I spent June 9, 2026, at the Atlanta CIO Community Executive Summit at the Marriott Northwest at Galleria, surrounded by peers wrestling with the same question in a hundred different forms: now that AI can do so much, what exactly should humans keep doing?
Here is my honest recap, the ideas that stayed with me, and the questions I am carrying back to my own organization.
The Keynote That Set the Tone: AI-Enabled Enterprise, Beyond Lazy Thinking
The opening keynote, featuring a leader from Mercedes-Benz, did not start with a demo. It started with a confession that most of the industry avoids saying out loud.
Yes, AI writes code today. AI analyzes reports. AI drafts our emails. And yet, we simply do not know for sure whether automation and AI will ultimately create more jobs than they destroy. Anyone who claims certainty on this is selling something.
The more interesting argument was historical. We have been offloading thinking to tools for decades:
- Calculators took over arithmetic
- GPS took over navigation
- Search engines took over recall
- AI is now reaching for judgment, critical thinking, and direction
That last one is different. The first three offloaded mechanics. This one offloads meaning.
The line of the day, the one I wrote down and underlined twice:
"Speed without understanding is debt with no repayment plan."
The Two Failures Every CIO Should Fear
The speaker framed AI risk as two opposite failure modes, and both are equally real.
The invisible failure. A junior developer ships AI-generated code they cannot explain. It works, until it does not. And when it fails, they cannot debug it, because they never understood it in the first place. Nothing looks wrong on any dashboard until the moment everything is wrong.
The visible failure. This comes in two flavors. Reckless adoption, which is speed without judgment. And paralysis, which is waiting for a certainty that never comes, while the ground is already lost to competitors who moved.
The escape from both is the same: training that teaches people how to think, not just which buttons to press.
Governance as the Foundation, Not an Afterthought
What impressed me most was that Mercedes-Benz has operated on published AI principles since 2019, long before the current wave made it fashionable:
- Responsible use
- Explainability
- Privacy protection
- Security
- Reliability
And across every single use case they shared, the same pattern repeated:
AI handles volume; humans handle judgment
AI drafts, humans decide
AI flags, humans investigate
The keynote closed with three questions every CIO should be asking, and I am stealing them for my next leadership meeting:
Where does speed serve us, and where does it cost us?
Are we building capability or dependency?
Who is accountable when the AI is wrong?
The winners of this era will not be the ones who deployed the most AI. This is not a technology transformation. It is a human transformation. And in true Pareto fashion, the vital few use cases will deliver most of the value while the trivial many consume most of the noise.
One final warning that deserves to be a poster on every office wall: AI will not make your organization lazy. It will make laziness hurt.
Breakout: Executive Workshop, What Is Good AI?
This session was less of a presentation and more of group therapy for technology executives. The discussion prompts were sharp:
Organizational alignment. How do you define good AI in the context of your organization, its goals, and its risk appetite? Notice that the definition is local. Good AI at a bank is not good AI at a startup.
Success and failure points. Peers shared examples where AI initiatives succeeded or failed based purely on alignment with principles, not on the quality of the model. The differentiator was rarely the technology.
Looking ahead. How does the definition of good AI change over the next 6 to 12 months as we move deeper into agentic and AI-native systems? When software starts taking actions instead of making suggestions, the bar for "good" rises sharply.
Beyond the bend. Where will it get harder to tell good AI initiatives from bad ones? My take: when everything has AI in it, the label stops meaning anything, and only outcomes will separate the two.
Breakout: The Modernization Imperative, Building for Agility and Scale
Kim Seabrook, CRO of OutSystems, and Ivan Palhegyi, Head of Transformation at Gen Re, made the case that modernization is no longer a project with an end date. It is a permanent operating posture. Legacy is not what you built ten years ago. Legacy is anything you cannot change quickly.
Keynote: The Trust Recession, How We Got Here, How We Get Out
This HP Workforce Experience session was the emotional core of the day. The argument: we are living through a recession of trust between employees and employers, and technology decisions are quietly making it worse.
Hybrid whiplash. The unwritten contract between employee and employer keeps changing. 37 percent of companies are enforcing office attendance in 2025, up from 17 percent in 2024. One in four bosses admitted the mandates were partly about attrition. People notice. And people do not trust the AI bills they are paying either.
The SaaSpocalypse. A word none of us in the room had heard before, and now none of us will forget. The average large enterprise runs 1,000-plus SaaS applications. Each employee switches between 11 or more apps just to get work done. Meanwhile, the tech industry has cut over 1.2 million jobs, up 66 percent year over year, while spending 650 billion dollars on AI. That contrast is the trust recession in a single sentence.
The way out: signals and the right kind of loop. Your network, your devices, your apps, and your meetings all generate signals. Used well, they place the right compute in the right place and rebuild the experience instead of surveilling it.
The most practical framework of the day distinguished two postures:
Human on the loop, where AI acts, and humans supervise: ticket routing, forecasting, device self-healing
Human in the loop, where humans decide, and AI assists: budgets, vendor selection
Knowing which loop each process belongs in is half the governance battle.
The Board Perspective: AI Is Not a Strategy
A recurring theme in the boardroom sessions: most boards care about three things.
- Strategy and risk
- Capital allocation
- Governance and oversight
AI is not a strategy. It is a means to an end. So the only question that matters to a board is: will it deliver those three? That is what you have to prove.
A useful shipping heuristic emerged for prioritization. If a use case is predictable and repeatable, like cloud cost optimization, ship it. If it is not predictable and repeatable, like a full data center exit, it is not yet ready for autonomous AI. Not never. Just not yet.
And one number that should keep every CIO awake: 70 percent of non-executive directors lack confidence in the value of current IT investments. The key to less pain and more gain is relentlessly linking IT investments to shareholder value. Not annually. Relentlessly.
Breakout: Empowering Human Capital, The Human Side of AI
The Zapier-flavored session on adoption struggles produced the most immediately usable playbook of the day:
Reward ideas, not just outcomes. Recognition drives adoption faster than mandates
Ask people to tell you the thing they hate most about their job, then aim AI at exactly that
Market internal wins loudly. Promotions and shout-outs after successful AI adoption send a signal
Run hackathons to surface hidden champions
Reframe the fear: AI is not going to replace the job; it will eliminate the mundane parts of it
Start with one enthusiastic person, prove it, then use that success to pull in more people
Adoption spreads person to person, not memo to memo.
Breakout: Emerging LLMs, The Next Security Challenge
The security conversation has matured. The discussion covered where vendors are now taking responsibility, the emergence of configuration standards, evolving GPU-level standards, and vulnerability management through whitelisted systems and continuous scanning. The takeaway: LLM security is becoming an engineering discipline with checklists, not a research topic with papers. That is progress.
What I Am Taking Back to My Team
If I compress the entire day into three commitments:
Audit every AI use case against the loop test. Is this human in the loop or human on the loop, and did we choose that deliberately?
Kill the invisible failure. No one ships what they cannot explain and debug. Understanding is the price of speed
Link every IT investment to shareholder value in language a board member would use, and do it before they ask
Thought-Provoking Questions
- If AI made your organization 10x faster tomorrow, which of your current mistakes would it make 10x bigger?
- Can your newest engineer explain and debug the last piece of AI-generated code your team shipped?
- Which processes in your company are human-in-the-loop by design, and which are that way only by habit?
- If 70 percent of your board lacks confidence in IT investments, whose job is it to change that, and what is the first proof point?
- What is the one mundane task your best people hate most, and why has AI not been pointed at it yet?
Top comments (0)