Do you really want an AI strategy?

Recently I was talking to a technology friend of mine who had been asked to come up with an AI strategy for his organisation.

I had a strong reaction to this – ‘any company asking for an AI strategy isn’t ready for an AI strategy’.

I explained that AI is such a broad topic that it’s meaningless to talk about it as a single thing. For instance, the strategic issues around generating marketing images are different to those around fraud detection processes and are different again to supplementing a call centre with robo-chat.

So after my friend told me off for making life harder, I offered them some more useful points.

1. AI is still not a technology problem - In most organisations AI should not be driven by Technology. There are a lot of useful roles the tech dept can be playing, but until it comes time to implement and operate it has to be the other areas of the business that are evaluating and driving adoption.

2. Good decision making is key – This is where organisations should be focused right now, and this where the tech dept can be a positive agent of change. Ask yourself – what can Technology do to foster the capability for AI decision making across the organisation.

3. Trifecta for good decision making: Knowledge, Risk and Governance.

4. Knowledge: look for AI training opportunities on the business side. It makes a lot more sense for a random person in the business to go and do an 'AI course' than for someone in tech; A tech person will have a hammer in search of a nail, a person from the business will have real problems that are worth solving.

5. Risk: A single 'risk appetite' for AI is meaningless. AI is too broad now, it can apply to 100 technologies and can in a dozen business domains, Organisations need to be able to articulate their AI risk appetite at a functional level.

For instance, can you answer these questions separately for HR, Finance, Marketing etc,
1. How tolerant are we of AI failure?
2. Are we happy follow others or are we prepared to invest in (and possibly waste) early effort?
3. Are we trying to save money (reduce headcount), reduce mistakes (automate) or improve quality (enhance what our people are already doing)?

2 and 1/2 years in from ChatGPT’s release and we are still riding the hype cycle. In most areas we have NOT seen the compelling general uses. Lots of claims, even lots of reasons to be positive, but it would be hard to build a clear business case for major expenditure in most areas.

I left my friend with this question: How will your company know when there is a compelling offering out there? This is what your strategy should be addressing.