Weekly Editorial, W44
Dayforce and SAP showed agents with no price, while Insperity and TriNet raised price per worksite employee and lost 2% and 11% of them. Buyers cannot yet budget the new unit, and the old one is shrinking under its price.
Theme 01
Dayforce, SAP and Oracle each held their event this week. A buyer leaving any of them could describe the agent, and could not budget it.
On 27 October Dayforce unveiled its Agentic HCM Platform, promising "Simple and predictable pricing, designed to give customers greater clarity and capacity to plan and scale their AI investment." The release gives no figure. Select agents are available now and others arrive "at different times through 2027."
The same day at TechEd, SAP said custom agents on its Joule Studio runtime stay free through December 2026. Its Q3 notes add that the agent builder then "becomes a premium AI feature." SAP has dated the meter and not priced it, and it shipped the spend dashboard first.
Workday reissued a rate card on 9 October with 34 metered lines and no currency price per credit. SAP gives a date with no rate. Dayforce gives an adjective. Oracle, at AI World, offers a free-to-start small-business bundle with $12,000 a year in cloud credits and no agent price.
The assistants kept opening meanwhile. Dayforce's Claude connector is in beta with general availability expected in 2027. Oracle put plugins in ChatGPT with no date. With Workday on 13 October, UKG on 20 October and ADP on 22 October, five large suites have opened to an assistant the customer already pays for, in a fortnight.
Dayforce drew the same line on autonomy as everyone else. Its time-to-pay agents flag missed punches and overdue approvals, compliance agents work "subject to customer review", and any change to workforce data through Claude needs "explicit human confirmation." None runs, changes or approves pay, and Dayforce does not say how often the agent is right.
The suite is settling into the rules and the record, with the request starting somewhere else. That makes the unit of value harder to bill, which helps explain why nobody has priced it. Until one does, a finance team can only approve these as a pilot with an uncapped later cost.
Theme 02
Three large payroll businesses reported this week. They disagree about whether the employee count is growing, and that disagreement is the useful part.
TriNet is the clearest case. Worksite employees fell 11% to about 295,000 and revenue fell 5%. Its CFO: "The impact of our first quarter repricing efforts on our WSE count and total revenues persisted throughout the third quarter." Adjusted EBITDA fell to $89m from $100m, ahead of its own forecast.
Insperity lost 2% of its worksite employees while revenue per worksite employee rose 4%, "due to higher pricing." Gross profit per employee rose 8% to $225. Its CFO said departing clients were markedly less profitable than those retained. Adjusted EBITDA more than tripled, helped by corporate staff cuts.
The common account has a flat unit forcing a price rise. TriNet says the reverse: its health plan repricing came first and the unit fell afterwards. In a bundled-benefits model the price is largely a pass-through, and volume is what remains after it. Insperity's pricing shows the same pattern without naming the cause.
ADP shows the other side. US pays per control rose 2%, and the full-year guide moved up to about 1%, which is employment still growing slowly inside the largest US payroll base. Bookings and retention guidance did not move. About $50m of the raised outlook is interest on client money.
The bundle follows the price. TriNet said on 27 October that Leave, Recruit and Spend will be included "without adding separate purchases", a reason to stay after the renewal. Neither PEO says where departing clients went, and ADP's own PEO worksite employees rose 2% to about 771,000.
Pure software vendors have no health plan to pass through, so a price rise on a flat base is a choice. Paycom and Paylocity report next week. That is the test of whether the seat is being repriced for the same reason across the market, or for two different ones.
Theme 03
Two more companies that hold the first relationship with a small business added payroll to it this week. One is a bank renting an incumbent, the other a membership for people with no staff.
On 28 October PNC announced PNC Payroll by ADP, bringing payroll, HR, time and retirement inside its digital banking platform. It is the third large US bank in 43 days after American Express and Chase, both on Gusto, and the first of the three to pick the incumbent. No launch date, price or customer count.
On 29 October Collective, a US back office for one-person businesses, announced "$100 million in new financing", and Intuit Ventures said it joined the round. Collective sells "Payroll, books and filings in one membership" at $199 or $349 a month, and says it will serve "those with employees", with no date.
A bank, a neobank, a payroll vendor, a state registry and now a solo-business membership all want to hold the record when the first employee arrives. PayFit reached new French companies through Infogreffe, and Gusto began forming US companies, on the same day last week.
The test of the membership model is that step up. An S Corp owner's salary is one employee in one state, while multi-state payroll for a growing team is a different product. Collective's claim stays a claim until it ships. Intuit's stake, in a company that markets against TurboTax, comes with no integration.
For a payroll vendor the implication is about channel. The next bank's choice of engine now says more about competitive position in US small-business payroll than a logo count does. And any buyer who arrives through one of these doors has had the evaluation made on their behalf.
Theme 04
The Budget on 28 October was the week's largest regulatory event for payroll software, and it split into what is now certain and what is still interim.
Mandatory payrolling of benefits in kind goes into Finance Bill 2026-27. Company cars and vans, fuel and medical benefits move first, from 6 April 2027, and most other benefits a year later. HMRC puts the change on "around 280,000 employers", replacing about 3.5 million P11D forms a year.
The specification exists: the RTI data item guide has been public since September, and it did not change on Budget day. The final employer guidance HMRC said would align with the Budget did not come. Its interim page is still dated 23 September, five months and one tax-year end before go-live.
PAYE by Direct Debit becomes a backstop after 80% of respondents opposed it. MTD penalty points are suspended for the first cohort. The LPC paused youth-rate alignment. In each case the published direction held, and the date or the force of it softened.
The P11D survives for globally mobile staff, for loans and accommodation, and for most other benefits until 2028, so many employers run both routes in 2027-28. Voluntary payrolling, including employer National Insurance on benefits, re-opens on 5 November and is the first practical test of readiness.
The Direct Debit outcome has a software consequence. A business that wants control of its cash from April 2029 pays before HMRC's mandate collects, which makes scheduling and approving the PAYE payment a payroll feature. It is a small line in the Budget and a real one on a UK product roadmap.
Next
Against W43. The Budget did not bring the final benefits guidance that last week's page said was due with it. Dayforce's pay agents stopped at a person, as every payroll agent did last week.