Q2 fiscal 2027 is the quarter a filing first puts a dollar on AI at Salesforce, and it is not in revenue: the 10-Q attributes of unrealized gains to the investment in Anthropic, of the quarter's revenue of , carried in other income and set by what investors paid for the lab's shares. The ledger records it as a reported non-operating gain and leaves it out of flow totals.
The product metric was recharacterized in the quarter it grew. Agentforce is stated above of annual recurring revenue, from , on a wider definition: from this quarter the figure counts Slackbot and Headless, which were outside it before, so the increase is part growth and part reclassification, and management does not split it. The release also adds a sentence that the metric may be updated for new products, feature offerings or acquired technologies, which makes further recharacterization a stated policy. Informatica Cloud annual recurring revenue, stated in both prior quarters, is withdrawn: the combined metric that contains it is still given and the component is not, so the data platform inside the combined can only be bounded. The ledger reads Agentforce revenue for the quarter at , on the wider definition. Premium editions gained levels: of sales and service users upgraded, at a premium of to .
On cost, the filing names generative AI for the first time: cost of revenues and research and development both rose in part on spend on hosting and generative AI technologies. Cost of subscription and support revenues ran above its prior-year share and research and development above its own. Management stopped giving the token count it gave in both prior quarters. The savings channels show a split: general and administrative expense was flat in dollars, relative to its prior-year share of revenue, while engineering, where the prior quarter's claim was headcount held flat, saw its line rise faster than revenue.
The toll management names remains bounded at zero by its own account: seats grew and attrition is approximately . The channel that moved against the company is the small one: professional services excluding Informatica is lower than a year earlier, which the filing puts down to demand for large engagements and the CEO's remarks tie to tools that do the work of service staff. Apart from the Anthropic gain, every dollar size here is the ledger's inference around management's run-rates and shares; channels measured only by counts, the token bill with no stated volume, and the attrition toll management denies are left unsized.
Sized channels against the income statement, Q3 2026
13 of 19 channels sized
Each blue mark is one channel's dollars for the quarter; a bar is the range of an estimate. Grey marks are the company's reported lines. The distance between them is the point: how large the AI channel is next to the line it sits in.
Total revenuesreported line
$11.35bn
Total operating expensesreported line
$6.37bn
Sales and marketingreported line
$3.86bn
Total cost of revenuesreported line
$2.65bn
Research and developmentreported line
$1.69bn
General and administrativereported line
$725mn
Restructuringreported line
$94mn
Unrealized gains on the investment in Anthropicrevenue in · reported in the filing
Building and deploying Agentforce: engineering and forward-deployed engineersspend · our inference
Agentforce subscriptions and credits (Agentforce ARR)revenue in · our inference
Data Cloud subscriptions counted in the AI and data ARR metricrevenue in · our inference
Informatica Cloud ARR counted in the AI and data ARR metricrevenue in · our inference
Premium edition upgrades with embedded AI (Agentforce One Edition, A4X)revenue in · our inference
Flex Credits and agentic enterprise licence agreementsrevenue in · our inference
Engineering headcount held flat by AI coding toolscost displaced · our inference
Core seat and edition revenue pulled through by agent adoptionrevenue in · our inference
Subscriptions sold to AI labs and AI-native companiesrevenue in · our inference
Professional services revenue displaced by agent-assisted implementationtoll · our inference
AI coding tools and generative AI used in engineeringspend · our inference
Pipeline generated by the company's own sales agentsrevenue in · our inference
$100k$1mn$10mn$100mn$1bn$10bn$100bn
AI channel, dollars for the quarter low to high of an estimate reported lineLog scale: each gridline is ten times the one before.
New money and old money, Q3 2026
7 new10 expanded2 relabelled
Each channel is tagged once for whether its money existed before language models, from the company's annual report and call at the start of the period. A bar splits one flow's sized dollars by that tag. The incremental total is the part that would not be there without the models: a new channel counts in full, an expanded one only for what AI added, a relabelled one at zero.
Flow, sized total
Split by novelty
Incremental total
Paid for AI$169mn to $889mn sized
new $450k to $46mnexpanded $169mn to $844mn
Incremental total $450k to $889mnpoint $6.1mn$422mn in 1 channel has no traced baseline
Cost displaced by AI$0 to $202mn sized
expanded $0 to $202mn
Incremental total $0 to $202mnpoint $84mn
Revenue arriving through AI$826mn to $1.32bn sized
new $304mn to $458mnexpanded $248mn to $538mnrelabelled $275mn to $325mn
Incremental total $304mn to $996mnpoint $429mn$297mn in 1 channel has no traced baseline
Cost imposed, or revenue lost, by others’ AI$0 to $22mn sized
new not sizedrelabelled $0 to $22mn
Incremental total $0
The sized total counts every channel the company credits to AI, including relabelled money that existed before language models and the ledger's own estimates for it. The incremental total counts relabelled channels at zero. A flow is split by layer where its dollars sit at more than one: end use, compute sold to builders, and hardware. The same dollar can be a buyer's spend, a cloud's revenue and a chipmaker's revenue, so the layers are never added together.
Paid for AI3 channels · $169mn to $889mn sized · $450k to $889mn incremental · 1 not sized
vendor bill
Model tokens and AI compute behind Agentforce
Not sized
No token volume is stated this quarter; the former estimate converted agentic work units, , to tokens at an assumed ratio, so the priced unit's volume is the ledger's, not a measured volume with a period (the methodology rule for volume times price). The filing's only cost statement names generative AI together with hosting services (claim c45), with nothing to separate AI's part.
Matched line moved : the whole line, not this channel.
described, no sizedisclosure: direction only· motive: product-defensive· before LLMs: new
For the first time the filing names generative AI: cost of revenues rose in part on spend on hosting services and generative AI technologies (claim c45). No amount is given. The token count management gave in both prior quarters is not given; the state stays directional, now on the filing's attribution rather than a count of AI work. The channel is unsized: no token volume is stated, and the filing names generative AI only beside hosting. The former estimate, , is no longer the size. The counterparty is mixed: third-party model labs paid by the token, and the cloud and data center capacity behind the company’s own models.
Evidence: 5 quotes, 3 figures, 4 confounds, 4 from before coverage
What the company pays model providers and hosts to run its agents: tokens bought from third-party labs and compute for its own models. It sits inside cost of subscription and support revenues, which the filing does not split by purpose.
Why this motive
Carried from the prior quarter: the cost is absorbed in the margin structure. The filing now confirms it is in cost of revenues (claim c45).
Before LLMs: new
Token purchases from model providers did not exist before LLMs; the anchor 10-K already lists the large language models behind the AI offerings among the third-party technology the company relies on, and earlier Einstein features ran on its own models. Cost of subscription and support revenues was in fiscal 2024, about a quarter.
