AI wrote it. Who's checking it?
Powered by Single Fin
Welcome to this week’s edition of CIOsurge!
In this edition:
Steven Singer, CIO and COO of a 110-plus-year-old food distributor, on where the human actually belongs in the loop
Our NYC Executive Forum is this Thursday. A few seats remain.
Why AI coding gains are disappearing into the review queue
The $250K cost of assuming your team already has the skills
Stay sharp. Stay ahead.
Where Is the Human in the Loop?
Food distribution has a reputation for adopting technology late, yet Steven Singer has put AI to real use as CIO and COO of a 110-plus-year-old distributor. When I asked what it took to get past the resistance, he went straight to governance.
For chat tools, that means rules and guardrails to keep hallucinations down. For agents with some autonomy, he wants a clear answer on where people fit in.
"the term is called human in the loop, right? But where in the loop is the human being? At what point in time? Is it past tense and too late where they're trying to fix an issue?"
His preference is to use AI as augmentation, where people see problems ahead of time and stay in control. Software development is his top example. A developer who can already write code on their own can save six to ten times the time by using a large language model to fill the gaps.
Holding both roles gives Steven two scorecards. The CIO in him values that developer productivity. The COO in him measures something else.
"show me the cash in the bank, show me the automation and collections, show me the things that you can do to make sure that the sales team is hitting quotas."
So he ties automation to early warning. Instead of an end-of-quarter conversation about missed numbers, reporting flags a salesperson's trend two or three weeks in, while there's still time to help them hit quota and earn their commission.
He admits the dual role creates tension, and sometimes that tension is internal.
"sometimes I'm arguing with myself, well, what's the CIO side versus what does the business actually need"
The takeaway for technology leaders: before you deploy an agent, decide exactly where the human sits in the loop, ahead of the action or after the damage. Then measure the result the way your COO would, in cash, collections, and quota.
🗓️ CIOsurge Executive Forums
Our closed-door dinners for CIOs, CISOs, and senior IT leaders. Off the record, with no vendor pitches. Seats are limited, and every registration is approved by our team.
New York | Thursday, October 8
Realizing AI ROI: Governing the Ecosystem Behind It
At Riverpark, with panelists Devon Bryan (Booking Holdings), Christopher Frenz (Rectangle Health), and Vamsi Tammana (Legend Biotech).
Request an invite
Los Angeles | Thursday, October 22
Realizing AI ROI: Governing the Ecosystem Behind It
At Perch, with panelists including Jeff Blair (CAA), Nick Reva (DoorDash), and former Dell CISO Tim Youngblood. I'll be moderating.
Request an invite
Atlanta | Tuesday, December 8
Proving ROI on AI Beyond the Pilot
At Chops Lobster Bar. Panel to be announced.
Request an invite
AI Coding Gains Are Disappearing Into the Review Queue
Bain & Co.'s 2026 Global Technology Report, based on a survey of nearly 300 senior technology leaders, found that AI coding tools help developers finish roughly 21% more tasks. The catch is that time spent reviewing AI output rose 91%. Developers are also juggling 47% more simultaneous workstreams, a sign that AI is generating work faster than teams can absorb it. Bain partner Purna Doddapaneni told CIO Dive the constraint has shifted from producing code to trusting it, since each change still needs a person to understand, test, and secure it.
This lines up with what Steven Singer told me in this week's interview. A developer who already knows how to code gets six to ten times the leverage from these tools. Someone who can't judge the output just produces more of it for someone else to check. So before you buy another coding assistant, look hard at your review capacity. Who signs off on AI-generated code, how long does it sit in the queue, and what happens when it ships without a real review? If you can't answer those three questions, your productivity gains are piling up as risk. Budget for reviewers the same way you budget for licenses.
— Zack Tembi
Skills Gaps Cost Most Companies $250K or More Last Year
A new Pluralsight report found most businesses wasted $250,000 or more last year because their teams' expertise didn't match their project ambitions. Cloud, cybersecurity, and AI projects were the most likely to be derailed, and 84% of tech leaders said skills gaps push projects over budget at least some of the time. More than three-quarters said they've watched employees finish training or earn certifications and still struggle to apply those skills on the job. CIO Dive also points to Randstad Digital research showing companies routinely spend more on AI platforms than on training people to use them.
That certification number should worry every CIO who reports training completion as progress. A badge proves someone finished the course. Whether they can do the work under pressure is a separate question. I'd rather see a team prove a skill on a live project, with a senior person reviewing the output, than collect another round of certificates. And before you greenlight the next cloud or AI initiative, run an honest skills check against the plan. Finding the gap in week one is cheap. Finding it in month six is how you end up as a statistic in next year's version of this report.
— Zack Tembi
🗞️ At A Glance
💡 CIO Spotlights
Nishit Sahay, Chief Information Officer, Altera
Joins Altera, the world's largest pure-play FPGA solutions provider, as CIO
Previously SVP and CIO at Marvell Technology, where he led corporate IT, the Data Office, and enterprise AI initiatives
Background in large-scale digital transformation, business process optimization, and technology integration
Mandate covers Altera's global technology strategy, with a focus on cybersecurity, expanding the use of data, and enterprise AI
CEO Raghib Hussain expects the hire to strengthen infrastructure, simplify how teams work, and give employees better tools
