Do you need a CAIO?

I got into a conversation with a few colleagues last month about the thorny issue of ‘AI Strategy’. Turns out that this is now a thing that organisations want. I kind of wonder why they want it if I’m honest, but the customer is always right and all. Anyhow, it got me thinking about who owns AI in a business. So let’s take a break from thinking about executive team roles and think it through. 

If you believe the hype then we are entering a phase of the economy in which automation radically changes the nature of work within a generation. AI experts will doubtless splutter at the idea that automation and AI are the same thing. I’m using the former in a more general sense. Artificial intelligence algorithms automate certain classes of decisions that hitherto required a subset of human biology to accomplish. By that definition AI tools are basically the same as other IT tools. An enterprise resource planning or warehouse management system automates administration tasks that a human being used to do in a ledger with a slide rule. It also automates the sending of instructions to other parts of the organisation about the actions that must be taken to fulfil the objective of a predefined business process. If we follow this logic then ownership of AI within a business should belong to the CIO, whose team will work with the business to make use of various AI platforms to enhance business processes.

The obvious challenge with this idea is that AI tools have the capacity to exercise a certain degree of judgement. In more advanced applications they have already demonstrated the ability to mimic elements of the flexibility exhibited by the human mind. Incidentally, that goes beyond answering calls/ whatsapps and driving cars, and includes quite complex tasks such as playing Dota 2. In this sphere, AIs have demonstrated the ability to ‘invent’ new strategies and tactics that break the typical human decision-making biases and can thus be very successful. The conceptual upshot of this new capability is that one could consider AIs in the same breath as human workers.

If we think a little deeper into the future of businesses we can stretch our thinking further. The paradigm in technology architecture is towards loosely-coupling functions using APIs. Some of those functions (again, broadly defined) will be proprietary to the organisation but the majority will be delivered by software as a service or function as a service providers. The only proprietary aspect is the data that the organisation applies those functions to. Competitive advantage thus comes from volume, scarcity and accessibility of data.

Thinking conceptually again, this means that I can ‘recruit’ AIs in much the same way that I ‘recruit’ human workers. I don’t need to license a software stack, integrate into my system and then maintain it and all of its (largely useless functions). Instead, I can hire an algorithm, feed it my data and make use of the results. As more-and-more aspects of the organisation enjoy decoupling of control software and the hardware that they relate to, this ‘hiring’ of AIs becomes ever more important.

Just to emphasise that point - the next generation of buildings, machines and vehicles will almost certainly be regarded by their designers as computers first, physical infrastructure second. You can see the trend spreading from Tesla to the traditional auto companies. If you add more sensors then you enable far greater potential for autonomy than the control algorithms could provide at the moment the hardware was created. I could conceivably hire different AI ‘drivers’ for different parts of a journey, for example. Or a different building manager for a busy day at the mall vs a sunny Sunday.

By this line of thought the key decision maker actually sits somewhere between the CIO and the CHRO (or equivalent).

There’s a further consideration. By some estimations the rise of AI is as seminal for businesses as the electrification of factories, or perhaps even of the steam revolution. By freeing humans from the limitations of their decision-making processes we could unlock a cascade of new innovations that will totally change our society. This makes AI a strategic issue that impacts the central value creating model of any organisation. Quite where this sits in the leadership team is open to question. Chief Strategy Officers are not a common feature on executive teams… and my view is that all executive officers should be strategic leaders, rather than letting someone else worry about the future. The CEO clearly needs an advisor who can help him or her understand and react to the predictable and unpredictable changes that AI throws up.

So do you need a CAIO? There is precedent for the creation of new executive roles from other recent perceived paradigm shifts. For example, when people realised that digital computers were here to stay (sometime in the early 2010s) it suddenly became fashionable to talk about ‘digital’ in all aspects of business. Some firms created Chief Digital Officers to represent how forward thinking they were at leadership level. For the most part, these people were and are reformed marketers (digital is just a channel, of course) or CIOs (IT is more than Windows!). With some notable exceptions, these CDOs have just continued the slow evolution that was already happening in their predecessor silos. Not a bad thing, but not actually very strategic.

The challenge with creating a new executive position is that it tends to make other members of the ExCo ignore that issue as something for their colleague to worry about. AI is like digital computing in that it impacts all aspects of the business. It differs because it enables the optimisation of resource allocation in very large scale systems in a way that was impossible with humans in the loop. That means that although a good IT system can support the business model of a traditional business, a good AI ‘system’ could well lead to it be picked apart.

If a new CAIO position is created then the chances are it will come from the HR or IT siloes (presuming that businesses are not ready for wholesale change). That will lead to the same failure that happened with CDOs (and CCOs before that) - the topic will be addressed narrowly when it needs to be considered as a systemic opportunity for transformation. So my view is that you do not need a CAIO, what you need is strategic leaders in ‘human resources’ (hate that term!) and IT who are able to engage each other and the rest of the business in an informed debate about the broad spectrum impacts of AI.

I expect the first FTSE 250 CAIO to be announced in the next 12 months.