Reading Room

Cloudflare announced a preview of Clairic, a new kind of AI model built only to pull structured data out of unstructured text (an extraction model), which it says works faster, cheaper, and more accurately than a general purpose AI model asked to do the same job. Turning messy text into clean, usable data is one of the most common AI tasks there is, so a cheaper specialized tool here could cut costs for any project that leans on it often.

He should watch this because it is an early sign that AI customer service bots might soon face real rules, relevant if any of his own projects ever puts a bot in front of a customer.

A post citing the news outlet Semafor says New York City wants every AI customer service phone line to offer a one touch way to reach a human being, instead of looping callers through endless menus. The post making the rounds does not name the specific agency or bill behind the idea, and no vote or deadline has surfaced yet.

The debate the poster raised: does a rule like this kill AI customer service outright, or does it just force companies to build bots that actually solve the caller's problem instead of stalling them? There is no detail yet on enforcement, or whether it would cover phone lines only or chat too.

Worth a second look once an actual bill or filing surfaces. Today's version of the story is one step removed from any primary source.

He watches whether money companies spend on enterprise software, a growing share of it AI flavored, actually turns into profit rather than just vendor talk. IBM's third quarter numbers, out today, offer one data point.

IBM reported revenue of $17.1 billion for the quarter, up 5 percent from the same quarter last year and above the $16.82 billion Wall Street had expected. Profit (net income, what is left after every cost) came in at $2.19 billion, up 26 percent from $1.74 billion a year earlier.

Software, the part of IBM's business that sells programs and subscriptions, brought in $7.8 billion, up 8 percent. Infrastructure, the part that sells hardware and the systems to run it, grew too, up 4 percent, though today's report did not give a dollar figure for that segment.

What is missing is any breakdown of how much of the gain came specifically from AI related products versus IBM's older software and hardware lines. That makes this a positive signal for enterprise tech spending in general, not proof that AI itself is the driver.

SAP said its cloud subscription revenue rose 25 percent from a year ago to €6.67 billion, ahead of the €6.57 billion analysts expected, while total revenue grew 12 percent to €10.27 billion. Its backlog of signed cloud contracts not yet booked as revenue grew 26 percent to €24.1 billion. Another sign that big companies are actually paying for cloud and AI tools, not just talking about them.

AT&T's in-house legal team, called LegalEdge, is working with OpenAI to make it easier for lawyers to find and use legal knowledge, rethinking how incoming requests get sorted and handled. A real example, inside a huge company, of AI changing day to day legal and office work, the kind of case study worth stealing ideas from.

Walkouts protesting AI related job losses hit Denver, Golden, and Fort Collins as part of a wider nationwide wave of protests, according to Denver7. A sign that public pushback against AI driven layoffs is turning into an organized, multi city movement rather than staying online.