enterprise-needs-an-ai-strategy-how-to-build-yours-f807b7e79b48