The market read Workday's backlog and skipped its agent adoption numbers

A finance systems lead who has an agent line item on next quarter's Workday renewal quote has to decide what it is worth, and this release hands her exactly two numbers to work with. Both are attach metrics, and neither is a usage metric. Start with the raise, because the arithmetic is smaller than the verb. The old full-year subscription revenue range was $9.925 billion to $9.950 billion; the new one is $9.94 billion to $9.95 billion ^[3]. The ceiling is the same number with a trailing zero trimmed. The floor came up $15 million and the midpoint moved $7.5 million, about 0.075 percent of a roughly $9.94 billion year ^[1]. The revenue beat was $10 million against a $2.64 billion consensus, some four tenths of a percent ^[2]. Earnings did the heavier lifting at $0.14 above target, around 5.4 percent ^[3], with net income up $404 million from the year-ago quarter ^[4]. The line with real growth in it is the backlog, and it splits in a way worth an hour of anyone's time. Twelve-month subscription backlog is $9.03 billion, up 14.2 percent ^[5]. Total subscription backlog is $27.4 billion, up 8 percent ^[6]. Subtract and the portion booked beyond the next four quarters is $18.37 billion now. Do the same subtraction a year ago (9.03 divided by 1.142 gives $7.91 billion near-term, 27.4 divided by 1.08 gives $25.37 billion total, leaving $17.46 billion) and that longer-dated piece grew about 5.2 percent ^[5]. Money customers have committed for the coming year is growing close to three times as fast as everything sitting behind it. Now set the agent numbers against that. Chief Executive Aneel Bhusri said AI drove more than 25 percent of new ACV and that more than 5,500 customers use at least one organic agent ^[7]. Against a base of more than 11,500 customers ^[8], that is roughly 48 percent breadth ^[6]. Here is what teams tell themselves those numbers mean: the agents are in production and doing work. Here is what they actually record: an AI component was on the paper when the deal closed, and somewhere in each of 5,500 tenants an agent has been switched on. One admin in HR ops trying a single agent once counts the same as a payroll team running them nightly. Attach at signature tells you the pitch closed the deal. The 12-month backlog is the nearest thing in this release to evidence that customers committed money after living with the product. Rebecca Wettemann of Valoir names the mechanism behind the constraint. She calls it a reverse AI moat: nobody pays for the agents without deploying the core finance or HCM platform first, so the pipeline depends on winning core deployments and then proving AI adoption on top of them ^[11]. She also says the AI pitch is strong enough to pull some customers into a platform switch, and that the agents are driving application sales rather than cannibalising them ^[13]. For the person running that switch, both are true at once: the agent demo is why the project gets funded, and the core migration is what the calendar actually holds. The flat after-hours print ^[9] also arrives after a month in which the stock jumped more than 18 percent on a report of Silver Lake buyout talks, gave it back when no deal came, and then took downgrades from analysts saying their own projections had been too optimistic ^[15]. Year to date it is down 9.8 percent while the S&P 500 is up 13 percent ^[10]. The forcing function to carry into a renewal is a two-column list. In the left column, write the number in your own systems that would look different twelve months from now if the agents were doing the work. In the right, write the number that would look identical if nobody opened them after week two.