YARDSTICK Take the audit Join the Waitlist

Yardstick Research tear-sheet / AI sales cohort

Methodology · how we score · rubric weights in plain sight · vendors received this sheet seven days before publication and could flag factual errors, never rankings

Salesforce Agentforce

Founded: 1999 (Salesforce, Inc.; NYSE: CRM). Agentforce launched at Dreamforce on Sept 17–19, 2024; GA Oct 29, 2024. [CITED — Salesforce GA press release] Headquarters: San Francisco, CA (Salesforce Tower) [CITED] Funding / financials: Public company. Benioff disclosed Agentforce annualized revenue "more than $500 million" growing "330% from a year earlier" on CNBC at Dreamforce '25 (Oct 14, 2025) — CEO-on-camera, vendor-claimed on the figure itself; Salesforce reports Agentforce as part of broader Data Cloud + AI bookings, not a separately audited segment. [CITED — CNBC] Generations: Agentforce 1.0 (Sept 2024) → Agentforce 2.0 (Dec 2024, Atlas Reasoning Engine + Slack-native) → Agentforce 3 (mid-2025) → Agentforce 360 (current; GA Oct 13, 2025) with Voice + Agent Script + expanded Builder. TDX 2026 (April 2026) revealed Headless 360 + Agentforce Vibes 2.0. [CITED — Salesforce press releases, Salesforce Ben] Archetype: Legacy CRM-native AI agent platform (Tier C in cohort — the most-watched legacy entrant; the bet that "agents, not seats, become the unit of revenue")

ICP (vendor-stated, verified against marketing): Mid-market and enterprise organizations already running Salesforce Sales Cloud / Service Cloud at meaningful scale, with a dedicated Salesforce admin or RevOps team and budget for Data Cloud. Named launch customers: Wiley, OpenTable, Saks, Wyndham Hotels & Resorts, ADP, The Adecco Group. [CITED — TechInformed; Salesforce customer stories.] Strongest customer-cited outcomes are Service / Marketing, not Sales SDR — every SDR Agent demo Yardstick located is on Salesforce's own channel; no independent customer SDR Agent walkthrough was found in the survey window.

Headline numbers

Observation Value Evidence
Total score (0–100) 30.0 (lowest in cohort; with category caveats on deliverability + data accuracy) weighted sum, scoring/scored-vendors.md
Personalization grade (sample prospect) 2/4 MEASURED — strong on inbound nurture / warm follow-up where CRM history is rich; weaker than Clay-driven workflows or Lavender-coached emails on cold outbound
Deliverability infrastructure score 0/4 MEASURED + CITED — absent by design (Agentforce rides the customer's existing email stack; no native warmup, no domain rotation)
Cost-per-seat efficiency 0/4 ESTIMATED — highest TCO in cohort by a wide margin; ~$210k–$260k Year-1 for a 25-rep team before Flex Credit overruns
Reply rate v2 — held-out test data forthcoming —
Cost per booked meeting v2 — held-out test data forthcoming —

Dimension scores (0–4)

