What changed this week — and what it means

So what · If you run Workday, this is the first agent wave aimed at your recruiting and payroll teams, and the questions to ask now are commercial and governance ones, not feature ones. Ask your account team three things: what early-adopter access costs and commits you to, how agent runs will be metered after general availability, and whether every candidate-facing step can be forced into human review. That last one will not stay optional: California's SB 947 requires a human to corroborate automated discipline and termination decisions from July 2027, and the EU AI Act's high-risk rules for employment AI apply from December 2027. The C1.ai deal is the sleeper story. Agents will need onboarding and offboarding just like people, so bring IT security into your agent pilot now rather than after go-live.

So what · Start with an inventory, because these laws reach beyond the HRIS: Morgan Lewis notes they cover general-purpose tools managers use on their own. For SB 947, map every point where an automated score or recommendation feeds discipline or termination, then write the human-corroboration step and the written-notice template well before July 2027. For SB 951, add one question to every California reduction-in-force checklist from January: did AI or automation substantially cause this? And check your monitoring tools for sentiment, "engagement" or emotion features now. AB 1883 makes them a liability on January 1, and that includes interview notetakers whose transcripts are later scored for tone.

So what · Two states, one pattern: the layoff notice is becoming an AI disclosure, and "the vendor's model decided" is no longer a defense. Decide now who answers the "did AI drive this?" question in a reduction in force, whether HR, legal or finance, and on what evidence, because the answer will be on file and quotable. The cheapest protection is the one Connecticut explicitly credits: documented anti-bias testing of every screening tool you use. Ask vendors for their latest results and the tool details the 2027 notices will require, in writing, at your next renewal.

So what · Few layoffs plus near-zero hiring is a frozen market, and it changes the AI conversation inside your company. With fewer exits and fewer hires, the productivity leaders expect from AI has to come from redesigning existing roles, not from attrition or replacement. That makes it an internal-mobility and reskilling brief for HR, not a headcount one. Keep an eye on the monthly AI share of cuts: it has fallen out of the top three reasons two months running, which is useful evidence for any board conversation that treats AI layoffs as inevitable.

So what · This is the number that ties the rest of this edition together. Workday's autopilots, California's corroboration rule and Connecticut's disclosure rules all assume someone owns the agent, and in three of four leadership teams nobody has agreed who does. Put a one-page agent accountability charter on your next leadership agenda: for each agent in use or in pilot, name the business owner, the human reviewer, the escalation path and who answers when it is wrong. HR is the natural author, because it already runs the equivalent for people. Gartner's point about judgment calls is the same one in our analysis of the judgment premium.

So what · For buyers, this is a market of well-funded incumbents and thinly funded newcomers, and your diligence should reflect that. When a startup is on your shortlist, ask about runway, the next funding milestone and what happens to your data if it is acquired or shuts down; the RFP question bank in our ATS comparison covers the data-exit question. When an incumbent is, remember why capital is flowing: investors are betting on workforce data as the context AI agents need. Your HR data is what makes their agents valuable, so negotiate data-use and model-training terms as carefully as price.