New Research: The CIO in 2030: Running the Enterprise Intelligence Economy thisisfrontier.ai/research/c… Fourteen operating imperatives for a strategic world where people, agents, software, compute, data, and trust are dynamically allocated to outcomes. #CIOChat

Wait until we get to the revenue-per-worker discussion, as enterprises have to start accounting for not just the cost of human labor, but digital labor (AI agents) as well. Given the more ephemeral nature of agents, we don’t really even have good KPI to use yet.

Agents can stay on task longer than ever (doubling every 7 months in length of time), but most enterprises are still reluctant to let them access systems of record. My take: Not just better guardrails are needed, but new database capabilities that can track the “red threads” of agentic data changes and their ripples through IT systems.

AI upgrades from @Vultr shift infrastructure buying toward control-plane readiness thisisfrontier.ai/news/vultr… The neoclouds continue to get smarter + wider in terms of the ecosystems capabilities they are building.

I believe we are currently underestimating how much of AI’s business impact will come from going deep into individual industries. Healthcare is giving us an early look at what that transformation could mean. Consider @ConcurrenceHQ. The company reports more than 10 million patient interactions. A September @Databricks case study puts its production AI at 14 models, 11.2 million model calls, and 100.8 billion input tokens every 30 days. That is credible initial operating scale when using advanced AI. Their agents support patient outreach in 100+ languages, scheduling, medication follow-up, and continued contact between appointments. They connect those interactions to patient records and clinical workflows, with escalation to the care team when needed. Underneath that is an assembly of AI capabilities that power it. Public documentation includes Claude agent models, @OpenAI realtime voice options, and specialist speech models. The value comes from making that intelligence useful right within the demands of actual care delivery. What interests me most is the possibility that solutions like this will directly arrive transformation of patient care. The infrastructure also deserves some precision analysis: Concurrence offers on-premises and cloud deployments, including customer-controlled and dedicated Azure environments. Where the models actually process patient data depends on the configuration. CIOs need to understand the inference path as thoroughly as the location of the medical record. And “superintelligent care” is its overall goal. Learning across consenting institutions and predicting individual care trajectories still require validation. But the ambition is the correct one and significant. This is why I expect vertical AI to become a major engine of the next wave of enterprise transformation. Healthcare, banking, insurance, and other regulated industries need intelligence that understands their work, can act within defined authority, and produces evidence people can trust. There is a historic business opportunity in bringing these pieces together. Domain expertise, workflow integration, governance, and accountability become part of the product itself. In short, watch this space.

Fascinating how hard collaborative work management tools like @airtable picked up on AI integration. It’s the correct move: Coordinate work in the full context of AI (and AI as workers) is where it’s all headed right now.