Dimension Weight Score (0–4) Evidence
Personalization quality 25% 2 MEASURED — "Smart Account Assist runs on Azure OpenAI GPT-4o … pulls from 'the last 80 meetings/calls and the latest 500 emails' per account" (agentforce.md §"Personalization"); but "the per-prospect personalization depth is less impressive than Clay-driven workflows or human-Lavender-coached emails for cold outbound. The strength is on inbound nurture and warm follow-up where the CRM history is rich, not cold prospecting where the data is thin." Demo dossier: every SDR Agent demo located is inbound or warm follow-up; cold-outbound prospecting from a list, sequence-style, is not shown in any Salesforce demo Yardstick surveyed.
Deliverability infrastructure quality 20% 0 MEASURED + CITED — Category caveat: "No native warmup, no native domain rotation, no native inbox-placement scoring — defer to whatever the customer's existing email stack provides" (agentforce.md §"Deliverability"). 0 because absent-by-design.
CRM integration depth 15% 4 CITED + MEASURED — "Salesforce-native: Agentforce is Salesforce — bi-directional read/write on every standard and custom object, governed by the customer's existing field-level security, sharing rules, and permission sets. No 'sync' because there is nothing to sync" (agentforce.md §"CRM integration"). Tied with Outreach/Salesloft on Salesforce-only depth; loses points for HubSpot/Pipedrive non-support.
Cost-per-seat efficiency 15% 0 ESTIMATED — "Year-1 total: ~$210k–$260k before any Flex Credit overruns" for 25-rep team; "Compare to 11x at ~$50k/yr per agent or AiSDR at ~$10k/yr per agent. Agentforce is not competing on TCO" (agentforce.md §"Reality-check TCO"). Highest TCO in the cohort.
Setup time 10% 0 MEASURED + CITED — "6–12 weeks for a production SDR/Service deployment, longer for orgs without an existing Data Cloud foundation." Salesforce marketing: "build an agent in minutes." Reviewer reality: 6–12 weeks. 0 because structural floor is weeks-to-months.
UI / UX 10% 1 MEASURED — "Reviewers consistently flag that the admin UI is dense and assumes Salesforce literacy; non-admin sellers see Agentforce primarily through Slack, Salesforce Mobile, or embedded panels in Lightning, where the experience is cleaner." Independent walkthrough (demos/agentforce.md #19) confirms admin-UI density outside stage demos.
Data accuracy 5% 0 MEASURED — Category caveat: "Salesforce does not own a third-party data graph the way Clay or ZoomInfo does, so cold-outbound research depends on whatever external connectors the customer wires up." Atlas reasoning-engine "33% accuracy / 2x relevance" claim is repeated as slide-bullet at the Agentforce 2.0 launch but methodology is never substantiated on-camera — VENDOR-CLAIMED only. 0 because absent-by-design for cold outbound.
Reply rate (v2) — v2 forthcoming —
Cost per booked meeting (v2) — v2 forthcoming —

Pricing tiers (CITED — salesforce.com pricing pages, May 15, 2025 Flex Credits press release, vantagepoint.io, oliv.ai, getgenerative.ai, saastr.com)

Pricing rebuilt twice in 14 months. Three coexisting models as of April 2026:

Model Price Notes
Per-conversation (legacy, Sept 2024) $2 / conversation "Conversation" = 24-hour interaction window with single end user. Service/CX legacy model. "Flex Credits and Conversations will not be supported in the same org" — buyers must pick one.
Flex Credits (consumption, May 2025+) $0.10 / action ($500 / 100,000 credits = ~5,000 actions per pack) Per-action consumption. Foundations starter: 100,000 Flex Credits free for Enterprise Edition+ customers (one-time, ~5,000 actions).
Agentforce add-on (per-user, 2025) $125–$150 / user / month Unmetered access to Agentforce agents within licensed cloud
Agentforce 1 Edition (top-tier) $550 / user / month list (~$650 reported real) Bundles Sales Cloud + Service Cloud + 1M Flex Credits/yr + 2.5M Data Cloud credits/yr + unmetered Agentforce

Mandatory baseline costs (not optional):

25-rep mid-market Year-1 TCO [ESTIMATED, conservative]: - Sales Cloud Enterprise: 25 × $165 × 12 = $49,500/yr - Data Cloud: 25 × $250 × 12 = $75,000/yr - Agentforce add-on: 25 × $125 × 12 = $37,500/yr - SI implementation (one-time): $50,000–$100,000 - Year-1 total: ~$210k–$260k before any Flex Credit overruns. Compare to 11x at ~$50k/yr per agent or AiSDR at ~$10k/yr per agent.

Pricing instability flag: Two pricing reboots in 14 months ($2/conversation Sept 2024 → Flex Credits $0.10/action May 2025 → per-user add-ons $125–$550/seat). Buyers should expect more change. The Flex Credits announcement (May 15, 2025) was handled in writing and earnings-call commentary — no marquee video event located. The absence of a video moment for the pricing rebuild is itself the story.

Integrations

Implication: if you are a HubSpot shop, Agentforce is functionally not for you. Agent platform integration is structurally Salesforce-only.

