Introduction
What passes for AI strategy these days has a very low bar that tells us little to nothing about the unique potential of AI.
We see this often from folks pivoting to become “AI experts” with little insights into the unique capabilities of AI.
My first AI design (patented ’96) was used in a commercial product (for improving capacity of Motorola cellular networks). I have used AI ever since (see example projects), so I have genuine insights into its unique commercial possibilities, many of which have only manifested in recent years.
What many posit as an AI strategy is often “Digital Transformation” with a bit of AI thrown in. This might be appropriate for tactical deployments, but insufficient for strategy.
Very few leaders have asked: How do I transform my business via the unique capabilities of AI? Or, how is corporate strategy fundamentally impacted by AI?
Indeed, there is a precursive question: How does the availability of “unlimited computation” impact corporate strategy? In many ways, AI provides the framework for answering that question.
Very few leaders are asking these questions because they don’t know how to interpret them within the context of corporate strategy. To assist with interpretation, I am currently documenting the unique principles of AI and how to deploy them as strategic tools.
I plan to share soon — to be honest, this is a kind of teaser post (and to explore if posting on LI is a useful medium).
As a precursor, let me share one example.
Recipe: AI Scaling & Innovation Networks
A recent discovery of AI is the so-called Scaling Law. It says that performance scales (as a power law) with the size of the dataset and AI model. Model complexity isn’t so important. Indeed, most of the “beyond human” capabilities of AI have come about from scaling.
With this in mind, a business might ask: How do we scale data as a strategic advantage?
Remarkably, many orgs do not ask this question.
They lack strategies specifically related to the accumulation of “data capital”. Data is still seen as a by-product of current processes versus an investment into future gains.
But there are some great examples from folks who really get it.
Recently, I encountered an alliance of farmers who understood that certain types of automation are inevitable in order to overcome common challenges, such as labor pressures.
They strategically identified the need for scalable data and so collaborated in the curation of a pooled repository of high-quality agricultural data.
Operationally, the alliance are organizing this effort via an “Innovation Network” in which resources and information are pooled and managed via a separate body.
The network taps into technology partners and key expert individuals who form what might be called “A Network of Invention”. This is a smart move because it leverages organic “network effects” from knowledge without the need to develop lots of in-house AI expertise, keeping in mind that such expertise is often hard to come by.
The alliance also works with an incubator to co-opt start-ups into the network. The data pool is planned to be a major input into the incubator program.
I have been involved with setting up several such networks, including for O2, the UK’s premier telco, via a loose collaboration of internal innovation teams (“labs”), strategic partners, corporate customers, incubators and informal “inventor” networks.
The farmer’s alliance doesn’t yet know for which applications the data will prove most valuable. But they realize that without the means to scale data they will have little chance of exploiting scaling laws. There is a prior expectation of utility from the data given the existence of such laws and related patterns, like Emergence.
In the alliance’s case, a key expectation from the data is reduced cycle time from AI experiments to in-field benefits, which is critical in agriculture where innovation cycle times are often overly long.
Summary
AI strategy must be formulated along the axes of AI’s unique attributes, such as Scaling Laws. This law alone gives rise to various strategic considerations, such as the accumulation of “data capital”.
Noting that scaling is hard, a possible strategic approach is the recipe:
Pooled data capital + innovation network
But there are many other recipes for accumulating “data capital” for strategic competitive advantage. More to come in later posts 🙂