Deja Vu with AI FinOps

Deja Vu for CIOs.

Back in the day we limited resource consumption by turning off lights, turning down the thermostat, etc. on factory floors when you didn’t have a third shift. I spent 10 years in factory automation and robotics so I could see our operations teams managing their spend with utilities — it was obvious when we were wasting resources.

Then we progressed to cloud consumption and deployed FinOps solutions to help us keep our monthly cloud bills under budget. In 2019, I attended a major annual cloud conference where I focused on FinOps sessions to learn how others were managing costs and driving down resource consumption.

Based on events I attended the last few weeks, we are all seeing this trend repeated with AI.

AI FinOps today looks different but goes after the same goals — meter the use, make folks accountable to be responsible with their resource consumption, and partner tightly with your finance leads for your forecast.

It’s new and complex.

In the last few weeks…

• I had an opportunity to talk about enterprise AI adoption in two recent podcasts. I am excited about what this can do for us, encourage experimentation not just exploration, and keep challenging myself to do what others are finding valuable in their everyday lives.
• I went to two sessions on AI Governance hosted by a local data and AI company. Those of us at companies with amazing histories and varied portfolios know that our foundational systems and data are what AI need to build upon — and since these data were not structured for this, we need to roll up our sleeves and get them there.
• I published a couple of blurbs in Forbes. Leaning in on leadership topics, my contributions are about the managerial challenges we face in times of technology dislocation — it’s tough but we are in this together.
• I met up with a group of CISOs, a group of CIOs, and am in a CIO group chat. In each of these settings I ask my peers what they are doing to monitor and manage token consumption — no easy solution here so we are sharing what works and what doesn’t.
• I was a panelist at an amazing educational leaders conference. Everyone has an opinion on AI in education — using it for writing papers, doing homework, and ideating can impede or enhance learning. Alleviating the productive struggle that comes with learning can stand in the way of us learning how to learn. <definitely a subject for a longer article>
• I also attended a neighborhood meeting on the adverse effects of data centers. If we all take a NIMBY perspective, the resources will be constrained which will affect supply and demand for service, likely leaving many behind. That said, responsible data center expansion must be tackled through policy and practice. <another one for a longer article>

We all agree that CIOs are on the hook for defining what responsible resource management and AI governance looks like as we navigate these waters together. Don’t just stand back and wait, watching the blinking lights while the invoices creep up. Get out in front of this and build an agent, buy a service, and get creative.

Restricting AI access won’t work — it will only increase your risks and decrease the productivity opportunity you have with this new tool. I love shadow IT and shadow AI since they tell me where I need to improve my service catalog. Create communities of practice to share what is working from content creation to agentic workflows. Listen to your internal customers and create a service offering for them.

Empower your teams with AI, train them and monitor their resource consumption to ensure your team is doing some experiments that lead to best practice for your business.

You got this.

It is deja vu for CIOs. When it came to cloud, you used an “if – then” approach. You assigned each workload a disposition for cloud depending on what that workload was doing.

• If cloud is being used for DR, then have it on demand.
• If you need seasonal 24×7 uptime, then scale with containers.
• If there is steady usage, then purchase reserved instances.

Let’s apply what we did before in new ways to handle AI token consumption and manage our resources responsibly and proactively.

If – then for AI tokens.

• If a simple task, then use a non-token seat license.
• If you have a complex need, then select a more sophisticated token-based service knowing it consumes more resources.

If – then for agentic AI tokens.

• If a simple request from your internal customers, then give them an SOP.
• If it’s more complex, then give them an agent on top of your knowledge management system.

If – then for avoiding AI tokens.

• Load log data for AI agents to analyze.
• If you know what you want, then automate it.
• If you are curious, then create and test an agent and then automate it.
• If you have superfluous data, then eliminate it.

If – then for eliminating AI tokens.

• If you have a process where you think you want to apply AI, then…
• First eliminate non-value-added tasks… then automate… then apply AI.
• In that order — be AI forward but not AI first.

You were cloud forward but not cloud zealous, do the same for AI. “Your mess for less” only lasted a couple of years. We got smart about our resources and found the right tool for the job. Same is true today.

What you can do today.

Take the journey one step at a time.

1. Map your business processes. Take a lean approach and eliminate the non-value-added tasks then automate the ones that provide the least value to your customers. I love Wardley strategy maps and there are other classics like Lean Six Sigma.
2. Focus your team and the organization on those business processes that mean the most to your customers. Enhance your customer engagement by giving them the personalized service that is enabled by automation and AI. Put this in your team’s objectives, celebrate learning, and reward success.
3. Empower your employees with assistive and agentic AI training and tools to increase customer engagement and decrease administrivia. You’ll increase retention and improve employee experience.
4. Continue to hire recent grads and interns — don’t just use agents for these entry-level jobs. AI enhances our performance and service delivery through all levels of the organization. We need to keep the talent pipeline healthy in this uncertain time. As a parent of twenty-somethings that learned through COVID, let’s continue to develop this next generation of leaders who learned the hard way how to push through adversity.

It’s a time for the CIOs who understand the business to provide exponential value. Take advantage of this confluence of events!

NB: I use assistive and agentic AI everyday. I chose not to use any AI for this article. Thoughts and words are my own and any misspellings or long-windedness are purely mine.