“The hardware, software, data and cloud computing platforms that we rely on, including, for example, the large language models leveraged in our AI offerings, may not continue to be available at reasonable prices, on commercially reasonable terms or at all.”
“Cost of subscription and support revenues primarily consists of expenses related to delivering our service and providing support, including the costs of data center capacity, certain fees paid to various third parties for the use of their technology, services and data, employee-related costs such as salaries and benefits, and allocated overhead.”
“The way we'll run is, we'll run like a model tournament, because one of the things everybody has to watch out is, it's great, but what about the cost to serve?”
Agentic work units delivered in the quarter · 2026-CQ3
Cost of subscription and support revenues · 2026-CQ3
Cost of subscription and support revenues, prior-year quarter · 2025-CQ3
Reported line it is matched to
Cost of subscription and support revenues was against , above the prior-year share of subscription and support revenue. The filing attributes the increase in cost of revenues to service delivery expenses, including hosting and generative AI technologies, and to amortization of acquired intangibles.
2026-CQ3: 2025-CQ3: 2026-CQ3: 2025-CQ3:
What else could explain it
line composition: The filing names hosting, generative AI and acquired-intangible amortization together as causes and does not split them.
acquisition: Informatica's costs and amortization are in the line this year and not last.
other: Part of the workload runs on the company's own models (claim c47), which cost compute rather than a provider fee.
other: The filing names spend on hosting services and generative AI technologies together as a cause of the rise in cost of revenues; nothing separates AI's part.
Quotes
“Cost of revenues increased in absolute dollars, and by one percent as a percentage of total revenues, for the three and six months ended July 31, 2026 compared to the same period a year ago, primarily as a result of an increase in service delivery expenses, including spend on hosting services and generative AI technologies, as well as amortization of intangible assets acquired through business combinations.”
“We intend to continue to invest additional resources in our AI, agentic and cloud services to allow us to scale with our customers and continue to evolve our security measures.”
“We are combining Claude's extraordinary intelligence and reasoning with Salesforce's trusted data and applications and workflows and business rules and governance.”
Q1 2026direction only · described, no size · product-defensive
Q2 2026direction only · our inference · product-defensive · $2.6mn to $69mn
Q3 2026direction only · described, no size · product-defensive
vendor bill
AI coding tools and generative AI used in engineering
0% to 0.4% of the quarter’s revenue
Incremental total: counts in full.
Matched line moved : the whole line, not this channel.
our inferencedisclosure: direction only· motive: efficiency· before LLMs: new
The filing attributes the rise in research and development as a share of revenue to employee costs and to spend on hosting and generative AI technologies (claim c49): a direction from the filing where the prior quarter had only the CEO's description. The call does not mention the company's own coding tools. The size, , keeps a seat-based floor and point and takes as its ceiling the amount by which the line exceeds its prior-year share of revenue.
Evidence: 1 quote, 2 figures, 2 confounds, 1 from before coverage
The outside bill for the AI coding tools and model usage of the company's own engineers. It sits inside research and development; management names the vendors and no amount.
Why this motive
Carried from the prior quarter: tools bought to hold engineering headcount flat.
Before LLMs: new
No outside AI coding-tool bill is in the anchor: on the anchor call the engineering chief describes the company's own code models in use by its engineers. Research and development was in Q4 fiscal 2024.
“But the other use cases which we are going to see, and in fact, I've rolled out our own code LLMs in our engineering org, and we are already seeing minimum 20% productivity.”
Research and development, prior-year quarter · 2025-CQ3
Reported line it is matched to
Research and development was against , above the prior-year share of revenue. The filing attributes the increase to employee-related costs and to spend on hosting services and generative AI technologies.
2026-CQ3: 2025-CQ3: 2026-CQ3: 2025-CQ3:
What else could explain it
line composition: The filing names employee costs, hosting and generative AI together; generative AI in research and development may include model development as well as coding tools.
acquisition: Acquired engineering teams are in the line this year and not last.
Quotes
“Research and development expenses as a percentage of total revenues during the three months ended July 31, 2026 increased by approximately one percent compared to the same period a year ago due to increased employee-related costs, as well as spend on hosting services and generative AI technologies.”
Q3 2026direction only · our inference · efficiency · $450k to $46mn
engineering
Building and deploying Agentforce: engineering and forward-deployed engineers
1.5% to 7.4% of the quarter’s revenue
Incremental total: counts at zero at the point and in full at the high end, with no traced baseline.
our inferencedisclosure: described· motive: offensive· before LLMs: expanded
Research and development was against , and the filing now names generative AI spend among the reasons the line rose (claim c49). The company agreed to acquire an AI service agent platform for (claim c51), which is capital rather than expense. The CEO says the new interface layer reduces the need for forward-deployed engineers, the investment the CFO named in Q4 fiscal 2026 (claim c53). The size, , is the ledger's assumed share of the line.
Evidence: 5 quotes, 3 figures, 2 confounds, 2 from before coverage
The company's own spend on developing its AI products and getting customers into production: the AI share of research and development and the forward-deployed engineers the CFO names. No split of either line is disclosed.
Why this motive
Carried: investment in support of a priced product. This quarter adds a new interface layer and an agreement to acquire an AI service agent company (claims c52, c51).
Before LLMs: expanded
Research and development funded the applications and earlier machine-learning features before agents: the line was in Q4 fiscal 2024, and the anchor 10-K says the company began adding AI resources at the end of that year. Forward-deployed engineers for agents are an addition of fiscal 2026 and 2027. The size is the whole of an activity that existed before.
“However, at the end of fiscal 2024, we began to invest in incremental AI resources to accelerate further growth and as a result our research and development headcount increased by five percent during fiscal 2024.”
Research and development, prior-year quarter · 2025-CQ3
Agreed price for Fin, an AI service agent platform · as-of 2026-07-31
What else could explain it
line composition: Research and development funds every product; the AI share is assumed.
acquisition: Acquisitions add engineers to the line and substitute purchased products for internal build.
Quotes
“Research and development expenses as a percentage of total revenues during the three months ended July 31, 2026 increased by approximately one percent compared to the same period a year ago due to increased employee-related costs, as well as spend on hosting services and generative AI technologies.”
“In June 2026, the Company entered into an agreement to acquire Intercom, Inc. (“Fin”), a customer agent platform providing autonomous, end-to-end AI service agents, for approximately $3.6 billion in cash, and subject to customary purchase price adjustments.”