AI will democratize capability while concentrating the proceeds. @vgr’s new “Cognitive Mercantilism” gets something important right: We keep treating today’s inventory of work as though it exhausts tomorrow’s possibilities. His argument is that cheaper intelligence will unlock entirely new ambitions, much as earlier economic transformations expanded what ordinary people could produce and consume. The demand for intelligence can grow as people — and agents themselves —discover what they can accomplish with it. I think that upside is profoundly underestimated. This connects directly to why I keep coming back to Jevons Paradox: Efficiency can stimulate enough new demand to increase total consumption. Give a small team capabilities that previously required a research department, a software company, and a consulting budget, and previously impractical projects become plausible. We have an enormous backlog of problems we have never had the capacity to address. That is also the opportunity behind what I call Mission Engineering: defining worthwhile outcomes, orchestrating people and agents, establishing constraints, and verifying results. Human ambition gets a much larger execution capacity. Turning that capacity into useful, accountable work becomes a discipline in its own right. But Jevons can keep the GPUs busy without keeping the payroll full. Expanding demand for intelligence does not guarantee expanding demand for paid human labor. Agents will probably perform much of the newly affordable work, too. Productivity and output can rise while workers lose bargaining power or receive a smaller share of the proceeds. The economic bridge from “people will want more” to “people will earn enough to participate” still has to be built. The agricultural analogy leaves that transition question open. A tractor could displace a farmworker; it couldn’t follow that worker into accounting. AI will simply compete in some of the new occupations we expect to absorb displaced workers. That makes the speed of adaptation, the creation of new human roles, and access to productive assets central to the outcome. Rao does acknowledge exploitation and the politics of access. I would put them at the center of the argument from the beginning. Cheap intelligence can coexist with expensive permission to use it commercially, dependence on proprietary context, and tightly controlled access to customers. An affordable agent subscription gives someone a powerful tool. Whether they can build an independent livelihood with it depends on who controls everything around that tool. His more speculative extension is the idea of autonomous AI systems becoming consumers themselves. That could generate real economic activity. But agents buying services from other agents tells us little about whose income rises, who owns the resulting assets, or who bears the costs. Economic interdependence is not proof of shared interests. For enterprise leaders, this is where the philosophy becomes an operating decision. Build pathways into mission ownership. Preserve apprenticeships. Give people paid time to acquire transferable skills. Share the gains, insist on portability, and make delegated authority accountable. These are practical ways to turn greater machine capability into greater human agency—the central concern of our Future of Work research. A headcount target is a remarkably small ambition for a technology with this much potential. I share Rao’s optimism about how much human potential remains unexplored. The test is whether ordinary people gain the power to decide what gets built, pursue their own purposes, and retain a meaningful stake in the value they create. Otherwise, cognitive mercantilism just gets a very efficient upgrade. contraptions.venkateshrao.co…⁠

Anthropic’s IPO will be industry-defining and will likely map a key trajectory that AI labs take afterwards. Our thoughts on what CIOs that depend on them should do to prepare, depending on how it goes: thisisfrontier.ai/news/anthr…

My take: With the growing bow wave of major math and science breakthroughs created by AI, it’s now clear that many orgs will need to prepare for extremely vigorous competition in the realms of innovation and R&D. Whole new markets are in the process of being created + old ones disrupted.

It what is probably a surprising turn of events for cybersecurity professionals, it turns out that zero-trust architectures with strict micro-segmentation are probably cheaper and easier than relentlessly hunting down and patching all possible security exploits…

This really blew up, and that’s good. It’s a vital conversation about AI that we must have now. “Risk is our business.” This is true if you’re using AI. And true if you’re not is the issue. How do we compartmentalize the extremely powerful tech advances happening now? Probably we can’t is the answer. But we must discuss it. A global debate.

A call to action by Anthropic’s deputy CISO: Things in cybersecurity will get worse before they get better. Plan for this for the long haul. CIOs must now go to their boards for long term resource increases, as difficult as that is right now. Get out ahead as far as you can.

Yes, focusing on the cost of tokens only is a fool’s errand. But no, managing the cost of tokens with clear goals in mind is very much not and must be part of modern cloud FinOps.

AI will soon dramatically impact the workforce. The majority of workers will go through meaningful to high levels of reinvention of what they do on a day-to-day + role basis. Figuring out the best paths forward with AI must a top priority of org leaders. Source: @McKinsey_MGI

Depreciation and amortization costs for the five big hyperscalers are projected to reach $590 billion in 2029. The big question: Will their profits exceed the cost? My take: They are certainly betting they are and I’m sure they have climb-down plans if not. The big money is on the token demand being there.

My thoughts of late have returned to this amazing scene by @WilliamShatner, made decades ago. As a young person, I found it tremendously inspiring. A code to live by. But it didn’t play out quite like we expected. The “fantastically advanced” life that we actually ended up encountering is right here, on Earth. All around us now. The key issue of our time is becoming this: Not everyone signed up for this mission, or the risk. And what we do about that, ultimately, is where the debate about AI will end up.

A key insight from @sama at @Stanford this week about the current global cyber situation: “I no longer think we can just patch all the bugs. I think we’re going to have to completely switch the defense model.”