Adoption reality (the headline gap)

Q1 FY26 disclosure (May 2025): ~8,000 of Salesforce's 150,000+ customers had Agentforce; ~4,000 paying; only ~800 in production (Salesforce Ben adoption analysis, Deep Analysis). That is roughly a 2–3% production-deployment rate against the customer base 18 months after launch, and only ~10% of paid Agentforce customers in production. By August 2025 earnings, Salesforce reported 18,500 Agentforce use cases (a softer metric than paid-production deployments) on a $10.9B quarter (diginomica). [CITED — all THIRD-PARTY.]

Editorial assessment

Agentforce is the most strategically loaded vendor in this guide and the most operationally cautious recommendation. Salesforce has bet the brand narrative on "the agentic enterprise" — Benioff stated in early 2025 that Salesforce would not hire engineers, claimed "30% engineering productivity gains" from Agentforce + AI, and framed the company's mission as "the No. 1 digital labor provider, period" (SF Standard). Twelve months later he was hiring 1,000 new grads to "build Agentforce and Headless 360" (Fortune, April 2026) — a meaningful softening. On Dec 4, 2025 Benioff told CNBC AI is "a commodity feature" — a notable rhetorical pivot from the Sept 2024 "agents are the new unit of revenue" framing. We capture both without speculating on his motives: the framing has a real effect on procurement-committee buying behavior at Salesforce-incumbent customers.

The product-vs-rhetoric gap matters. Yes: native CRM data, native Slack, Data Cloud grounding, Atlas reasoning, Voice + Agent Script in 360, single-vendor procurement, and enterprise audit are real assets that no AI-native startup can match. No: pricing has been rebuilt twice in 14 months, production adoption is ~10% of paid customers a year+ in, deliverability/cold-outbound is structurally absent, TCO is the highest in the cohort, and the SDR Agent specifically lags the Service / Marketing / Field Service personas in customer-cited outcomes — every SDR demo Yardstick located is on Salesforce's own channel; no independent customer SDR Agent walkthrough was found. The OpenTable "73% of all restaurant web queries after three weeks" outcome (CXToday) is service deflection, not sales outbound; no comparable cold-outbound customer metric exists in any video Yardstick surveyed.

Best for

Salesforce-incumbent mid-market and enterprise customers who: (1) already pay for Sales Cloud and have a Salesforce admin / RevOps team; (2) have or plan to budget for Data Cloud; (3) are buying an agent platform (sales + service + marketing + field service from one vendor with shared governance), not a point AI SDR; (4) have a use case in inbound nurture, lead qualification, account research for an existing book, or service deflection — not pure cold outbound; (5) prefer single-vendor, single-DPA, single-bill consolidation. Stage fit: Optimization (yes), Scale (conditional — Salesforce shops with Sales Cloud + Data Cloud already in place), Pilot (no — sandbox provisioning requires paid Sales Cloud contract), Foundation (no — $200K+ Year-1 TCO floor).

Not for

Any team that does not already run Salesforce, any team buying primarily for cold outbound, any SMB or Foundation-stage GTM team (Sales Cloud + Data Cloud baseline alone exceeds the entire 11x or Apollo bill), HubSpot/Pipedrive shops, buyers requiring contractual price stability (two rebuilds in 14 months).

Right of reply

Pending — to be solicited 7 calendar days before guide publication. Open diligence questions raised for the right-of-reply window: (1) production deployment rate as of most recent quarterly disclosure (has the May 2025 ~10% figure materially moved?); (2) which named launch customers are using the SDR/Sales persona specifically vs. Service/Marketing; (3) FedEx as an Agentforce reference (cited in trade press but no quantitative case study located — confirm or strike); (4) Flex Credit consumption math for a fully-automated outbound prospect-touch (research → draft → send → log) in production deployments; (5) price-protection clause for Agentforce 1 Edition customers given the two-rebuild track record.

Affiliate disclosure

No current affiliate relationship at compile time (2026-04-29). Tested on the same rubric as every other vendor in the cohort.