Cost displaced by AI3 channels · $0 to $202mn sized · $0 to $202mn incremental · 2 not sized
customer support · cheap to verify
Customer support handled by Agentforce on the help site
Not sized
The measures are a cumulative count of conversations and a resolution share by count, applied to no reported line (claim c54); counts size nothing (the methodology rule for counts).
described, no sizedisclosure: direction only· motive: efficiency· before LLMs: expanded
The help agent has passed customer conversations with resolved autonomously (claim c54). The basis changed: the prior quarter counted inquiries handled autonomously, this one counts all conversations and a resolution share, so the two counts cannot be differenced. No cost or headcount is given, so the state stays directional on counts of AI work; support cost sits inside . The channel is unsized: its measures are counts. The former estimate, , is no longer the size.
Evidence: 2 quotes, 3 figures, 3 confounds, 2 from before coverage
The company's own customer inquiries resolved by its agent on its help site and phone line instead of by support engineers. The cost it would displace sits inside cost of subscription and support revenues.
Why this motive
Carried from the prior quarter: work moved off the staffed line, now with a stated share resolved without a human (claim c54).
Before LLMs: expanded
Support was delivered by the company's own support group over the web, telephone and email at the anchor, with the cost inside cost of subscription and support revenues: in fiscal 2024, about a quarter. The size is the change AI made, not the whole line.
“Cost of subscription and support revenues primarily consists of expenses related to delivering our service and providing support, including the costs of data center capacity, certain fees paid to various third parties for the use of their technology, services and data, employee-related costs such as salaries and benefits, and allocated overhead.”
“Our global customer support group responds to both business and technical inquiries about the use of our products via the web, telephone, email, social networks and other channels.”
Q1 2026described · described, no size · exploratory
Q2 2026direction only · described, no size · efficiency
Q3 2026direction only · described, no size · efficiency
engineering · cheap to verify
Engineering headcount held flat by AI coding tools
0% to 1.8% of the quarter’s revenue
Incremental total: counts in full.
Matched line moved : the whole line, not this channel.
our inferencedisclosure: described· motive: efficiency· before LLMs: expanded
The call says nothing about engineering productivity this quarter; the filing repeats that savings from generative AI efficiencies are reinvested in the roadmap (claim c50). The state falls from directional to described. The line moved against the claim: research and development was against , above its prior-year share of revenue, which the filing attributes in part to generative AI spend (claim c49). The size, , is held from the prior quarter's assumptions.
Evidence: 2 quotes, 2 figures, 2 confounds, 2 from before coverage
Engineering hiring avoided, or output gained, from the company's own use of AI coding tools. The line it would show in is research and development; the filing says the efficiencies are reinvested.
Why this motive
Carried from the prior quarter, when flat engineering headcount was attributed to AI. This quarter the call is silent and the filing repeats that savings are reinvested (claim c50).
Before LLMs: expanded
Engineering staff cost is the line that predates the tools: research and development was in Q4 fiscal 2024. On the anchor call the engineering chief already reports a productivity gain from the company's own code models, and the anchor 10-K shows research and development headcount still rising that year. The size is the change AI made, not the whole line.
“However, at the end of fiscal 2024, we began to invest in incremental AI resources to accelerate further growth and as a result our research and development headcount increased by five percent during fiscal 2024.”
“But the other use cases which we are going to see, and in fact, I've rolled out our own code LLMs in our engineering org, and we are already seeing minimum 20% productivity.”
Research and development, prior-year quarter · 2025-CQ3
Reported line it is matched to
Research and development was against , above the prior-year share of revenue: the line rose faster than revenue, the opposite of what a net saving would show.
2026-CQ3: 2025-CQ3: 2026-CQ3: 2025-CQ3:
What else could explain it
acquisition: Acquired engineering teams are in the line this year and not last.
other: The spend on the tools that produce the saving is in the same line, and this quarter it outweighed any saving visible there.
Quotes
“We plan to reinvest savings from efficiencies realized from the rapid deployment of generative AI technologies to accelerate our product roadmap.”
“Research and development expenses as a percentage of total revenues during the three months ended July 31, 2026 increased by approximately one percent compared to the same period a year ago due to increased employee-related costs, as well as spend on hosting services and generative AI technologies.”
Company-wide efficiency from internal agents and Slackbot
Not sized
The only measure is annualized hours saved, (claim c56), a count of AI work with no rate; it sizes nothing (the methodology rule for counts).
Matched line moved : the whole line, not this channel.
described, no sizedisclosure: direction only· motive: efficiency· before LLMs: expanded
Slackbot is credited with annualized hours of productivity gains, against a quarter earlier (claim c56). General and administrative expense was against , relative to its prior-year share of revenue, and the filing calls it about flat in dollars (claim c58). Restructuring of was booked, mainly severance (claim c59). Hours saved are a count of AI work, so the state is directional. The channel is unsized: its only measure is a count. The former estimate, , is no longer the size.
Evidence: 6 quotes, 4 figures, 3 confounds, 3 from before coverage
Employee time and overhead saved by the company's use of its own agents across functions, which it calls being its own first customer. The lines it would show in are general and administrative expense and employee costs generally.
Why this motive
Hours saved are restated at a higher level and general and administrative expense was flat in dollars while revenue grew (claims c56, c58): the cost-line tell of the efficiency motive.
Before LLMs: expanded
Overhead staff costs predate agents: general and administrative expense was in Q4 fiscal 2024. The anchor 10-K attributes that year's fall in overhead headcount to restructuring and a hiring pause, with no mention of AI. The size is the change AI made, not the whole line.
“Our general and administrative headcount decreased by 20 percent during fiscal 2024 driven by the Restructuring Plan and our hiring pause that was in effect during fiscal year 2024.”
“I think the fully autonomous cases, for example, in our own internal use cases with our models, we are able to detect 60% of instance and auto-remediate.”
Annualized employee hours of productivity gains attributed to Slackbot · as-of 2026-07-31
General and administrative · 2026-CQ3
General and administrative, prior-year quarter · 2025-CQ3
Restructuring · 2026-CQ3
Reported line it is matched to
General and administrative expense was against , which is relative to the prior-year share of revenue (a negative figure is below it). The filing describes the line as relatively flat in absolute dollars and does not attribute that to AI.
2026-CQ3: 2025-CQ3: 2026-CQ3: 2025-CQ3:
What else could explain it
transformation program: Restructuring of , against a year earlier, removes headcount; the filing does not tie it to AI.
operating leverage: General and administrative expense growing slower than revenue is ordinary at this scale.
other: Hours of productivity are self-reported and are not a reduction in any cost line.
Quotes
“Slackbot is driving 8.1 million hours of annualized productivity gains for our employees. That's more than double quarter-over-quarter.”
“We expect that general and administrative expenses may decrease as a percentage of revenues over time as we continue to invest in process efficiency initiatives, which includes the use of AI and agents.”
“For the three and six months ended July 31, 2026, general and administrative expenses were relatively flat in absolute dollars compared to the same period a year ago.”
“In the three and six months ended July 31, 2026, we incurred approximately $94 million and $174 million, respectively, of costs related to our restructuring initiatives, which were primarily related to employee transitions, severance payments and employee benefits.”
