It turns out that tech output productivity has tilted up sharply the last couple of years.
It is clear that there is now a new ghost in the machine.
And it looks a whole lot like an LLM.
It turns out that tech output productivity has tilted up sharply the last couple of years.
It is clear that there is now a new ghost in the machine.
And it looks a whole lot like an LLM.
Overheard: The model is now the commodity. And the context is now the moat.
24.02.2026 02:16 — 👍 1 🔁 0 💬 0 📌 0
2/ The mistake orgs make is waiting for costs to fall before acting.
The expensive phase is the learning phase.
Act now to build governance, agent FinOps, operating models for fleets of digital labor.
Those who learn to manage agents now will run tomorrow’s digital workforce.
1/ Many debates today around statements like this:
The marginal cost of running an agent will drop essentially to the cost of electricity.
Right now, running agents is indeed very expensive.
But make no mistake, like all tech, the cost in the long term will fall towards zero.
Open-weight models no longer chasing proprietary AI. They’re drafting behind it at full speed.
Gap likely closes for good by 2027 as OSS outpaces frontier scaling.
CIO warning: Don’t lock strategy to model leaders. Advantage shifts to orchestration, data control, cost governance, not raw model IQ.
My latest research:
The Great CIO Platform Reset: Why Agentic AI Is Forcing a 2026 Reckoning
futurumgroup.com/press-releas...
Bottom line: AI agents are reshaping enterprise platform strategy.
My comments in Infoworld on GitHub readying agents to automate repository maintenance (oh, and coding)
www.infoworld.com/article/4133...
My take: They are careful to position it as agents tending grunt work in repos, but they’ll basically be able to do most coding tasks.
The 7 biggest S/4HANA migration hurdles — and how to overcome them
www.cio.com/article/4130...
My take: Many CIOs I speak w/ are accumulating SAP instances faster than they can combine/rationalize them.
ERP industry needs to adapt to AI speed if it doesn’t want to be left behind.
Most CIOs regret AI vendor, platform decisions
www.ciodive.com/news/cios-re...
Nearly 3 in 4 CIOs wish they hadn’t made certain AI vendor choices. Over half have had their CEO challenge them.
My take: The AI budget runway will run out soon unless major results start coming in.
Just wrapped my latest global CIO survey, many fascinating industry shifts.
We’re seeing double-digit moves across several core enterprise platforms, the kind that signal structural transition, not normal budget churn.
Something bigger than AI adoption is underway: Reimagining
Overheard: Is on-premise the new cloud?
For AI, the answer appears to be yes.
One of the next big competitive business domains is token economics.
AI Inference Will Drive Increased Cloud Native Software Consumption
cloudnativenow.com/features/cnc...
Which is going to be a key issue: Are AI coding models actually onboard cloud-native bandwagon? I will be doing some research to find out.
The struggle for good AI governance is real
www.cio.com/article/4128...
Not just real, but critical in many orgs. Creating sanity while servicing stakeholders’ often unrealistic expectations is the order of the day.
Global sovereign cloud spend to increase 36% in 2026
www.ciodive.com/news/global-...
Proximate cause: The geopolitical risks of keeping data in other countries has sharply risen, along with desire for cost control, more vendor responsiveness, and desire for regional cloud growth.
Are they able to be used and will they be used? Certainly.
Are they actually ready for the enterprise? Not hardly. But they have to start somewhere. ;-)
These adds/changes to the roadmap force org change, cost model resets, and new operating muscle.
Low risk equals incremental.
2026 Azure is transformational.
That’s why the risk is real. And why the upside is too.
The good news: Microsoft will likely get many of these out.
For the 1st time in a while, the Azure enterprise roadmap is all orange/red.
Not because Microsoft can’t ship, but because what’s ahead now = hard.
Agents, governance, confidential AI, Arc as a control plane, platform consolidation aren’t features. They’re architectural shifts.
My take: OpenClaw mostly uncovers the security vulnerabilities which were always there but are now exposed at human scale.
While it does have some of its own security issues, basically it's also a way to find all the security vectors you never knew lived in your IT stack.
Mostly a good thing.
5/ Prepare for what’s next:
Far more packaged enterprise offerings, hardened skills, deep IT integration, governed access to enterprise data, private and on-prem models, policy layers, auditability, and cost controls.
The enterprise era of OpenClaw is just getting started!
4/ Early enterprise OpenClaw players, cont’d:
“Zero Devops” w/ backups: xcloud.host/openclaw-hos...
Trusted execution: near.org/blog/how-to-...
Put others in comments below, please!
3/ Here’s the early players:
Most secure: runlayer.com/openclaw
Fully-managed + scalable: elest.io/open-source/...
The most enterprise features: kilo.ai/kiloclaw
Managed + affordable: v2cloud.com/solutions/op...
Cont’d👇
2/ It just became the fastest-growing open-source project in history, w/ pure, personal agentic AI.
But OpenClaw’s most important legacy might not be hobbyists. It is likely businesses.
It’s very early.
But the initial enterprise ecosystem is forming fast.
Here’s the 1st wave👇
1/ OpenClaw for the enterprise has just begun to arrive. It emerged out of sheer necessity.
Once workers try to use OpenClaw agents to run workflows, touch credentials, access systems of record + act for the business, then security, governance suddenly becomes basic survival.🧵👇
We keep re-inventing management theory with metaphors from computing.
Mainframe era → Company = a hierarchy
PC era → Company = a file system
Internet era → Company = a network
Social era → Company = a social graph
AI era?
Company = a living model
Management → architecture
AI gold rush to drive 2026 IT spending — as IT services get the squeeze
www.cio.com/article/4128...
Overall IT spend increases by huge 10% but it’s bursty, somewhat uncontrolled as AI remakes the computing landscape. Some SaaS starting to get moved out, as the AI surge continues.
My Take: OpenClaw signals a real platform shift. But it’s still a power user’s tool. High leverage, high risk.
The breakout opportunity is obvious: Simplify, commoditize + de-risk personal agents for the masses. Whoever cracks this layer becomes the OpenAI of the agentic era.
My take: I think OpenClaw will advance the development of AI agents more than just about anything else in 2026.
It has high highs and low lows, having taking it for a real spin over the last week.
But it’s also moved us closer to a commonly accepted model for agentic work.
Claude has really come up in enterprise adoption of frontier LLMs.
They took the top spot from OpenAI last year, particularly due to the high capability of Claude Code, but they have continued to press their market share lead into 2026.
Google is also catching up. A real race.
IT leaders grapple with AI agent sprawl
www.ciodive.com/news/it-lead...
More than 80% of IT leaders think the proliferation of AI agents will yield more complexity than value.
Yet users are often not receiving the access to AI that they are hoping for at work. Shadow AI a real issue.
The net margin in software will shift to agents
With agents expected to account for 60% of software economics by 2030, more $ will flow through agentic workloads rather than traditional SaaS.
This is also true of companies. More + more revenue will be due to agents vs. humans.