Hiring a Chief AI Officer is about more than AI

Several board consultants have recommended hiring a CAIO (Chief AI Officer) reporting to the CEO. Peter Diamandis said in an Inc. article:

There are two kinds of companies this decade: those that are fully utilizing AI, and those that are out of business.

Even if a company doesn't adopt AI, VR glasses, robots, and whatever else comes along, their customers will, so the company needs to adapt to support those changes -- changes that are accelerating exponentially toward a "technical singularity" when everything changes all at once.

This article evaluates the top objections and the reasons:

• AI is different
• AI is pervasive
• AI is changing fast
• AI needs to be managed differently

A. Why not also hire a Chief Copy Machine Officer?

Such a quip underlies the misunderstanding of why a CAIO is needed: the potential impact of AI -- both for good and bad -- is so much more than installing any previous technology. AI is about more than a single technology.

AI is being inserted into software used by every part of your company. AI capabilities are being embedded into every major software offering:

• Microsoft has added AI into every product is sells: Chat in Edge browsers and Microsoft 365 suite (Word, PowerPoint, Excel, etc.)
• Salesforce has added it's "Einstein" AI capabilities in all its products
• SAP has added Generative AI capabilities in all of its platforms

And they have been doing it for years. So your company needs to move faster.

Such pervasiveness means your prospects, customers, and employees would be expecting it from what your company offers.

To compete in an AI world, companies need to break out of "business as usual" in order to stay ahead of competitors and keep top talent. Changes in how AI works and what it impacts are changing fast. [1] CAIOs would work with company product teams on the capabilities your prospects and customers will be asking for.

B. Isn't AI just another feature?

No. AI makes use of a major shift in how software is produced.

AI is the enabling technology for the "Fifth Industrial Revolution". Unlike the Industrial Revolution, companies who think they are saving money by not training the gig workers they employ limit their future to what individuals can learn on their own. New kinds of workers and companies are needed.

Traditionally, software has been created using human programmers to create logic programmatically. Such programs can be tested by running them line by line.

However, AI models are created based on automatic derivation from data.

Unlike hand-coded programs, the logic an AI model makes are not transparent. This has resulted in EU legislation for "explainable" AI models.

The use of historical data means that AI models "bake in" historical biases and even accentuate the biases further -- a part of the controversy surrounding LLMs (Large Language Models) that form the basis for "Generative AI" products such as ChatGPT. Companies that build the models have not been transparent about the "corpus" of data from which their models are based. Was the model based on Stackoverflow or other sources that are known to contain some bad advice?

C. Aren't CAIOs salespeople for AI companies?

No. Actually, the opposite needs to be true. [2] An effective CAIO would inject a voice of reason and knowledge to validate claims made by salespeople who could otherwise mislead individuals. Like "Chief Medical Officer" some companies brought in during the pandemic, CAIOs provide an ally for people in the company to address the dark side and downsides to using AI which salespeople have a self-interest to ignore. For example, ChatGPT and other "Generative AI" are known to "hallucinate" (come up with false facts) while sounding authoritative. Alternatives (such as Perplexity.io) address that concern by citing its sources. But the concern remains.

[3] CAIOs can provide foresight on the implications of AI adoption. For example, lawsuits from Stackoverflow or some other entity claiming compensation from your company allegedly profiting from their data. Could the strike by Hollywood creatives around royalties for AI have been averted by more proactive dialog? What conflicts would arise if your company used artificially created faces and text rather than real people -- and others call that out publicly? [4] CAIOs som be monitoring laws the US and EU are drafting today and what they might enact in the future so they can prepare for compliance (as a competitive advantage).

D. Do CAIOs prepare for mass layoffs?

AI capabilities are causing a realignment of what people do in every profession. That's why companies incorporate AI features into their products.

You may not be replaced with AI, but you'll be replaced by someone who uses AI.

In 2024, Google, Facebook, and many other tech companies had massive layoffs while they posted jobs requiring AI skills.

[5] A CAIO may prevent disruptive layoffs. To maintain employee loyalty, a CAIO would work with HR and each department manager to define the skill sets and jobs needed, then help implement plans to transition managers and employees. This is an iterative process. AI is so new that there are very few experts. So a company might as well enable employees to learn to use AI. Allowing employees the freedom to learn and experiment with AI can enable a company to innovate faster than competitors, through people who already know their company's industry.