“the head of all of our operating units have been using Claudeforce to run their businesses inside Salesforce this quarter. It is one of the reasons why we are getting such great performance.”
Q1 2026described · described, no size · narrative-defensive
Q2 2026direction only · described, no size · efficiency
Q3 2026direction only · described, no size · efficiency
Revenue arriving through AI11 channels · $826mn to $1.32bn sized · $304mn to $996mn incremental · 2 not sized
product revenue · cheap to verify
Agentforce subscriptions and credits (Agentforce ARR)
2.6% to 3.3% of the quarter’s revenue
Incremental total: counts in full.
our inferencedisclosure: quantified· motive: offensive· before LLMs: new
Agentforce is stated above of annual recurring revenue, up over , from a quarter earlier (claims c4, c1). The figure was recharacterized: from this quarter it counts Slackbot and Headless, which were outside it before, and the release says the metric may be updated for new products or acquired technologies (claim c5). The increase is therefore part growth and part reclassification, and management does not split it. The ledger reads the quarter at . Accounts in production grew , with paying customers added, and work units in the quarter were .
Evidence: 11 quotes, 7 figures, 3 confounds, 5 from before coverage
Recurring revenue from Agentforce: agent access sold as premium editions and as consumption credits. Management reports it as annual recurring revenue of executed agreements at each period end; no income-statement line carries it, and the revenue category later named Agentforce Apps is the existing applications, not this product.
Why this motive
A separately priced AI product with a disclosed recurring value, bookings that doubled and paying customers added (claims c4, c9, c8): unchanged.
Before LLMs: new
No agent product was offered at the anchor. The anchor shows Einstein, a set of predictive machine-learning features inside the applications, and the first generative assistants announced in 2023; Agentforce launched in the third quarter of fiscal 2025. Management's growth rate puts its recurring value at at the end of fiscal 2025, against a year later.
“Einstein, our AI productivity and development platform, brings AI into Salesforce apps and workflows and offers the ability to deploy conversational, generative AI assistants that empower teams to get work done without compromising data security and privacy.”
“Of course, we have been a leader in this area with Einstein, more than a trillion transactions delivered this week. These are primarily predictive transactions built on machine intelligence, machine learning, and deep learning.”
“I think the way I see it is this AI technologies are on a continuum. There is predictive, then there's generative, and the real long-term goal is autonomous.”
“Can you just distill for us, if you don't mind, what is new about generative AI as far as Salesforce's opportunity is concerned, netting out against what Einstein has been able to accomplish for you, for the company?”
“From time to time, we may update this metric to incorporate new products, feature offerings, or acquired technologies that meet its definition. Beginning in Q2 FY27, Agentforce ARR includes our AI offerings Slackbot and Headless 360.”
“The Company defines Agentforce and Data 360 annual recurring revenue ("ARR") as the annualized recurring value of active Data 360 and certain generative artificial intelligence ("AI") subscription agreements, including those for Agentforce, generative AI products and features, and Headless 360, that were executed at the end of the reporting period.”
“50% of the bookings came from customers refilling the tank. So they consume, they use the Flex Credits, they want more, they raise their hand, we go there”
Premium edition upgrades with embedded AI (Agentforce One Edition, A4X)
0.79% to 2% of the quarter’s revenue; overlaps another channel, not added into totals
our inferencedisclosure: quantified· motive: offensive· before LLMs: expanded
For the first time management gives levels: of sales and service users have upgraded, at a premium of to (claim c13). Bookings more than doubled on the quarter, and the new agentic products and headless access are gated to the premium editions (claims c12, c15). The state moves from directional to quantified. The filing still says pricing was not a significant driver of revenue growth (claim c23). The size, , remains the ledger's assumed share of the Agentforce estimate, inside which it sits.
Evidence: 7 quotes, 3 figures, 2 confounds, 5 from before coverage
Existing application seats upgraded to higher-priced editions that include unlimited agent use by employees. Management counts these inside Agentforce bookings, so the channel is contained in the Agentforce channel.
Why this motive
Management gives a price premium and a penetration rate for the AI-bearing editions (claim c13): a separately priced product with a disclosed attach.
Before LLMs: expanded
Edition upgrades on seat licences are in the anchor: the fiscal 2024 10-K names upgrades and enhanced editions among the sources of subscription growth, on subscription and support revenue of for the year, and says pricing was not a significant driver. The AI-bearing premium editions were introduced during fiscal 2026; no dollar baseline for edition upgrades is disclosed. The size is the whole of an activity that existed before.
“Einstein, our AI productivity and development platform, brings AI into Salesforce apps and workflows and offers the ability to deploy conversational, generative AI assistants that empower teams to get work done without compromising data security and privacy.”
“Our future success also depends in part on our ability to sell additional features and services, more subscriptions or enhanced editions of our services to our current customers.”
“The increase in subscription and support revenues for fiscal 2024 was primarily caused by volume-driven increases from new business, which includes new customers, upgrades and additional subscriptions from existing customers. Pricing was not a significant driver of the increase in revenues for either period.”
“Pricing and packaging strategies for enterprise and other customers for subscriptions to our existing and future service offerings, including for our AI offerings, may not be widely accepted by new or existing customers.”
“Not only do you have the free version of Slack, not only do you have the per-user version of Slack, but then you have the additional LLM version of Slack. For each one of our products, in every single one of our categories, there's that opportunity to upsell and cross-sell into the next version of generative AI.”
“If you want Claudeforce, if you want headless, if you want all these value-added agentic additions and capabilities, you need to move to our premium edition”
“The increase in subscription and support revenues for the three and six months ended July 31, 2026 was primarily caused by volume-driven increases from new business, which includes new customers, upgrades, and additional subscriptions from existing customers. Pricing was not a significant driver of the increase in revenues for either period.”
“We are combining Claude's extraordinary intelligence and reasoning with Salesforce's trusted data and applications and workflows and business rules and governance.”
Flex Credits and agentic enterprise licence agreements
0.79% to 2% of the quarter’s revenue; overlaps another channel, not added into totals
our inferencedisclosure: direction only· motive: offensive· before LLMs: new
Of Agentforce bookings, came from customers who used their credits and bought more (claim c10). Pricing now spans credits bought as used, as overages, or up front, and outcome-based terms are being negotiated (claims c17, c16); the CEO says pricing is still partly tied to per-user models (claim c18). No credits revenue or share of the metric is given. The size, , is the ledger's assumed share of the Agentforce estimate, inside which it sits.
Evidence: 6 quotes, 2 figures, 2 confounds, 1 from before coverage
Consumption credits for customer-facing agents, bought as used, as overages or up front, often inside an agentic enterprise licence agreement. Management counts these inside Agentforce bookings, so the channel is contained in the Agentforce channel.
