IT STRATEGY

Getty Images

As anyone who’s ever tried to ask an over-caffeinated coworker about their weekend knows, a simple, open-ended question can sometimes lead to the most unexpectedly lengthy responses.

And when the respondent is a large language model (LLM), lengthy responses can translate into serious budget issues, especially if an IT team is querying their AI tools multiple times per day.

For example, a software developer might ask their LLM of choice to help build an internal chat app from scratch, which could lead to excessively long outputs as the AI attempts to solve several massive architectural challenges. And since most AI tools charge by the token, epic-length outputs (and inputs) can quickly strain an AI budget.

(A token is the smallest data processed by a large language model. One token corresponds to roughly 4 English-language text characters, according to OpenAI, or roughly 100 tokens is equal to 75 words. GitHub in April announced metered, per-token pricing; ChatGPT’s API usage is charged by the number of tokens used; and Anthropic’s models have token-based cost structures.)

How IT pros are keeping AI outputs short and sweet.—BH

      

Sponsored By Harness

Engineers don’t really write code anymore. Now, they’re shifting mainly to directing AI instead (AI coding agents and agentic pipelines, we’re lookin’ at you).

Want to hear directly from the people navigating this shift? Harness is putting a very familiar face on the stage that longtime fans of the Brew will definitely recognize. Yep, it’s none other than Alex Lieberman, entrepreneur and co-founder of Tenex, Morning Brew (!!!), and storyarb.

Save your seat to the free Harness virtual conference on Sept. 30. Tune in for Alex’s session, From Code to Factory: Building Software in the Age of AI, to hear his thoughts on the software factory concept taking over engineering orgs in 2026.

Register for free.

IT STRATEGY

Patrice Williams-Lindo

As employers turn to AI tools to sift through a flood of résumés and find qualified candidates for IT jobs, many are designing “knockout criteria” to help determine who gets an interview.

But establishing that criteria isn’t just a job for HR, according to Patrice Williams-Lindo, CEO and founder of workforce and AI governance advisory Career Nomad: “I think that HR owns the why. IT owns the how.”

For IT pros grappling with that “how,” Williams-Lindo highlighted three important strategies that can prevent AI and human hirers from accidentally filtering out top candidates.

Audit your rejections—not just your hires. Run what’s known as the four-fifths test, which analyzes if the selection rate for a demographic (including race, sex, or ethnicity) reaches at least 80% for the highest-rate pool. For example, if 25 out of 50 male applicants are hired (a 50% selection rate), and 20 out of 50 female applicants are hired (a 40% rate), the selection rate is 80% (i.e., 40% divided by 50%), which passes the test.

The art of the knockout question.—BH

      

HARDWARE

Getty Images

AI is changing the tech stack—and for Apple, that’s an opportunity to enter markets traditionally dominated by PCs.

Enterprises rushing to deploy AI solutions are hitting the reality that building infrastructure takes time and money—and hardware isn’t always available. Delays can impact project timelines and even an organization’s health.

Fixit. Many IT teams are turning to Apple hardware to pick up the slack when it comes to the on-prem training and use of AI models. Matt Vlasach, SVP of product and solutions engineering at Jamf, told IT Brew that a number of factors including token cost, data sovereignty, and the capabilities of Apple’s chipset are at play in moving to the company’s hardware.

“It’s created a lot of interest in leveraging Apple devices as a great tool for proving out local device inference and exploring on-device models to figure out what sort of workloads can be handled locally for the sake of cost reduction, as well as for the sake of data sovereignty,” Vlasach said.

One enterprise built an AI cluster for how much?—EH

      

Sponsored By Tines

Wild, wild code. AI allows for faster builds, but unmanaged “wild” code can open you to new risks. Join Tines for a session on how IT can build the foundation for AI-powered innovation. Learn how getting governance right can unlock meaningful AI adoption for your teams. Register here.

patch notes

Francis Scialabba

Today’s top IT reads.

Stat: 22%. That’s the proportion of UK companies that have been hit with an AI-related breach in the past year. (ITPro)

Quote: “They have to be able to translate business needs to technology teams and explain how technology creates value for the business so that everyone can understand and trust that information.”—Lenka Pincot, chief of staff to the CEO at the Project Management Institute, on the qualities that make a good project manager (CIO)

Read: How employees are circumventing AI mandates in the workplace. (Aftermath)

^{*A message from our sponsor.}

Guoya/Getty Images

Learn what IT Brew’s State of the Industry Survey revealed about IT staff and their knowledge on AI policies.