Weekly Editorial, W33
Silver Lake is in talks to take Workday private at about $43bn, and Zalaris filed results the next morning showing its own change of owner has turned a positive operating profit into a negative one. Dayforce completed the same move in February and its accounts are no longer public at all.
Theme 01
A rumour, a set of accounts and a completed deregistration. Read together they describe a direction and put a number on what it costs.
On 13 August Reuters reported that Silver Lake is in talks to acquire Workday, whose market value before the report stood at about $43bn. The talks have run for months, are ongoing, and carry no guarantee. Silver Lake could bring in further investors. The stock closed up 17.78% at $206.45 after an intraday move near 25% forced a trading halt, its best day in a decade. Neither party has confirmed, denied or commented on terms.
Twelve hours later Zalaris filed its half-year report and, without intending to, priced the exercise. Revenue was NOK 344.2m, up 0.5% in constant currency. Adjusted operating profit was NOK 37.6m. Reported operating profit was negative NOK 14.6m against positive NOK 36.6m a year earlier, a swing of NOK 51.2m that the company attributes to advisory and share-based payment costs from its acquisition by Kona BidCo. Chief executive Hans-Petter Mellerud called it "stable underlying performance during a period of strategic transition."
Dayforce is the completed case. It terminated its registration in February after its own take-private, and its headcount, retention and segment economics are no longer obtainable from any public filing.
CFO buyer. The consequence that reaches a customer is not the purchase price, it is that audited disclosure stops. A buyer evaluating a platform whose owner has just taken it private loses the ability to check whether the vendor is financially healthy, at precisely the moment the vendor's ownership incentives change. Dayforce has already demonstrated it. Workday is currently among the most transparent large vendors in this market, and if it follows, that visibility ends. Nobody diligences the continued existence of a vendor's accounts, which is why this arrives as a surprise rather than as a risk anyone priced. Platform vs product. Workday is the largest pure-play HCM company and one of very few running HR and finance on one data model. A financial sponsor is buying an installed base and a switching cost, not a growth story, and such purchases are typically funded with debt serviced from that same installed base. The question a buyer should carry into a renewal is what that structure implies for price and for research investment. Stated limit. The Workday item is a sourced report, corroborated but unconfirmed. Nothing links Silver Lake's interest to Zalaris's transaction costs beyond both being take-privates in the same market, and that is the observation rather than a mechanism.
Theme 02
The company that owns the largest job board in the United States published the cleanest evidence this market has produced on where growth comes from when volume stops.
Recruit Holdings, which owns Indeed, reported that the total number of US job postings fell about 4% year on year while US revenue rose 30.0% to a quarterly record of $1.64bn. All of the growth came from revenue per posting, which rose 35%. The company then supplied its own comparison: the previous revenue record, $1.61bn, came off a posting base roughly 57% larger that was itself growing 24%, at revenue-per-posting growth of 1%.
Guidance assumes the contraction continues and still raises the full-year US revenue growth outlook from 13.6% to 25.1%. Read alongside ADP's guidance of flat-to-1% employment growth inside its own installed base, two independent measures of US labour volume now point the same way while the money attached to each unit rises.
CFO buyer. The cost of filling a role through the dominant US posting channel is rising at 30 to 35% a year while hiring volume falls. That is a talent-spend line moving in the opposite direction to headcount, and it appears in no per-employee software comparison. A finance team modelling recruitment cost as a function of hiring volume will be wrong in both directions at once. Platform vs product. The prevailing assumption is that a vendor whose countable unit stops growing must expand into adjacent products. Recruit's unit did not merely flatten, it shrank, and the company grew 30% without entering an adjacent category and without an acquisition, by moving customers up a feature tier on the unit it already had. Raising the price of a shrinking unit is a live route and it is currently underweighted in how this market thinks about growth.
Theme 03
Britain's first mandatory digital filing deadline produced its first real number this week, and it arrived alongside a consultation to collect a second tax through payroll.
HMRC reported that more than 436,000 sole traders and landlords filed their first quarterly update under Making Tax Digital for Income Tax, out of over 570,000 signed up. Two statements set the enforcement position: "In 2026 to 2027, there are no penalty points for late quarterly updates", and "From September 2026, HMRC will begin signing up customers who should be using MTD... but have not yet done so." Roughly three quarters of the registered population filed, and registration is itself the smaller number, which is why the authority is preparing to enrol the remainder itself.
Separately, HMRC's consultation on timely payments in Self Assessment sets out reforms for taxpayers with PAYE income "who will be required to pay their forecasted ITSA liability in-year" from April 2029, following an announcement at Budget 2025. About 12 million file Self Assessment and roughly 7 million of those also have PAYE income. Self Assessment is not a payroll tax, and collecting a forecast of it through PAYE makes the payroll run the collection mechanism for a liability the employer does not compute and cannot verify.
Multi-regional regulatory operator. The operative fact for anyone selling readiness against the August deadline is that it was not an enforcement gate. Penalty points are disapplied for the whole year and the date that carries consequence remains the annual return in January. The commercially significant item is the September auto-enrolment: a defined population acquires a filing obligation by letter, without having chosen software first. CFO buyer. April 2029 is far enough away that nothing needs building and close enough that the design is being fixed now, in a consultation that closed in early August, which most payroll vendors were not part of. The direction is consistent across both items: obligations keep landing on the employer's administration layer rather than in the withholding calculation. Construction vertical. Sole traders earning above the threshold describes a large part of the UK subcontractor base, which already operates inside the Construction Industry Scheme and now carries quarterly digital filing alongside it.
Theme 04
The largest agentic release of the week measured itself in coverage rather than capability, and the pattern is now consistent enough to be the market's default.
Oracle added thirteen agentic applications and AI agents across talent management, spanning role design, skills inference from work-system data, job requirement generation, learning production, manager coaching, career planning and workforce supply-and-demand analysis. On governance the release offers a single clause, that the agents act "while keeping people in control of business-critical decisions." There is no availability date, no pricing, no adoption figure and no accuracy figure anywhere in it.
Thirteen agents is a completeness claim, and completeness claims are answered by counting rather than by measuring. Every one of these agents makes a judgement: inferring an employee's skills, drafting the requirements a role is hired against, telling a planner where the gaps are. Not one measure of how often those judgements are correct accompanies them. Elsewhere this week another vendor passed ten days beyond the general-availability date it announced for an agent built specifically to catch payroll errors, still without publishing a detection rate, an error rate, an evaluation method or a price.
CFO buyer. A buyer comparing this against a rival release can compare surface area and nothing else. There is no price to negotiate, no availability date to plan a programme around and no accuracy figure to underwrite a business case, so the entire burden of establishing whether any of it works transfers to the buyer's own evaluation. Platform vs product. The governance pattern across the market is now consistent: vendors control what an agent is permitted to touch and say nothing about whether it is right. Those two properties are independent. An agent can be perfectly scoped and routinely wrong. The first vendor to publish a competence measure alongside its method will not be matched quickly, because publishing one requires having measured it, and the uniform silence across the tracked set reads more like nobody has than like everybody is hiding a good number.
Next