Why this motive
Carried: consumption credits priced per unit of agent work. This quarter adds outcome-based pricing as an option (claim c16).
Before LLMs: new
Priced units of agent work did not exist before LLMs; the anchor mentions pricing and packaging for AI offerings only as a risk and names no credit or per-conversation price. Flex Credits were introduced in calendar 2025. No dollar baseline is disclosed.
“Pricing and packaging strategies for enterprise and other customers for subscriptions to our existing and future service offerings, including for our AI offerings, may not be widely accepted by new or existing customers.”
Share of Agentforce bookings from customers buying more credits · 2026-CQ3
Agentic work units delivered in the quarter · 2026-CQ3
What else could explain it
other: Credits are booked when bought and earned when consumed.
bundling: Credits are included in enterprise licence agreements rather than priced deal by deal.
Quotes
“50% of the bookings came from customers refilling the tank. So they consume, they use the Flex Credits, they want more, they raise their hand, we go there”
“Some of them want it by consumption, and that is important to them. Some of them are just basic usage customers, and they want to pay that way. Some of them even want to pay by outcome, and that outcome could be by a transaction outcome or a business outcome.”
“That we monetize with ILAs, with Flex Credits, and here we meet customers where they are in their journey. Sometimes they want to buy as they go, sometimes we just charge for overages, and many times they buy up front a bunch of credits.”
“Additionally, our transition toward more complex pricing structures, including AI-driven consumption models, may make it more difficult to optimize our pricing, predict attrition rates, and accurately forecast revenue.”
Q2 2026direction only · our inference · offensive · $70mn to $195mn
Q3 2026direction only · our inference · offensive · $90mn to $225mn
product revenue
Data Cloud subscriptions counted in the AI and data ARR metric
2.2% to 2.9% of the quarter’s revenue
Incremental total: counts at zero at the point and in full at the high end, with no traced baseline.
our inferencedisclosure: bounded· motive: offensive· before LLMs: expanded
The combined metric is nearly . Less Agentforce, is the data platform and Informatica Cloud together; management withdrew the Informatica Cloud figure this quarter, so the data platform alone is bounded rather than given (claims c3, c1). The state moves from quantified to bounded. The size, , subtracts an assumed Informatica level from the remainder.
Evidence: 4 quotes, 2 figures, 2 confounds, 4 from before coverage
The data platform (formerly Data Cloud) and other generative AI subscriptions that management counts with Agentforce in one annual recurring revenue metric. Sized as that metric less Agentforce and less Informatica Cloud.
Why this motive
Carried from prior quarters: the data product is sold on consumption credits and attached to the AI sale.
Before LLMs: expanded
At the anchor the data platform was offered as Data Cloud, a data engine that unifies a customer's records, and management called it the company's fastest-growing product before any agent was offered. Neither anchor source gives its revenue or recurring value; outside Agentforce and Informatica Cloud the metric's remainder was at the end of fiscal 2026. The size is the whole of an activity that existed before.
“Data Cloud is our hyperscale, trusted data engine native to Salesforce. It brings a company's disconnected, enterprise data into Salesforce to deliver an actionable, comprehensive, 360-degree view of a customer.”
“By bringing together structured and unstructured data, Data Cloud offers fast and secure entry into AI for outcomes that are accurate, relevant and grounded with a company's data.”
“Data Cloud is the heart of Customer 360, and now our fastest-growing cloud ever. Data Cloud creates a real-time intelligent data lake that brings together and harmonizes all of our customers' data in one place.”
Agentforce and Data 360 annual recurring revenue, including Informatica Cloud, approximate · as-of 2026-07-31
Data 360 and Informatica Cloud annual recurring revenue together: the metric less Agentforce · as-of 2026-07-31
What else could explain it
relabel: The data platform predates agents and is counted in a metric presented as AI and data; products moved between the Agentforce figure and the remainder this quarter.
other: The remainder is stated less precisely than before: the combined figure is nearly its level and the Agentforce figure is a floor.
Quotes
“Agentforce ARR, it hit $1.5 billion. Our ARR for AI and data is about to cross $4 billion.”
“The Company defines Agentforce and Data 360 annual recurring revenue ("ARR") as the annualized recurring value of active Data 360 and certain generative artificial intelligence ("AI") subscription agreements, including those for Agentforce, generative AI products and features, and Headless 360, that were executed at the end of the reporting period.”
“Headless platform, Data 360, and other growth was fueled by continued momentum in Informatica and Data 360, partially offset by license revenue headwinds and volatility in integration and analytics.”
Informatica Cloud ARR counted in the AI and data ARR metric
2.4% to 2.9% of the quarter’s revenue
Incremental total: counts at zero.
our inferencedisclosure: withdrawn· motive: exploratory· before LLMs: relabelled
Informatica Cloud recurring value was stated in both prior quarters (, then ) and is withdrawn this quarter: the combined metric that contains it is still given and the component is not. The release keeps the definition and drops the number. Informatica still contributed of revenue, of it subscription and support. The size, , assumes a level near the last one stated.
Evidence: 4 quotes, 2 figures, 2 confounds, 1 from before coverage
Cloud data-management subscriptions of Informatica, acquired in November 2025, which management includes in the annual recurring revenue metric it presents for Agentforce and the data platform.
Why this motive
Carried from the prior quarter: an acquired subscription base inside an AI-labelled metric, with no tell about why its customers pay; exploratory, since nothing in the sources contradicts itself.
Before LLMs: relabelled
Informatica reported this cloud recurring revenue as an independent company until November 2025; at the anchor the company's own data integration was offered through MuleSoft. Informatica Cloud entered the metric on acquisition at : the same subscriptions under a new owner and an AI and data label.
“With MuleSoft, customers connect any data, or AI model securely and automate tasks and processes, using discoverable and reusable APIs and integrations, to transform businesses and drive faster time to value.”
“The acquisition of Informatica in November 2025 contributed approximately $456 million and $900 million of total revenues for the three and six months ended July 31, 2026, respectively.”
“Headless platform, Data 360, and other growth was fueled by continued momentum in Informatica and Data 360, partially offset by license revenue headwinds and volatility in integration and analytics.”
Core seat and edition revenue pulled through by agent adoption
0% to 1.8% of the quarter’s revenue
Incremental total: counts in full.
our inferencedisclosure: quantified· motive: offensive· before LLMs: expanded
The measure changed: customers that adopt agents are said to be at double their annual order value on average, where the prior quarter gave a spend multiple for a named cohort (claim c22). Seats grew in sales, service and Slack (claims c20, c21). Subscription and support revenue was against , with from Informatica, and the filing attributes the growth to volume (claim c23). The size, , is the ledger's assumed share of the organic increase.
