Market Pulse
Market Pulse
When AI Models Get Cheaper, Who Keeps the Savings?
When a customer pays SAP a fixed price for each task SAP's AI handles, the model doing the work can change while the price stays put. SAP is making that swap, moving work onto cheaper models and keeping the difference. In a written answer to Reuters, it said so plainly: "The goal is healthy margins as scale grows, not an absolute reduction in AI expenditure."
Read from the buyer's side, the arrangement raises a question about what a fixed per-task price quietly rewards, for SAP and for the customers paying it.
When AI Models Get Cheaper, Who Keeps the Savings?
When a customer pays SAP a fixed price for each task SAP's AI handles, the model doing the work can change while the price stays put. SAP is making that swap, moving work onto cheaper models and keeping the difference. In a written answer to Reuters, it said so plainly: "The goal is healthy margins as scale grows, not an absolute reduction in AI expenditure."
Read from the buyer's side, the arrangement raises a question about what a fixed per-task price quietly rewards, for SAP and for the customers paying it.
Research Summaries
AgentAbstain: Testing Whether AI Agents Know When Not to Act
The authors call it post-hoc abstention: the agent takes an irreversible step, then says it should have stopped.
Deterministic sandboxes and single-run, single-turn tasks describe the tested setups, not how often deployed agents misbehave.
Research Summaries
Take-Up and Targeting: Experimental Evidence from SNAP
New enrollees had higher incomes and were less sick than control-group enrollees, even though no helper chose among applicants.
Lower friction raised participation unevenly. A rising enrollment count, read alone, would not show who remained outside.
Consumer Demand Note
Uber and DoorDash have now counted the food orders that AI agents placed for customers. These agents are software that carries out a purchase on a person's behalf. Both totals are in the hundreds. Read quickly, that sounds like weak demand for delegation. I would read it as a weak test.
Dinner already works. An app built for the job gets food to the door in a few taps, so an agent has almost nothing to rescue.
The need for help shows up where the existing path breaks down: the insurance appeal left unfiled, the refund nobody chased, the benefits form a daughter completes after her own workday. Nobody has measured delegation there yet.
