CRM (Trans): What Do 28.6T Tokens and 3.8bn AWUs Buy?

Dolphin Research recap of$Salesforce(CRM.US) FY27 Q1 earnings call. For our take on the print, see Salesforce: No growth, no profit — is the AI story still working?.

I. Key takeaways from the print

1. Largest shareholder return ever: Executed a $25 bn accelerated share repurchase (ASR) this quarter, the biggest in company history and half of the $50 bn buyback authorization. The ASR reduced Q1 diluted shares by 10% YoY, with a single-quarter reduction of 103 mn shares (11% of float), lifting non-GAAP EPS by $0.23 and GAAP EPS by $0.14.

2. Raised full-year revenue guide: FY27 revenue guided to $45.9–46.2 bn (midpoint up). Q2 revenue guided to $11.27–11.35 bn, implying ~10% CC growth; subscription & support revenue CC growth for the year maintained at ~11%.

3. Cash flow guide cut: Due to a ~5 ppt headwind from debt issuance tied to the ASR, F2Q27 operating cash flow and free cash flow are both guided to grow 4%–5% YoY. Management highlighted the ASR-related financing as the primary drag.

4. Margin beat this quarter, but FY guide trimmed: GAAP OPM of 21.1%, up 130 bps YoY, beating expectations. Full-year GAAP OPM guidance was nudged down to 20.6%, mainly on higher restructuring charges.

5. New revenue disclosure live: FY27 debuts the new revenue taxonomy (Agentforce Apps / Data 360, Platform & Other) with this quarter the first under the new framework.

II. Earnings call details

2.1 Management commentary highlights

1. Overall performance

a. Q1 revenue was $11.13 bn, +13% YoY (+12% CC), above guidance; cRPO reached $33.6 bn, +~14% YoY (+13% CC). Operating cash flow was $6.7 bn.

b. Recorded 98 new ACV deals over $1 mn, an all-time high. The top 10 deals saw annual incremental bookings up 60% YoY, with combined TCV of $800 mn (~2.5x the prior-year cohort).

c. With AI coding tools, engineering doubled features and code shipped YoY while reducing defects and incidents. Productivity gains were broad-based across the stack.

2. Agentforce & AI

a. Agentforce ARR surpassed $1 bn for the first time; combined with Data 360 and Cloud, AI + data ARR reached $3.4 bn.

b. Processed 28.6 trillion tokens in the quarter, +152% QoQ, translating into 3.8 bn Agentic Work Units (AWUs), +111% QoQ. Usage momentum remained strong across cohorts.

c. Half of Agentforce and Data 360 bookings came from expansions by existing customers; the top 10 AWU customers lifted overall Salesforce spend by 1.5x over the past year.

d. Advanced SKUs (A1E and A4X, inclusive of Agentforce) bookings rose nearly 60% YoY; in-production Agentforce customers grew 50% QoQ.

e. On help.salesforce.com, Agentforce autonomously handled over 4 mn service requests, now 2x human-handled volumes. Automation is scaling smoothly.

f. Three AI monetization paths: upgrade existing human seats (unlimited AI usage), unlock new seat scenarios as re-architected clouds improve ROI and attract previously priced-out users, and sell Flex Credits for customer-side AI use cases (6 of the top 10 deals were AELA with sizable Flex Credits).

3. Agentforce Coworker & Headless 360

a. Newly launched Agentforce Coworker is embedded in every Salesforce app, enabling AI answers and auto-execution from in-app search with no heavy configuration. This upgrades ~1 bn monthly in-app searches into actions.

b. Headless 360, delivered via MCP, API-ifies the full Salesforce platform, allowing customers to invoke Salesforce data and capabilities from any tool (e.g., Claude, Cursor, ChatGPT). Since launch in Apr, it has processed 4.5 mn MCP calls; core platform API calls were nearly 1 trillion in Q1.

c. Strategically, Headless extends TAM to previously unmonetized surfaces, marking the next growth phase toward the FY30 framework. Management is working with customers and partners on fair monetization in new interaction contexts.

4. Slack

a. Slack contributed nearly half of new $1 mn+ ACV deals this quarter, +80% YoY; Slack AWUs rose 350% QoQ. Large-deal momentum is broad.

b. Slackbot is the fastest-adopted AI tool in Salesforce history, creating ~3.8 mn hours of annualized productivity. Slack MCP reached 1 mn users within six weeks, with over 50 mn tool invocations.

c. In Q1, the community built 3 mn custom apps on Slack (+8x QoQ), including 250k AI Agents (2x QoQ, +8x YoY). Management sees Slack on track to become Salesforce’s fourth $10 bn cloud.

5. Informatica & Data 360

a. Post-acquisition integration of Informatica is ahead of plan, with bookings and revenue accelerating from single-digit to double-digit growth. It provides the heavy-duty data management infrastructure needed to move from pilot to production.

b. The timing of Informatica cloud and on-prem revenue was a key driver of the beat this quarter, alongside professional services timing. Both contributed to outperformance.