Evidence: 4 quotes, 3 figures, 3 confounds, 2 from before coverage
Ordinary subscription revenue that arrives because agents raise the return on the applications: seats added in deals that include agents and higher total spend at customers that adopt them. Distinct from the AI products themselves.
Why this motive
A measured movement in order value that management attributes to agent adoption (claim c22), with seat growth in the largest applications (claim c20).
Before LLMs: expanded
Seat subscriptions are the business that predates agents: subscription and support revenue was in fiscal 2024, about a quarter, and the anchor 10-K attributes its growth to new customers, upgrades and added subscriptions. Management's measures of pull-through (seats added in the largest deals, spend multiples of agent adopters) begin with the 2026 calls. The size is the change AI made, not the whole line.
“Our future success also depends in part on our ability to sell additional features and services, more subscriptions or enhanced editions of our services to our current customers.”
“The increase in subscription and support revenues for fiscal 2024 was primarily caused by volume-driven increases from new business, which includes new customers, upgrades and additional subscriptions from existing customers. Pricing was not a significant driver of the increase in revenues for either period.”
Subscription and support revenues, prior-year quarter · 2025-CQ3
Informatica contribution to subscription and support revenues · 2026-CQ3
What else could explain it
mix shift: The order value of agent adopters includes the AI products themselves; the customers who adopt first are likely the fastest-growing ones.
operating leverage: Seat growth has ordinary causes: new customers, new geographies and more sales capacity.
other: The cohort measured is not the one measured the prior quarter and its size is not given.
Quotes
“Our seats were supposed to decline. Instead, Agentforce sales, service, and Slack all saw year-over-year growth, and we were told customers would abandon us, but attrition is near its lowest level ever.”
“Customers that start the agentic journey with Salesforce, on average, they are already at double the AOV with a perspective of nearly quadrupling, 3 x - 4 x the AOV.”
“The increase in subscription and support revenues for the three and six months ended July 31, 2026 was primarily caused by volume-driven increases from new business, which includes new customers, upgrades, and additional subscriptions from existing customers. Pricing was not a significant driver of the increase in revenues for either period.”
Subscriptions sold to AI labs and AI-native companies
0.03% to 0.73% of the quarter’s revenue
Incremental total: counts in full.
our inferencedisclosure: quantified· motive: offensive· before LLMs: new
of the leading AI companies are said to use Salesforce and Slack, with spend up year over year (claim c27). This is the first growth rate for the channel and it has no dollar base. These payers are funded largely by investors; one of them is a company Salesforce holds a stake in and buys models from. The size, , is a reference-class ballpark that the stated growth would not change the order of.
Evidence: 3 quotes, 2 figures, 2 confounds, 2 from before coverage
Slack and CRM subscriptions bought by AI labs and AI-native companies, which management cites as customers. Kept apart from enterprise revenue because these payers are funded largely by investors, and some are also suppliers of models to the company.
Why this motive
Spend growth of the leading AI companies is stated and offered as evidence that frontier models depend on the applications (claim c27): a measured movement.
Before LLMs: new
AI labs were already named as Slack customers on the first-quarter fiscal 2024 call, one of them from its first day. The cohort at its present scale did not exist before LLMs, and no revenue from it is disclosed for any period.
“The powerful part about that was I realized that everything from day one at OpenAI had been in Slack.”
“In Q1, we saw amazing momentum with customers like the California Office of Systems Integration, Paramount Global, Revel, and OpenAI, and rolled out an AI-ready platform, Slack Canvas, and app integrations with ChatGPT and Anthropic's Claude.”
Leading AI companies management counts as customers · as-of 2026-07-31
Spend of those AI companies, year-over-year growth · 2026-CQ3
What else could explain it
other: A growth rate from a small base; the base is not disclosed.
other: The customer, supplier and investor relationships with the same lab are not at arm's length in every respect.
Quotes
“Nine of the 10 AI companies, like you just heard from Anthropic, that standardized on Salesforce, use Salesforce and Slack. Their spend is up 435% year over year.”
“Companies like Anthropic, OpenAI, Lovable, Replit, on and on and on, I think there is more than 1 million companies on Slack now, are all building their agentic products directly into Slack because Slack is where their customers want them.”
The measures are active users, a quarterly growth rate and a multiple of edition upgrades (claims c30, c31); users of a feature included in paid editions are a count, not seats sold at a seat price, and an upgrade multiple has no level (the methodology rules for counts and for volume times price).
described, no sizedisclosure: direction only· motive: offensive· before LLMs: expanded
Slackbot has active users, up on the quarter, and upgrades to premium Slack editions have tripled since its launch (claims c30, c31). From this quarter it is counted inside Agentforce recurring value (claim c4), so the channel now overlaps the Agentforce channel. The channel is unsized: its measures are counts and a multiple with no level. The former estimate, , is no longer the size.
Evidence: 7 quotes, 2 figures, 2 confounds, 3 from before coverage
Slack revenue that exists because Slackbot, an employee agent included in paid Slack editions, moves customers from free to paid or to higher editions. Slack revenue is not reported separately.
Why this motive
Edition upgrades tripled since launch and an active-user count is given (claims c31, c30): measured movement attributed to the product.
Before LLMs: expanded
Slack was offered in a free version and a per-user version at the anchor, with the cost of its free user base inside cost of subscription and support revenues; on the anchor call the CEO described an added LLM version as the next step up. The fiscal 2026 10-K describes Slackbot as recently introduced. No Slack revenue is disclosed. The size is the change AI made, not the whole line.
“Our future success also depends in part on our ability to sell additional features and services, more subscriptions or enhanced editions of our services to our current customers.”
“Also included in the cost of subscription and support revenues are expenses incurred supporting the free user base of Slack, including third-party hosting costs and employee-related costs, including stock-based compensation expense, specific to customer experience and technical operations.”
“Not only do you have the free version of Slack, not only do you have the per-user version of Slack, but then you have the additional LLM version of Slack. For each one of our products, in every single one of our categories, there's that opportunity to upsell and cross-sell into the next version of generative AI.”
“From time to time, we may update this metric to incorporate new products, feature offerings, or acquired technologies that meet its definition. Beginning in Q2 FY27, Agentforce ARR includes our AI offerings Slackbot and Headless 360.”
“Companies like Anthropic, OpenAI, Lovable, Replit, on and on and on, I think there is more than 1 million companies on Slack now, are all building their agentic products directly into Slack because Slack is where their customers want them.”