Some ways to make AI commonplace:

• Ensure there is a budget for employees to learn and cloud time to experiment.
• Evaluate and select training organizations (such as OReilly.com, Coursera, and cloud vendors) to provide training for better consistency. Focus on improving problem-solving, decision-making, and critical-thinking skills.
• Provide a way to onboard people quickly yet securely. This includes assigning GitHub, API keys, and cloud credentials under "Least-Privilege" principles but without an arduous approval process.
• Include in training how to leverage automated "MLOps" (Machine Learning Operations) pipelines that scan data for errors and bias, then scan code for security vulnerabilities.
• Provide an easy way to make AI backend assets visible via front-end websites and API gateways with the security.
• Hire tutors and mentors from other companies in another industry to guide learning, clear roadblocks, and provide inspiration. [6] These and the in-house CAIO would customize messaging appropriate to each type of group in the organization (directors, managers, developers, salespeople, etc.).
• "Drip" short emails and/or Slack messages to cover important content, send surveys, and announce events.
• Arrange parties where innovations can be demonstrated, and record the presentations for private viewing within the company.

The above defines what it means to really "work hard" toward AI. But a CAIO also reminds people of this crucial point:

"We experiment with AI not for the sake of adding a technology, but to improve customer retention, security posture, and employee productivity."

E. What else can CAIOs do?

Currently, it's too early to tell which AI vendor will dominate the market. Each cloud vendor has been providing AI capabilities for years and announcing surprising capabilities amid their rebrandings:

• Amazon has its "SageMaker" and "Alexa"
• Microsoft transitioned from their "Cortana" to "Cognitive" to "ChatGPT" from OpenAI
• Google has their "Bard", recently renamed "Gemini"

[8] A CAIO would help companies avoid the huge penalty in switching among vendors. If you have one part of the company building creating assets using Google's Tensorflow (with Keras) and other teams creating in Facebook/Meta's Pytorch, it would be much harder for the teams to collaborate and innovate as a whole.

LLMs (Large Language Models) that incorporate billions of words have thus far come from cloud vendors who have the deep pockets to afford the millions of dollars to takes to train large models. So the choice of an AI model can be tied to a choice in cloud vendor.

This tie-in may mean that companies implement "multi-cloud" systems that enable easier switching among different cloud vendors. For example, would it be better to use Google's edition of Kubernetes or the edition that comes out later to operate on Microsoft's Azure and AWS clouds?

[8] A CAIO may help with various decisions for their companies to achieve greater resilience to change by "reinventing" themselves. For example, companies are increasingly called to make a stand on issues (LBGT, abortion, etc.). Aligning with one side would cause boycotts by the other side. AI adoption is becoming a divisive issue. One solution may be to split the company into different companies - one going all in for AGI (Artificial General Intelligence) where robots can rule the world, and another "AI free" using traditional coding values.

F. Wouldn't a CAIO create more politics?

Some caution that adding a CAIO would disrupt the status quo of power dynamics among other executives. That is the point of getting a CAIO.

[9] CAIOs can be the "referee" in balancing how concerns from each part of the company are addressed to benefit the company as a whole. As before, there are likely "data silos" which need integration.

[10] CAIOs can work with others to leverage company-wide adoption of systems to support human-AI collaboration:

• Dashboards projecting the plan and progress toward objectives (such as Grafana, Tableau, etc.)
• Progress is captured automatically from systems to make status and potential issues visible
• A common calendaring system that makes it easy to arrange both meetings between people and automated events (such as security certificate renewals)
• A common task tracking and project planning system that issues alerts (such as Jira, Asana)
• Common tools to create dynamic diagrams (such as Lucid) and videos
• A system to escalate tasks not resolved in a timely way (such as PagerDuty, ServiceNow)
• A system to take regular surveys of opinion and trivia to measure the mindset of people
• Open and private discussion forums (such as Slack)
• Regular scheduled "fire drills" to ensure the organization responds quickly to incidents such as ransomware or sudden failure in AI systems

Anything else, please let me know.

// Wilson (https://linkedin.com/in/wilsonmar)