6. FY30 framework progress

a. Management reaffirmed the ‘Rule of 50’ targets (revenue of $63 bn+ including Informatica, with margin improvement). The Customer Zero program and a lean Agentic Enterprise are central to execution.

b. The engineering org is ~15k and has been stable for ~2 years, with AI tools driving ongoing efficiency; sales capacity continues to expand to cover a broader opportunity set.

2.2 Q&A

Q: Agentforce metrics look strong. Can you share more on customer pipeline signals and internal usage?

A: Marc said Salesforce itself is the deepest user of Agentforce. The help.salesforce.com and 1-800-NO-SOFTWARE service queues are now largely autonomous, with Agentforce handling authentication, answers, and handoffs to humans. In Q1, the Agentforce sales module alone autonomously worked 220k leads, generating $42 mn in pipeline, and the most important tech change this quarter was Agentforce replacing the global search bar to instantly upgrade AI across all apps at zero implementation cost.

Q: Given cRPO only met guidance the past two quarters and Tableau/Commerce Cloud were drags, what underpins confidence in an H2 organic re-acceleration?

A: Robin noted the key forward indicator is ‘net-new AOV growth outpacing growth in the AOV base’. This signal emerged in H2 FY26 and is expected to continue in H1 FY27, driving a re-acceleration in core revenue in H2. The full-year guide is up, Agentforce and Data 360 pipelines are very strong, and cRPO is timing-affected with its lead-indicator role evolving as consumption models mix up.

Miguel added that post-acquisition, Informatica bookings accelerated from single-digit to strong double-digit growth over two quarters, and revenue is now growing double digits. The top 10 deals’ annual incremental bookings grew 60% YoY with TCV ~2.5x last year, and seven added human seats; net-new AOV outpacing AOV growth is the org-wide north star, with attrition improving internally and externally. Headless 360’s fourth monetization path (new interaction surfaces) is not fully in the guide, offering upside optionality.

Q: How does Headless 360 balance expanding TAM with the risk of value disintermediation from Salesforce (customers building their own or competitors piggybacking)?

A: Patrick said Headless is a mature tech concept (decoupling UI from underlying capabilities). Salesforce has long been the most-invoked API set per Postman, and Headless formalizes an open stance: ‘use Salesforce however you want’, with customers seeking more ways to extract more value from Salesforce rather than to replicate it themselves.

Srini added that Headless MCP tool calls have exceeded 1.5 mn and Slack MCP tool calls surpassed 50 mn, revealing significant previously suppressed demand. Miguel cited Adecco, whose agents built in other AI labs are now connecting to Salesforce data via MCP, and Anthropic, whose Sales Cloud usage jumped 5x in Q1 by accessing it through Headless from Coworker, Slack, and other entry points, making Sales Cloud more strategic to them than ever; Marc emphasized Salesforce’s API-first, metadata-driven architecture, noting Headless does not ‘remove apps’ but opens every app to any interface.

Q: With 28.6 trillion tokens and 3.8 bn AWUs, how does usage convert to revenue, and why wasn’t GPM hit by the token bill?

A: Marc said the engineering team has held near 15k for ~two years, with AI tools continuously lifting engineering efficiency, a key driver of margin expansion. On token costs, Salesforce procures substantial volumes of OpenAI Codex and Anthropic tools, and both are also deep customers of Salesforce across Slack and Sales Cloud/Service Cloud, creating mutually beneficial commercial relationships.

Robin added that the top 10 AWU customers lifted total Salesforce spend by 1.5x over the past year, demonstrating monetization via consumption and platform expansion as usage scales. The FY30 ‘Rule of 50’ requires both revenue growth ($63 bn+) and margin improvement, supported by Customer Zero and the lean Agentic Enterprise; revenue attrition improved further in Q1, and net-new AOV continues to accelerate, tracking slightly ahead of plan.

Q: What role does Slack play as an entry point in Agentic adoption, and what is the impact on large-deal bookings?

A: Miguel said Slack has become the preferred multi-user collaboration front end for builders and knowledge workers to access enterprise apps, with Slackbot lifting company-wide productivity by ~3%. Slack MCP calls mainly come from developers building apps that need Slack’s rich context, and Slack’s bookings and net-new AOV are strong with very low churn.

Srini added Slack is the most intuitive human-in-the-loop collaboration surface: in engineering channels, a developer posts a PR request and agents respond immediately; knowledge workers increasingly query and execute cross-app actions through Slackbot, making leading AI labs and engineering-heavy orgs the most active Slack MCP users. Marc concluded that Slack has evolved from a collaboration tool into a knowledge repository plus execution layer for external tools; Slack contributed nearly half of $1 mn+ deals with AWUs up 350% QoQ, and ‘in two years, agents on Slack will outnumber humans’, making a fourth $10 bn cloud after Sales, Service, and Data a matter of time.

<End of transcript>

Risk disclosure and disclaimer:Dolphin Research Disclaimer & General Disclosure