Q1 2026described · described, no size · product-defensive
Q2 2026direction only · described, no size · offensive
Q3 2026direction only · described, no size · offensive
pricing packaging
Agent and API access to the platform (Headless, MCP)
Not sized
The only measure is a multiple of agent calls through MCP and the command line, (claim c34), a count of AI work, and the CFO says the company is 'working through headless monetization' (claim c33). A count sizes nothing (the methodology rule for counts).
described, no sizedisclosure: direction only· motive: exploratory· before LLMs: expanded
Agent use of the applications through MCP and command-line calls rose times (claim c34). Headless is counted inside Agentforce recurring value from this quarter and named in the metric's definition (claims c4, c36), yet the CFO says its monetization is still being worked through (claim c33). The state moves from described to directional; the former estimate, , an assumed share of Agentforce, is no longer the size: the call multiple is a count and monetization is still being worked through.
Evidence: 8 quotes, 1 figure, 2 confounds, 2 from before coverage
Access to the applications and data by outside agents and coding tools through MCP servers, APIs and the command line, which management describes as a surface it has not previously monetized.
Why this motive
The CFO says monetization is still being worked through (claim c33) while the release counts the product in recurring value and the CEO gates it to the premium edition (claims c4, c15). The tells conflict; the less durable motive is kept.
Before LLMs: expanded
Platform and API access for third parties and developers came with the subscription at the anchor and was not separately priced; the anchor also shows API management offered through MuleSoft. Management calls the Headless surfaces ones it has never monetized. The size is the whole of an activity that existed before.
“With MuleSoft, customers connect any data, or AI model securely and automate tasks and processes, using discoverable and reusable APIs and integrations, to transform businesses and drive faster time to value.”
“We also enable third parties to use our platform and developer tools to create additional functionality and new applications that run on our platform, which are sold separately from, or in conjunction with, our service offerings.”
“MCP and API usage grew significantly in the quarter as customers embed Salesforce intelligence, the data, metadata, and semantics directly into their workflows.”
“From time to time, we may update this metric to incorporate new products, feature offerings, or acquired technologies that meet its definition. Beginning in Q2 FY27, Agentforce ARR includes our AI offerings Slackbot and Headless 360.”
“If you want Claudeforce, if you want headless, if you want all these value-added agentic additions and capabilities, you need to move to our premium edition”
“The Company defines Agentforce and Data 360 annual recurring revenue ("ARR") as the annualized recurring value of active Data 360 and certain generative artificial intelligence ("AI") subscription agreements, including those for Agentforce, generative AI products and features, and Headless 360, that were executed at the end of the reporting period.”
Q2 2026described · described, no size · exploratory
Q3 2026direction only · described, no size · exploratory
sales force · cheap to verify
Pipeline generated by the company's own sales agents
0% to 0.05% of the quarter’s revenue
Incremental total: counts in full.
our inferencedisclosure: described· motive: offensive· before LLMs: expanded
No lead or pipeline figure is given this quarter, so the state falls from quantified to described. The sales organization describes agents it built on a third-party model to inspect the pipeline, and the CEO credits internal use of the new product for the quarter's performance (claims c37, c38). Sales headcount is still growing, to about account executives (claim c39); sales and marketing was against . The size, , carries the prior quarter's pipeline with a wider range.
Evidence: 4 quotes, 3 figures, 2 confounds, 3 from before coverage
Demand the company creates for itself by running its own sales agents: leads qualified and pipeline generated that its sales staff would not have reached. Recorded as revenue rather than savings because management says sales headcount keeps growing.
Why this motive
Carried from the prior quarter: pipeline attributed to the company's own agents. This quarter gives no count, only that leaders run the business through agents (claims c37, c38).
Before LLMs: expanded
Lead handling at the anchor was done by telephone sales staff and self-service, inside marketing and sales expense of in Q4 fiscal 2024; the anchor call describes a planned automated motion for the low end of the market. No agent-generated pipeline is disclosed before 2026. The size is the change AI made, not the whole line.
“We sell our services primarily through our direct sales force, which comprises telephone sales personnel based in regional hubs, field sales personnel based in territories close to their customers and self-service offerings.”
“How do we drive a self-serve motion, an automated motion at the low end of our market to bring our account executives upmarket to drive higher productivity in the sales organization?”
Sales and marketing, prior-year quarter · 2025-CQ3
What else could explain it
other: The last stated pipeline is a quarter old.
operating leverage: Sales performance in the quarter also reflects added sales capacity.
Quotes
“So we build our own agents based on Claude, and that agent inspects, they are our deputy CROs for all our businesses, and they inspect the business, the pipeline.”
“the head of all of our operating units have been using Claudeforce to run their businesses inside Salesforce this quarter. It is one of the reasons why we are getting such great performance.”
“We expect that sales and marketing expenses may decrease as a percentage of revenues over time as we continue to focus on leveraging our self-serve and partner-led channels and increasing our sales productivity, which includes the use of AI and agents.”
23.8% of the quarter’s revenue; a non-operating gain, not added into totals
reported in the filingdisclosure: quantified· motive: exploratory· before LLMs: new
The 10-Q names the investment and attributes of the quarter's unrealized gains to it, inside net strategic gains of ; the stake is carried at (claims c41, c42). This is the only dollar on the Salesforce exhibit that a filing reports and ties to AI. It equals of the quarter's revenue, but it is not revenue: it is a non-cash gain below operating income, set by what other investors paid for the lab's shares. It is recorded as a non-operating gain and left out of flow totals.
Evidence: 4 quotes, 3 figures, 2 confounds, 2 from before coverage
Mark-to-market gains on the company's equity stake in an AI lab, recorded in gains on strategic investments below operating income. Non-cash and not revenue: it is recorded as a non-operating gain, traced like any other figure, and left out of flow totals.
Why this motive
Carried from the prior quarter: a non-cash mark set by the investee’s funding rounds, read exploratory since none of the operating tells applies and the sources do not contradict each other.
Before LLMs: new
The stake dates from 2023: the anchor call says the company had just invested in Anthropic. Strategic investments were carried at at the end of fiscal 2024 and the line recorded a net gain of in fiscal 2022, so marks on venture stakes predate LLMs while this holding does not. The fiscal 2026 10-K reports of unrealized gains from one privately held investment it does not name.
“We invest in companies that we believe are digitally transforming their industries, improving customer experiences, helping us expand our solution ecosystem or supporting other corporate initiatives, including AI.”
Carrying value of the investment in Anthropic · as-of 2026-07-31
Gains on strategic investments, net · 2026-CQ3
Unrealized Anthropic gain as a percent of total revenues · 2026-CQ3
What else could explain it
one time item: A mark-to-market gain recurs only if the investee raises at a higher price again; it can reverse through impairment.
other: Non-cash, outside operating income, and funded by investors' valuation of an AI lab rather than by any customer's use of AI.
Quotes
“The unrealized gains for the three and six months ended July 31, 2026 include gains of $2.7 billion and $3.0 billion, respectively, related to the Company’s investment in Anthropic.”
“As of July 31, 2026, our strategic investment portfolio consisted of investments in over 450 companies with a combined carrying value of $11.3 billion, including the Company’s investment in Anthropic PBC (“Anthropic”) which represented approximately $5.1 billion of the total strategic investments portfolio.”
