WANTED: Chief AI Officer

This article was originally published on Medium.

Chief AI Officer (CAIO) is the hot new job title: it shows up on LinkedIn, both on people’s profile and on job postings; on-line courses and certifications are being offered for it (see https://www.chiefaiofficer.com/); the US Justice Department recently appointed one (see The Justice Department gets a chief AI officer — The Verge); and today the US Vice President announced "they have directed all federal agencies to designate a chief AI officer".

CAIO, why now?

AI has been around for a very long time: the ELIZA chatbot (ChatGPT's grand-grand grandmother) was introduced at MIT in 1966. So why did we have to wait more than 50 years for the job title to emerge?

What makes things different today is this new breed of AI — Generative AI (GenAI for short), embodied by ChatGPT launched in November 2022 — that looks like a magical tool that keeps getting better and can be used by anyone for good and evil with new use cases being invented every day.

• GenAI is accessible. Anyone with computer access and an approximate mastery of a human language can enjoy it.
• GenAI is transformative. I would draw analogies with transformative technologies (e.g. nuclear energy, electric vehicles, content streaming or cloud computing) that brought new value creation in exchange for serious risk/benefit trade-offs.
• GenAI is fast-paced and appears limitless.
• GenAI is already being used for good and evil.
• The science underpinning GenAI is hard to understand when not totally unexplainable.
• GenAI is a one-way door decision. Once you have opened Pandora's box, it is impossible to close it.
• GenAI is seen by some as the doom of humanity.

"War is too serious a matter to entrust to military men." said Georges Clémenceau. If the same is true for AI, we need someone AI can be entrusted to. And that would be the Chief AI Officer.

Chief AI Officer, as the AI champion

The term champion implies the notion of fight in the context of a cause. The cause is to leverage AI to make the company more successful. The fight is against the status quo (e.g. myths about AI, innovator's dilemma, etc.)

As the AI champion, the CAIO relentlessly advocates for the use of AI inside the company. They are the early adopter and lead by example, often being the first one to use AI for their own job. They are a catalyst encouraging transformation using AI and providing air cover when needed. They also foster a culture of continuous experiments.

Chief AI Officer, as the AI guru

From its etymology, the term guru implies expertise, wisdom and teaching.

As the AI guru, the CAIO stays abreast of all the developments in the space (new companies, new products, new advances): they are simply the best informed person inside the company in terms of AI. They use their experience, common sense and story-telling skills to guide people in the right direction. They educate people and evangelize new ways of doing things. Ideally, rather than forcing people, they nudge them towards a meaningful AI transformati

Chief AI Officer, as the AI officer

As a member of the C-suite we should expect the CAIO to embody attributes such as strategic vision, leadership & influence, decision-making & risk management, communication & relationship building, and adaptability & resilience. The CAIO should be the AI-whisperer for the company's CEO.

A critical part of the job is to align the risks and opportunities of AI with the company’s business goals including revenue, market share, customer satisfaction, brand, risk management, employee happiness and retention, operational efficiency, innovation, etc.

CAIO, what profile?

Nobody really knows yet what the job of Chief AI Officer entails or what the ideal candidate should look like. See here for my exchange with ChatGPT on this topic.

Here is my personal check-list for a strong CAIO candidate

1. an innovator, looking at the world as it could be and wondering why not
2. a domain expert in the AI space with the right intuition and the capacity of going deeper when needed
3. a domain expert for the domain your company operates in
4. a curious tinkerer with a hands-on attitude
5. a story teller
6. a team player

I would go for someone with a MSc or PhD (preferred) in a quantitative field, with both research lab and corporate experiences; someone who still enjoys getting their hands dirty playing with AI technologies; someone who has successfully led projects with a sizable and cross-functional team. I would be biased towards a senior candidate with a successful technical-product-manager or product-engineer career and some leadership experience.

Closing thoughts

You need a champion to look at risks and opportunities related to AI. This person urgently needs unlimited access to most if not all teams, with their roadmaps, pain points, aspirations, etc.

In lots of organizations, creating a new C-level position (and the team that goes with it) is probably the best solution if you can afford it. But you should expect some friction and maybe jealousy from other C-suite members or teams (who does not want to be in charge of AI?). Can your existing Chief Technology Officer assume this role? Maybe. But at the risk of not getting their full undivided attention.

Also keep in mind that the CAIO role may be temporary as some of the uncertainties around AI may disappear and their responsibilities may get transitioned to other teams.

In any case, having an AI champion to help the CEO and the entire organization navigate the current foggy AI skies is critical. The last thing you want is to fly blind.

Feedback is welcome.

Appendix A: AI / business goals alignment

Here are a few examples of such alignment between AI and business goals, with a focus Generative AI large language models (LLMs), for a hypothetical company in the business of providing data and analytics.

The good

Building a conversational interface powered by an LLM on top of your data product will grow revenue (you can sell your product with a premium), expand your marketshare (your product becomes more appealing), increase your customer satisfaction (who doesn’t like asking questions in plain english). And from a brand point of view, this will send the message that you are a true AI company.

Embracing tools like Microsoft CoPilot or AWS Code Whisperer internally will make your engineers more productive and happy to use bleeding edge technologies.

Leveraging LLMs internally will make your domain experts and your knowledge team more productive and happy.

Looking at company data as valuable assets for AI that can be used internally or shared or sold externally (think Reddit's recent agreement with Google).

The bad

When you are in the data business, you must return the correct answer all the time, along with a reference to the source material. Unfortunately, LLMs are known for confabulations.

Without guardrails, LLMs may return the wrong answer; you may lose your customers’ trust and then their business. The “WANTED” illustration opening this article is a good example (visible typos and some gibberish).

The ugly

The full legal ramifications of LLMs are still uncertain, with numerous pending litigations going on. Using an LLM for commercial use may put your company at risk.

When used internally by developers, LLMs can leak information to the outside or can inject poisoned code into your company codebase.

Appendix B: recent news coverage on the topic

• Every US federal agency must hire a chief AI officer (Mar 2024)
• The AI Role All Companies Are Looking To Hire | MOONSHOTS (Mar 2024))
• Hottest Job in Corporate America? The Executive in Charge of A.I. (Jan 2024)
• The Rise Of The Chief AI Officer: Powering AI’s Corporate Revolution (Feb 2024)
• Chief AI officer: What it takes to land the C-suite’s hottest new job | CIO (Nov 2023)
• Chief AI officer is tech’s hottest new role — but finding workers with the right skills is tough (Feb 2024)
• What does a Chief AI Officer do? (Jan 2023)

Acknowledgments

• ChatGPT4 for some fragments of this article.
• my partners in crime at Gro Intelligence: Jacques Paye , Jackson Sargent , Elia Schneider , Himanshu C. , and David Ismay .
• all the people who provided feedback on earlier versions of this article including Robert Cheetham , John De Oliveira , Beth Simone Noveck , Henri Palacci , Jerome Pesenti .