Framework for an AI strategy

AI is not IT.

In the case of IT – a computer is programmed (defined a set of rules / logic) and fed data to achieve an outcome. In the case of AI, the computer is fed data and the intended outcome and it creates the rules to achieve the outcomes. IT is deterministic, AI is probabilistic. Given this fundamental difference every aspect of dealing with AI from strategy to adoption needs to be looked at with a fresh perspective.

Here is a framework for an AI strategy...

Establish purpose:

• What leapfrog changes will AI bring to the industry and organisation in question?
• Humans and AI – what roles / activity does the organisation expect each of them to play in their business of the future?

Answering these questions will help organisations with three important insights

1. If or by when should they adopt AI in their business
2. What aspects of their business will be AI powered / led and 
3. Importantly, establish if AI is supplementary or complementary to humans in their future mode of operations

This clarity is essential to developing the purpose and the Constitution for AI in the organisation. The constitution for AI is a document that sets out the philosophy and principles of leveraging AI by the organisation for itself and its customers. Organisations should use the constitution as the guiding light to develop policies (laws) on privacy, security, access, applications, terms of use etc.

Create platform:

Data is the oil (fuel) for AI engine. Sustainable and continuous supply of data is key for successful leverage of AI.

This means eliminating data siloes within the business, having a clear forward looking view on organic (capture) and in-organic (buy) data sources, setting up humming processes for data engineering (cleansing and making available real time), creating a definitive view on how the data will be used and monetised etc. In short a data strategy.

It does not matter if vehicle [organisation] has a powerful V8 engine [ML Model], it won’t move an inch without clean fuel [data]

Talking GPUs won’t cut it! Training, Inferencing, Data processing (key dimensions of an AI project) at scale demands that organisations reimagine the underlying hardware infrastructure for aspects of business that will be driven or led by AI. Already there are companies making IPUs and DLUs (Deep Learning Unit) which are specialised hardware for AI. This means having an AI & IT hybrid infrastructure strategy.

Organisations that are referring to Cloud and On-prem as Hybrid infrastructure, have some catching up to do…

Hybrid infrastructure now means Cloud and AI infrastructure!

MLOps – This is the DevOps equivalent in the AI world. MLOps is the de facto way Data Scientists and production teams collaborate. Organisations must ensure they include ways of working and collaboration approaches especially in the context of remote working as a part of the strategy rather than as afterthought.

AI talent will never be available at scale in one geographic location!

Once the purpose and platform are established, the final frontier is co-opting people.

Co-opt people:

As Peter Drucker said, “Culture eats strategy for breakfast”. Culture is made by the people – employees, customers, the eco-system and stakeholders. The trust paradigm has shifted. People trust technology more than the society (individuals in the society) or institutions.

If organisations decide to shift their business to be driven by AI, Explainable AI is an imperative.

Enterprises need to ensure they build Explainable AI systems so they can ensure traceability, transparency and hence build trust in the systems that people will come to depend upon.

Explainable AI the opposite of Black box AI. Explainable AI systems can show (visually) how their inference engine created the logic / set of rules to achieve the outcome. Employees (developers and sales personnel) will transfer the trust to customers who will cascade it further!

The key to wide spread adoption of AI will be Explainable AI systems.

Organisations can save hundreds of thousands of $$ if not millions by investing in a strong AI strategy first instead of hurrying up on pilots and PoCs.

Starting with a pilot / PoC for AI is a sure way of failing.

An AI strategy will enable organisations (1) understand if, when, where and why they can leverage AI, (2) create the necessary infrastructure in terms of data, hardware and processes and (3) create a cultural environment where humans look forward to AI adoption instead of fighting it!

Let the conversations begin! :-)