“As of July 31, 2026 and January 31, 2026, the Company’s privately held investment in Anthropic PBC (“Anthropic”) had a carrying value that represented approximately 45 percent and 22 percent, respectively, of the Company’s total strategic investments portfolio.”
Q1 2026described · described, no size · exploratory
Q2 2026described · described, no size · exploratory
Q3 2026quantified · reported in the filing · exploratory · $2.70bn
Cost imposed, or revenue lost, by others’ AI2 channels · $0 to $22mn sized · $0 incremental · 1 not sized
customer cohort
Subscription revenue lost to customers' AI (seats and attrition)
Not sized
Management denies the effect: seats grew, attrition is near a record low and frontier models are said to depend on the applications (claims c60, c20, c62, c61). A toll whose money management says has not started gets no ballpark (the methodology rule for money not started).
described, no sizedisclosure: bounded· motive: imposed· before LLMs: new
Management answers the displacement thesis point by point: seats grew, attrition is near a record low, and frontier models are said to depend on the applications (claims c60, c20, c62, c61). The filing puts attrition at approximately , as in prior quarters, and keeps would-be customers who build in-house among its competitors (claims c63, c64). A customer on the call describes replacing vertical software built in-house while keeping the platform; that is the customer's account and is not quoted here. The channel is unsized: management attributes none of the attrition to AI. The former estimate, , is no longer the size. The counterparty is mixed: frontier-model labs, AI-native vendors, and customers building their own applications with AI.
Evidence: 9 quotes, 2 figures, 2 confounds, 3 from before coverage
Revenue that leaves because customers need fewer seats or replace applications with AI built on frontier models or by AI-native vendors. Sized as the revenue that would have been billed at the prior rate; management bounds it through attrition and seat growth.
Why this motive
Carried: a toll, not the company's choice.
Before LLMs: new
Attrition existed before LLMs for ordinary reasons: about at the end of fiscal 2024, excluding Slack, and about at the end of fiscal 2026. The anchor 10-K already lists would-be customers that build their own applications among competitors; a part of attrition caused by customers' AI would be new, and none is disclosed.
“As of January 31, 2024, our attrition rate, excluding Slack, was approximately eight percent.”
“traditional platform development environment companies and cloud computing development platform companies who may develop toolsets and products that allow customers to build new apps that run on the customers' current infrastructure or as hosted services, as well as would-be customers who may develop enterprise applications for internal use.”
other: Attrition is a trailing twelve-month measure that excludes Slack self-service and acquisitions; a recent change would show late.
mix shift: Licence revenue, integration and analytics remain weak for reasons the sources do not separate from AI.
Quotes
“This is not the SaaSpocalypse. As I said earlier, we have been hearing about this for the last two quarters, these dire predictions about the end of software and how the models eat everything.”
“traditional platform development environment companies and cloud computing development platform companies who may develop toolsets and products that allow customers to build new applications, including AI-augmented applications, that run on the customers’ current infrastructure or as hosted services, as well as would-be customers who may develop enterprise applications for internal use.”
“Our seats were supposed to decline. Instead, Agentforce sales, service, and Slack all saw year-over-year growth, and we were told customers would abandon us, but attrition is near its lowest level ever.”
“I think that has been the huge shock for us, that these new next-gen tools, the Cursors, the Replit, the Claudes of the world, make our product better.”
“Additionally, our transition toward more complex pricing structures, including AI-driven consumption models, may make it more difficult to optimize our pricing, predict attrition rates, and accurately forecast revenue.”
Professional services revenue displaced by agent-assisted implementation
0% to 0.2% of the quarter’s revenue
Incremental total: counts at zero.
Matched line moved : the whole line, not this channel.
our inferencedisclosure: described· motive: imposed· before LLMs: relabelled
The CEO says coding tools now do customization and administration that once needed service staff, so a customer is running in a month (claim c66). Professional services revenue was against , and excluding Informatica's services it is lower than a year earlier, a wider gap than the prior quarter's . The filing attributes the decrease to less demand for large transformation engagements (claim c67). The size, , is an assumed share of the decline.
Evidence: 4 quotes, 4 figures, 2 confounds, 3 from before coverage
Implementation and customization work the company no longer bills because customers and partners deploy the platform faster with coding agents. The line it shows in is professional services and other revenues.
Why this motive
Carried: a toll on services revenue from faster, agent-assisted implementation.
Before LLMs: relabelled
Professional services and other revenue was in fiscal 2024, about a quarter, and was already falling at the anchor on less demand for large multi-year transformation engagements, the cause the filings still give. On the anchor call the engineering chief said generative tools for developers reduce the cost of implementations.
“The decrease in professional services and other revenues was due primarily to less demand for larger, multi-year transformation engagements and, in some cases, delayed projects.”
“We are going to roll Einstein GPT for our developers in the ecosystem, which will not only help the low-code developers bridge the gaps, where there's a talent gap, but also it reduces the cost of implementations for a lot of people, so there's a lot of value.”
“Also, in Q1, our professional service business started to see less demand for multi-year transformations, and in some cases, delayed projects as customers focus on quick wins and fast time to value.”
Professional services and other revenues · 2026-CQ3
Professional services and other revenues, prior-year quarter · 2025-CQ3
Informatica contribution to total revenues · 2026-CQ3
Informatica contribution to subscription and support revenues · 2026-CQ3
Reported line it is matched to
Professional services and other revenue was against ; with Informatica's services removed it is lower than a year earlier. The filing attributes the decrease to less demand for large multi-year engagements.
2026-CQ3: 2025-CQ3: 2026-CQ3: 2026-CQ3:
What else could explain it
acquisition: Informatica's services revenue is in the line this year and not last.
other: The filing gives weaker demand for large transformation engagements as the cause; much implementation work is billed by partners and never passes through this line.
Quotes
“That all of a sudden it can customize the data, the metadata, set up the preferences, do the administration, do things that maybe you needed a bunch of service folks to do. Now all of a sudden, you're getting going in a month.”
“I think that has been the huge shock for us, that these new next-gen tools, the Cursors, the Replit, the Claudes of the world, make our product better.”
“The decrease in professional services and other revenues for the three and six months ended July 31, 2026 was primarily due to less demand for larger, multi-year transformation engagements, which may continue in the near term.”
Q3 2026. Growing slower than revenue: general and administrative (−1.2%). A displaced cost shows up as a line that stays under the dashed revenue line. These are the audited lines, as first reported; nothing here is attributed to AI by the filing. Not drawn: restructuring, which one-time items move by more than 60% in a quarter; the values are in the table below.
Total cost of revenuesResearch and developmentSales and marketingGeneral and administrativeTotal operating expensesRevenue