The Reality on AI Transformation

Ethan Mollick's AI book is a brilliant must-read. If you read just one AI book, read his. Ditto some of his podcasts (for example, with Azeem). He is smarter and better prepared than me on paper and in the real world. But maybe not close enough to the teach-infused realities inside companies today.

The particular issue of testing a firm's AI 'seriousness' by suggesting you benchmark last week's upgrade of the Gemini model to adjust your plans is borderline nonsense.

As a general rule, it's not smart to not agree with Ethan, but with lots of admiration, here goes.

The reality for companies is that this is an interrelated nested set of big problems/challenges.

I. Layer One - Strategy Refresh

It is a strategy refresh moment that has to include the board level, a place where few board members understand the current change waves, including AI, and even fewer have any view over the horizon. Strategy work is hard and precarious, let alone during times of foundational shift—double that, given the speed of the change waves.

II. Layer Two - Tech Transformation

Inside of that, it is a tech transformation moment when, honestly, most IT shops are not great nor revered by their business. This is in a world where IT is typically not great at change management/program management, further exacerbated by a talent shortage. To make this even harder, every IT shop (and board) should review its security posture because one of the places where cool new tech is being adopted quickly/deeply is the Dark Web. The threat vectors are changing and worsening (now).

III. Layer Three - AI Transformation

Inside of that, most firms need a particularly hard and substantially different AI transformation plan, not just GenAI but a broad AI transformation for what is usually a hodgepodge of mismanaged earnest efforts. This must be value-anchored and future-proofed. It must now also include how you go about mining and farming tacit knowledge and adjusting processes to help that. If that doesn't make immediate sense, you need help seeing the near-term future of GenAI with RAG/CAG. Genius Alan Kay used to say context is worth 80 IQ points. Maybe it's more than 80 in this GenAI/AI-infused future.

In layers two and three, it is essential that this be co-led by IT-aware, enlightened folks from the CFO's organization (yes, that's hard, too). Finance needs to think strategically about cost structure, not the usual shorter-term default that is cost.

IV. Layer Four - Gen AI - The Rugby Model

All the GenAI stuff below the surface is pretty complicated. It is moving faster than any tech I have seen in a long time, maybe ever. It is so broad and easy to use that it is changing consumer tech expectations. There will be a growing demand and win for products, services, experiences, and brands that are intelligent (Toby's 10Ps) versus not intelligent. The second letter in AI stands for Intelligence...stew on that for a minute.

But understanding how LLMs really work and what they generate almost requires deep expertise in AI, Information Architecture, Phenomenology, Semantic/Symbolic Logic, statistics, and maybe Stochastic Calculus, so good luck with that.

While the emergent properties of GenAI/LLMs in all their forms are astounding, you really don't need to be able to explain how they 'think'. To start with, it doesn't think, and honestly, I'm not sure if anyone really has figured that out beyond; it does something that looks a lot like thinking on a foundation of having ingested trillions of elements.

But having said that, you absolutely must get your hands on this inside your company, use it, experiment, play, fail, learn, and then do more of that. Let your front-line ideate and create.

It is simple. This amazing, game-changing tech is like rugby.

While at some level it is simple to explain and the functional basics are quite straightforward (just like rugby), you don't become proficient at rugby by watching it and reading about it or by having some innovation group play rugby down in the basement somewhere.

You become great at rugby by spending time on the field and getting the bumps and bruises that come with that. This is not your typical plan; create a plan and monitor the plan for action vs. plan results as we do on so many things, with rewards for hitting plans and the opposite for not. That rubric, by the way, just teaches people how not to stretch and how to play safe. For sure, have a plan to get folks on the field and let them play/learn/build, create excitement and wrap safety around that, for your firm. Honestly, when was the last time someone in your firm got a reward or the spotlight in a positive way for trying something innovative and falling flat in their face?

This cultural change will be diametrically opposed to the implied culture at most firms. This will not work well without real work up at level one and a rethink of the culture (read that as the behaviors we purposefully propagate) needed for this new era.

But not dabble. Make sure you land real impacts, aim high. Take a look at what Klarna has done with customer service or the analysis BCG did or Takeda with process design or DIcks with marketing or what VANTIQ has delivered.

V. Layer Five - Magic

So, this is more difficult than it seems, and I know it looks truly difficult.

In times of massive tech change, including foundational leaps forward infiltrating business, science, and consumers' real lives, and as the buzzword-intensive media goes buzzword-intensive, it is hard to see both sides of what is a Creative-Destructive cycle.

Historically, it is easy to see those. British Steel/Empire, the textile industry, 80% of the vibrant auto manufacturers in 1912, Kodak, and GE are all examples of knowing there's a creative destruction cycle. However, from the 'board' on down, few seriously consider they could be on the destruction side, so Layer One gets short shrift. This is doubly so for the more successful firms.

So managing these five layers in an integrated, loosely coupled and semi-autonomous way (which is inherently contradictory) so it doesn't turn into a bureaucratic rabbit warren of a program is just so very hard. In parallel, doing the right work early up at the board to ensure what is being asked for, incentivized, and tracked leads to the needed new behaviors is a level of abstraction that almost never gets thought through clearly.

In this new world, innovation, including experimenting/failing/learning, must be rewarded and managed positively. That is a completely new behavior at many firms outside of some usually lame innovation team that no one really cares about. Aligning comms, rewards, and punishments drives behaviors. The sum of your behaviors at a firm is your culture. This requires a culture change, and that's how you do that. It is not changing the plaques on the wall and giving speeches about innovation.

For any new exploration/journey, you should take journey talent with you. This is a talent that has done that journey before. This is in the trenches as much as it is up top. Realizing you need that and finding it is pretty hard. This is not just tech talent. It is transformational leadership talent with a tech context.

Getting all of layer five right requires a level of magic.

Conclusion and Back to the Future

Unsurprisingly, during Industrial Revolution level change, there was a tremendous amount of destruction in the Creative Destruction Cycle.

So being afraid of ending up on the destruction side is pragmatically healthy.

In 1900, Baker sold one of the most popular lines of cars in America. Baker supplied the first fleet of cars to the Whitehouse. Baker made electric cars. The company offered a range of models, including runabouts, roadsters, Queen Victoria, Landaulet, Stanhope, Coupe, and Surrey styles. In 1900, 38% of the cars on the road in the US were electric— Yep, 38%!

Then the production line and the electric starter motor hit the market, not a GenAI scale change, but a pretty big change. Within 30 years (which is about 6 years in AI time) only 40 of the 400 profitable automotive companies still existed and of those three had 80% of the market. Very few CEOs thought they would be on the destruction side of the creative destruction cycle.

So, Ethan's thought that you should pause your GenAI plans to go evaluate and do internal benchmark work because Gemini did a new release is maybe looking at this from the wrong vantage point.

P.S. Toby's 10Ps: this is what intelligent actually is, with applied AI, for a company, you become: Proactive, Predictive, Pattern Matched, Preventative, Permissioned, Process Performant, Peer Connected, Personalized, Precise and Pragmatic; while your enemy does not. This is better outcomes, agility, and cost structure which, if you don't screw up the strategy level, is the winning combo.