Why you need a Chief Artificial Intelligence Officer… for now
AI deployment in the corporate world is at a critical inflexion point, often stuck at the proof of concept (POC) stage. The need for translation between data science teams and business stakeholders means that corporates should consider appointing a Chief AI Officer (CAIO) to set strategy, support technology choices, and drive roll-out. As machine learning becomes business as usual, the ambition should be to make them redundant. But until you are there, a CAIO may be just the shock to the system that your business needs.
A familiar scenario
The Chief Executive has attended the conference, her senior team have had the briefings. The Board have asked the questions. Multiple startups have been invited in for coffee. The pilot projects are fascinating – if inconclusive. Chatbots have proliferated. Every team has some version of ‘How AI will transform the business’ written up for sales or strategy purposes. But somehow, projects remain stuck at the POC stage. Mobilising the IT and operational resources to deliver on wildly promising but slightly woolly business cases is proving complex. The ROI on the expensive data scientists and strategising workshops remains limited. Sound familiar?
Getting AI live in the organisation is hard work. The challenge is two-fold. Firstly, expectations about the technology probably need resetting. That self driving car is still half a decade away and chatbots remain frustratingly easy to ‘break’. There is a hype blip to be lived through – but the opportunity is real. At Best Practice AI we have captured over 600 use cases for AI, many of which are implementable tomorrow, and over 1,200 case studies where this has happened in reality.
The second critical challenge is that getting machine learning (the main driver of current AI) into operational production is hard. Getting to proof of concept is complex and costly: accessing the right combination of data science and domain expertise skills, identifying and optimising appropriate data sets and then managing the compute costs and technology is not a simple task. But it is often simpler and cheaper than the next stage of embedding the machine learning output in the business and wider organisation.
Calling on the CAIO
Overcoming these hurdles will get easier over time — each iteration reduces the challenges for the next one, and scaled roll-out will ultimately make this simply one tool amongst many. But the situation at the moment is analogous to the early days of the world wide web. To make change happen, many organisations turned to a Chief Digital Officer. Over time this became redundant as digital became something that every process had baked in. But, initially, they had a critical role: part prophet, part teacher, part strategist, part operator and part venture capitalist. To make AI work your organisation may need the modern equivalent: the Chief AI Officer. Broadly, CAIOs will be responsible for five things.
1. To set a vision of where the organisation should be going. This will involve explaining AI to stakeholders across the organisation but also working with them to formulate a strategy.
2: To set standards, ensuring that common tools are used and that ethical issues are thought through. Part of this will involve setting up the right governance structure to make sure that there is a plan to handle problematic AI issues.
2. To build capacity, recruiting and retaining the right data science talent, ensuring that the data engineering capacity exists and finding the right vendors. This may be about brokering conversations between specific use case vendors, e.g., natural language processing (NLP) vendors and the legal team, supporting the IT team in cloud computing vendor negotiations or finding the right consultants to solve key problems.
3. To make things happen, operating at the right level with the right stakeholder buy-in to convene key people, manage the project office and, where necessary, kick the doors down to ensure delivery. This will be an unashamedly political role and will need superlative stakeholder management skills, especially as the new technology’s reputation as a potential job-killer will need handling sensitively.
4. To be the public face of the new technology internally and externally, working across the industry, networking in the right places, and finding the right academics to consult if necessary. And, let’s be honest, if something goes wrong, they will be the one to take the blame.
Creating a home for the CAIO
The CAIO might not be at the Executive Committee level, but beware the various other departments reaching out to own the role. AI often gets its initial traction through innovation teams – but is then stymied in the transition to broader business ownership. The IT function has many of the requisite technological skills but often struggles to make broader business cases or to deliver on change management. The data team would be a good home for the CAIO, but only if they are operating at the ExCom level: a strong management information (MI) function is a world away from a full AI strategy. Key functions may be strong users of AI – digital marketing teams or customer service teams with chatbots, for example – but they will always be optimising on specific things.
So, who will make a good CAIO? This is a hard role to fill — balancing data science and technology skills with broader business change management experience is a fine line. Ultimately it will be circumstances that dictate where the balance should be struck. Factors include the broader team mix and the budget available, but above all the nature of the key questions that the business faces.
The CAIO has one final task: to make themselves redundant. Their objectives and incentives should reflect this. This should not be a problem for them. If they get this right then there will be plenty of other opportunities elsewhere. And, if they get it really right and they transform your business, then they might just be looking at the top job.
For more insights from our guest contributors, sign up to our free weekly newsletter.