Where Should AI Live?

Seems obvious doesn't it? AI is an algorithm. Algorithms are data science. Data science is part of Information Technology.

Except it isn't that simple.

I am suggesting that in fact, AI is more human resources. Now, before we start using the "A" word (anthropomorphize), let me present my points of consideration.

1. Evolution

As AI has evolved over the years, the models have become more advanced, the learning capacity and method of improvement have become increasing automated, and in fact for some tasks, there is now a level of self-sufficiency and autonomy, especially with agentic AI deployments where models use other models to inference and apply RAG.

As a result, the operation of and interaction with these models are more humanlike now than ever before. I'm not saying they're sentient, or even AGI, but they can perform tasks in tandem with a human.

2. Lexicon

Think about the terminology we use when working with AI. The left column is how I "grew up" in the business. We were programmers. We coded, debugged, looked for malware, and found errors.

Now, prompt engineers teach models, inference with them to make them better, look for poisoning attempts, and try to avoid hallucinations.

The entire AI industry has moved from traditional technical terminologies to human behaviors, ailments, and processes.

We have already acquiesced the point that these models need to be viewed as human analogs.

3. Support

In the past, IT teams understood their discrete functions but all had general knowledge of coding, scripting, IT policy and business process.

Now, the people who build models are deep experts in mathematics, well beyond what most IT staff would grasp. Read any detailed paper on a new model or methodology for AI on arXiv.org and you won't be more than 2 pages in before having to pull out 300 level college texts on theorems, proofs, and advance calculus.

But the people who train and use these models are the opposite; they're focused on business process, ethics, use cases, and how to "speak" to the models. While not experts in math, they're experts in law, policy, governance, and business requirements.

And the two sides don't really need to work closely together. A model designed for language/human interaction needs to be built with the same accuracy, capacity and parameters no matter what the use case may be.

4. Engagement

When we interact with models today, we speak in our native languages, not computer languages, and the systems return that favor. Be clear and specific

"ask follow-up questions or for clarification if my response isn't what you need"

The entire process of interaction could just as easily be with another human. The Imitation Game could be played to its logical conclusion.

The Point

You may be thinking to yourself, "ok, I see these points, but how does this mean it's an HR function now?"

I'm glad you asked.

The IT team is a cost center - it is designed to deploy technology to support the business and its users. It stores data, presents that data for consumption, makes sure everyone has compute resources and access to information. It ensures everything is secure.

In this context, modern AI models are users of IT resources, not an IT resource

The HR team is responsible for training on corporate policy, who employees can interact with and what they can say, what's considered to be sensitive information, identify bias, avoid sending sensitive information to people outside the organization, etc.

Now, here's the critical thinking question -

Which of these two functions is more essential?

If IT resources are down, business can still function. If there are advanced AI models and humans at a company without the HR functions, the business will collapse.

We evolved AI to be humanlike. We build them using human terms. We support them in an organization as human analogs. We interact with them in natural language. We govern them under the policies written for humans.

Sounds like AI models are effectively "staff augmentation", all of whom report to... HR.

Just something to consider. Remember, the value of AI is not measured by its parameters, performance, or cool factor. It's measured by its ability to advance business process and human work. If HR isn't involved in a leadership capacity, lack of governance and increased business risk are genuine considerations.

As with all topics, the content here is my thoughts alone and does not reflect that of my company.