Microsoft 365 went dark. Could you still operate?

Powered by Single Fin
Welcome to this week’s edition of CIOsurge!
In this edition:
Why token usage showing up in developer reviews should make CIOs pause
Where to find me and your peers over the next few months
Microsoft 365 went dark for 24+ hours. Could your business still operate?
99 flaws in 90 hours, inside devices nobody can patch
Three open leadership roles, including a CISO and a CIO seat
Stay sharp. Stay ahead.
Token Usage Is Showing Up in Developer Reviews. That Should Make CIOs Pause.
Last Wednesday I moderated The Economics of AI, a Mind and Machine x SMU Cox Marketing special edition at SMU’s EY Gallery. Amy Chaney, SVP, Technology COO at Citi, gave the keynote. My panel was Todd Smith (Taho Labs), Bryan Tharpe (Slalom), Daryan Dehghanpisheh (Palo Alto Networks), and Jeff Calusinski (Reknew, former CTO at USAA).
We covered a lot: who owns the AI budget, hard-dollar ROI versus “time saved,” agents running up bills nobody approved, and when a frontier model is overkill. But the point that stuck with me came from somewhere else: how engineering leaders are starting to review their developers.
For years, the signal was GitHub activity: commits, pull requests, code published. Now leaders are looking at AI token usage alongside that, and sometimes instead of it.
That should get a CIO’s attention, because one number is being asked to do two jobs.
Token usage is a cost line. Every prompt and agent call shows up on an invoice. It’s also being read as a productivity signal: the engineer using more AI must be shipping more.
Those two readings pull in opposite directions. Treat token spend as a cost to crush, and people hide their usage or go back to slower tools. Treat it as a performance score, and people game it. A messy prompt, a retry loop, or an agent calling three other agents can burn a lot of tokens and still leave you with code you wouldn’t put in production.
Neither tokens nor commits tell you whether the work was sound, maintainable, or even necessary.
My take: use token usage as an input, not a scorecard. Put it next to what actually shipped, defect rates, and rework. And be clear with your engineering leaders about what you’re rewarding. Commits alone measure yesterday’s workflow. Tokens alone mean paying for a lot of motion that isn’t progress.
One thing to do next quarter: ask your engineering leaders whether token usage has quietly become part of the developer scorecard. If it has, make sure quality is in there too.
— Zack
🗓️ CIOsurge Executive Forums

Join me and your peers at our upcoming CIOsurge Executive Forums: closed-door, off-the-record dinners for senior CIOs, CISOs and IT leaders, with no vendor pitches.
Realizing AI ROI: Governing the Ecosystem Behind It
Atlanta · Dec 8, 2026 · 5:30 PM ET
Sunday, 4am: Your AI Agent Has Been Hijacked. How Do You Respond?
New York · Jan 28, 2027 · 5:30 PM ET
CIOsurge Executive Forum, Miami Beach
Feb 11, 2027 · 5:30 PM ET
CIOsurge Executive Forum, San Francisco
Feb 25, 2027 · 5:30 PM PT
CIOsurge Executive Forum, Los Angeles
Mar 4, 2027 · 5:30 PM PT
See all upcoming dinners on the CIOsurge Events page.
Microsoft 365 Went Dark for 24+ Hours. Could Your Business Still Operate?
Microsoft 365 suffered a global outage that began Thursday afternoon and ran into Friday, cutting off access to many services and stopping work for businesses around the world. Microsoft told CIO Dive the outage was not a malicious attack and wasn’t caused by AI agents. It blamed a technical issue from an earlier network capacity upgrade that surfaced after a recent network change, and it plans to publish a post-incident report.
In a research note, Gartner said the outage exposes how fragile provider concentration is. Microsoft 365 has become the “connective tissue” tying together corporate identity, endpoint compliance, security and AI workflows, so when it degrades, an isolated infrastructure failure turns into an enterprisewide disruption. Gartner advised CIOs to add monitoring that catches interruptions, set up out-of-band communications and fallbacks, audit single-vendor concentration in areas like email and authentication, keep independent backups, and stress-test AI workload dependencies before scaling.
The question in the headline is the one I’d put to my own team this week. Most of us bought Microsoft 365 for email and documents, and over time it became the place where identity, device compliance, security and now AI workflows all meet. That makes it a single dependency with a lot hanging off it. Gartner’s checklist is a good place to start, and I’d begin with two items: an out-of-band way to run an incident when Teams and Outlook are both down, and an honest count of how much of authentication runs through one vendor. Will Lassalle judges vendors by how things look three months after go-live. A week like this is a good time to judge our own fallback plans the same way.
— Zack Tembi
AI found widespread IoT flaws more quickly and cheaply than humans could
Security firm Forescout used a specialized AI harness to find 99 vulnerabilities in 90 hours, work its researchers estimated would have taken nine months without it. The team tested 10 popular TCP/IP stacks, most of them open source, including Contiki for low-power devices like street lights, lwIP, and wolfIP for real-time operating systems in safety-critical devices.
Sixty-six of the flaws were new, and roughly half were memory-corruption bugs. Forescout said many foundational flaws had survived for years in code that its own researchers, other experts and the vendors had already reviewed. It also warned that unsupported legacy products are everywhere, and that cheap AI-assisted research can now find dangerous flaws in them that no vendor will be able to patch.
The speed is the headline, but the part that worries me is the patch that will never come. Most of us have embedded devices in buildings, plants and networks that run code like this and sit outside any support contract. This quarter, I’d ask the team for three lists: which connected devices we own, which still have a vendor willing to fix them, and which don’t. The ones that don’t get segmented and monitored now, and end-of-support dates go into every new hardware contract we sign.
— Zack Tembi
🗞️ At A Glance

💡 CIO Spotlights
Kobi Sivilya, CISO and CIO, Checkmarx
Appointed to a dual CISO and CIO role, leading Checkmarx’s corporate security and IT functions
Brings more than 20 years of experience across cybersecurity and technology
Joins from Outbrain, where he was CISO and VP of IT and ran both security and enterprise technology
Previously VP of Security Engineering at Fireblocks, building security engineering for a blockchain platform that protects digital assets for financial institutions
His stated focus is making internal technology and security support how teams work today while keeping the business ready for what comes next
🤝 Jobs
See every open CIO, CISO and IT leadership role on the CIOsurge Jobs board.





