Q1 FY27 Preview: Agentforce on Trial

May 27 is sixteen days away, and it is shaping up to be the most consequential earnings call Salesforce has hosted in years. The company confirmed the Q1 fiscal 2027 release date in a May 6 investor notice: results land after market close, with the live webcast at 2:00pm Pacific. Between now and then, Marc Benioff has to convince a market that has dragged CRM down roughly 30 percent year-to-date that the Agentforce story is real, the Data 360 story is bigger than the headlines suggest, and the $25 billion debt-funded buyback that is currently absorbing the company's free cash flow is the right use of capital.

This is not a normal preview. It is a stress test of the agentic-enterprise thesis that Salesforce has spent the last eighteen months selling.

The setup: numbers that have to hold

The Q4 FY26 print in February was, by any reasonable read, a beat. Revenue came in at about $11.2 billion, up roughly 12 percent year over year. Adjusted EPS was $3.81 versus a consensus of $3.05, a 25 percent surprise. Agentforce and Data 360 combined had pushed past $1.4 billion in annual recurring revenue, with more than 9,500 paid Agentforce deals closed and 3.2 trillion tokens processed across the platform in the quarter. cRPO grew 11 percent.

And the stock fell anyway. It has kept falling. By the close on May 8 it was sitting near a 52-week low, off about 30 percent for the year. The market is no longer pricing this as a beat-and-raise growth story; it is pricing it as a SaaS incumbent in the middle of a generative-AI rerating, and the rerating has not been kind.

Here is what is unusual about this moment. The company is not in operational trouble. The fundamentals are fine. The argument is entirely about the future shape of the business. Bears think agent revenue cannibalizes seat revenue faster than it replaces it. Bulls think Agentforce becomes a consumption-priced AI distribution layer that compounds on top of the CRM seats, not under them. The Q1 FY27 call is the first quarter where the new reporting structure Salesforce signaled in February should let analysts see the split more clearly.

That reporting structure is the first thing to watch on May 27.

What's actually changing in the reporting

In the February earnings call, CFO Robin Washington signaled that fiscal 2027 disclosure would "break out AI-related revenue more clearly." She did not commit to a line item, but multiple analyst notes in March and April have anticipated some form of explicit Agentforce ARR or AI consumption revenue, separated from Sales Cloud, Service Cloud, and the rest of the legacy seat business.

If that breakout shows up on May 27, the call gets easier. Analysts will get a number to model. If it does not, expect a brutal Q&A session in which every analyst tries to back into Agentforce ARR from cRPO and consumption commentary. Either way, the company has to address attach rate (what percentage of existing customers are buying Agentforce on top of their seat licenses) and consumption growth (how many flex credits and tokens those customers are burning).

The State of Sales 2026 report that Salesforce dropped on May 9 is, in part, marketing scaffolding for this question. Its top numbers, 87 percent AI penetration in sales orgs and 54 percent of individual sellers having used an agent at least once, are positioned exactly where Wall Street needs to see traction. Expect them to be cited on the call.

The buyback is the other story

Sitting alongside the operating questions is a capital-structure question that no software company has had to answer at this scale before. In March, Salesforce issued $25 billion of senior notes, with coupons running from 4.5 to 6.7 percent and maturities stretching as far as 2066. That debt is funding the first half of a $50 billion buyback authorization the board approved in February. The first wave of accelerated repurchases began in March; depending on average execution price, $25 billion buys somewhere between 95 and 110 million CRM shares, or about 10 to 11 percent of the float.

The bull case on the buyback is straightforward: Benioff has said publicly he thinks the market's AI-disruption sell-off has made the share price unjustifiably cheap, and using cheap-ish investment-grade debt to retire shares at a depressed price is textbook good capital allocation if you believe earnings power is intact. The bear case is also straightforward: layering $25 billion of fresh long-dated debt onto a SaaS balance sheet right as the business model is being questioned is unprecedented, and the interest expense alone will run roughly $1.3 billion a year before any of the new agent products have proven they can refill the operating-income line.

Expect at least three analysts on the call to ask about the second tranche of the $50 billion authorization, what the timing looks like, and whether more debt is involved. The honest answer is that it depends on where CRM trades, which is exactly what Benioff will say.

Product evidence: the Claude integration matters more than it looks

While the financial story dominates the headlines, there is one product development from the last two weeks that should not get lost. Salesforce and Anthropic confirmed that Anthropic's Model Context Protocol Apps are now the official bidirectional bridge between Claude and the Salesforce platform. The first surface is Slack: Claude users can pull live Salesforce context into a Claude conversation and push the resulting actions back into the org. The same plumbing extends across Agentforce 360 in the coming months.

The detail that makes this strategically interesting is the trust boundary. Anthropic is the first large-language-model provider whose models run fully inside Salesforce's tenant isolation. Customer data never leaves the Salesforce-managed VPC, and the same security controls customers already trust for Apex execution and metadata operations cover the Claude calls. That removes the single biggest enterprise objection to LLM-driven workflows in regulated industries.

Why does this matter for the earnings call? Because the bear thesis on Agentforce is that customers will bypass Salesforce entirely, talk to Claude or ChatGPT directly, and treat the CRM as a commoditized system of record. The Anthropic deal is a direct counter: Claude becomes a Salesforce distribution channel rather than a competitor, and the regulated-industries angle, banking, healthcare, public sector, is exactly where Salesforce already has the deepest install base. Expect Benioff to spend several minutes on this on May 27, probably framed as evidence that Agentforce 360 is the only platform with the trust posture to make enterprise AI usable.

The mechanics are worth understanding even if you do not care about the stock. When a Claude user inside Slack asks "show me the at-risk renewals in EMEA above $500K," the MCP App resolves that request inside the customer's Salesforce tenant. The relevant Opportunity and Account records are surfaced as structured context, not raw exports. If the user then says "draft a save plan for the top three," the resulting plan is written back to Salesforce as Tasks attached to the right records, governed by the same field-level security and sharing rules that already protect those records. There is no shadow data store, no copied CRM dump in a vector database that the security team will need to audit. The same pattern works through the new Slack external client app flow that supplies the OAuth scope.

What to listen for on May 27

If you only have time to track four things during the call, track these.

One. Agentforce ARR or paid deal count disclosed cleanly. Anything less specific than the $1.4 billion combined number from Q4 is a step backward and the stock will reflect it. Anything that shows clean growth, even at a smaller absolute level, is bullish.

Two. Data 360 consumption commentary. This is the unsexy half of the AI story. Without Data Cloud ingestion, the agents have nothing to ground on, and consumption-priced Data 360 is where some of the most durable AI-era revenue actually sits. Watch for token counts, ingested-row counts, or active-customer counts.

Three. Margin trajectory and operating leverage. Agentforce has model-cost-of-goods that the legacy SaaS business does not. The question is whether the company's gross margin compresses as agent revenue grows. Robin Washington has been guiding to flat-to-up margins; if that guidance changes on the call, the stock moves.

Four. Capital return commentary. The market needs to hear what happens to the second $25 billion of the buyback authorization. Silence on this point will be read as preserving optionality. Specifics will be read as confidence.

Sixteen days. Salesforce has put more pre-earnings signal into the market in the last week, the Anthropic deepening, the State of Sales report, the Summer '26 release going live in sandboxes, than in any quarter in recent memory. By May 27 the question will no longer be whether Agentforce is real. It will be whether the rate at which Agentforce is real is fast enough to outrun the rate at which the rest of the SaaS bundle is being rerated.

We will cover the call live in the next news roundup.