← Case Library

Agentforce 360 (Salesforce)

Q3
Enterprise SaaS / CRM · Shape forming · The Mirage Star
Researched as of April 2026
ShareLinkedInX
Shape Forming — Star proposed, not yet confirmed

Downgrading from the on-record Star: the narrative explicitly debunks the ecosystem-coordination claim via its own '24-hour outage test' — an Agentforce outage would degrade customer-service deflection but core CRM, billing, and pipeline operations continue unaffected, meaning customers do not operate through Agentforce as infrastructure the way Star requires. The well-evidenced core is Q3 ticket-deflection automation (dominant share 0.647) at Salesforce-internal, Heathrow, Wiley, and Reddit deployments; the AgentExchange/hyperscaler-partnership language (Q4) is real but marketing-grade rather than platform-grade, and Salesforce's own late-2025 pivot to deterministic architectures after 'declining trust in LLMs' further undercuts the autonomous-coordinator framing.

AdaptAugmentAssist
Q3Automator
Q6Optimiser
Q9Conductor
Q2Tactician
Q5Partner
Q8Explorer
Q1Operator
Q4Weaver
Q7Navigator
StructuredDynamicComplex
3/9 cells occupied
Shape Read

Salesforce made Agentforce 360 generally available on 13 October 2025. That names the portfolio. It does not add a cell. Agentforce carries every rhetorical signature of a platform-coordinator Star, ecosystem language, named hyperscaler partners, a marketplace, and real financial momentum ($540M ARR, up 330% year over year), yet the operating evidence sits almost entirely inside Salesforce's existing CRM customer base rather than in genuine cross-party coordination. The strongest deployments, Salesforce's own internal customer service, Heathrow, Wiley, and Reddit, are all credible Q3 Automator ticket-deflection wins, not evidence that customers or partners operate through Agentforce as infrastructure. The clearest diagnostic is the 24-hour outage test: an Agentforce outage today would degrade customer-service deflection and internal productivity, but core CRM, billing, and pipeline operations would continue unaffected, the signature of marketing-grade rather than platform-grade dependency. Salesforce's own late-2025 pivot toward deterministic, rule-bound agent architectures, after publicly conceding declining trust in large language models, is an implicit concession that the autonomous-coordinator framing outran what the technology and the ecosystem could support.

Trajectory: The case remains an open test: continued AgentExchange and cross-customer growth could close the gap toward a real Star, while the December 2025 deterministic-architecture pivot signals an implicit walk-back from full autonomy claims. The 13 October 2025 general availability names Agentforce 360. It does not move this grid.

Inside the company · not a cell

Nothing quoted yet. A rate or a saving will sit here when it is on the record. It does not light the grid.

Initiative Map
CellInitiativeConfidenceEvidence
Not a cellAgentforce 360General availability, 13 October 2025🔶Salesforce announced the general availability of Agentforce 360 on 13 October 2025. The release names the portfolio. Customer rates on that page stay rates. The launch does not add a cell.As of October 2025
Q3AutomatorIndeed employer onboarding"Trusted, autonomous AI agents"🔶Routine question handling for employers (job posting flags, account verification); reduces time on simple repetitive tasksAs of April 2026
Q3AutomatorPepsiCo retail-store ordering agent"Agentic enterprise" framing🔶Used to help 1.5M retail stores manage orders; targeting 5M by end of 2026As of April 2026
Q3AutomatorVivint customer support"Autonomous agent" framing✅Agentforce sometimes failed to send satisfaction surveys despite explicit instructions; required co-built deterministic-trigger system with Salesforce to override LLM behaviour; ~2.5M-customer support baseAs of April 2026
Q3AutomatorSalesforce-internal customer-service agent (Customer Zero, help.salesforce.com)"Agentic layer speaking to the customer"; "Customer Zero for our new agentic service product"✅2.2M agent conversations vs 1.5M for human engineers; 84% case resolution; 4% intentional human handoff; >$100M annualised cost savings; "I don't know" rate dropped 30% to <10% via knowledge-base cleanup; supported reduction of customer-service headcount from ~9,000 to ~5,000As of April 2026
Q3AutomatorSalesforce-internal AI SDR / agentic sales"Agentic salesperson is calling every single person back"; "the agentic enterprise has arrived"✅100M previously uncontacted leads being called back; >10,000 conversations per week; >$1M in deals attributed to AI SDR; 49% productivity boost claimed for Salesforce ANZ teamsAs of April 2026
Q3AutomatorHeathrow "Hallie" passenger service"Digital travel companion"; agent that "acts as" companion✅90% chat resolution without human transfer; 40% efficiency gain; 50%+ fewer chat exchanges; 95% accuracy on case summaries; 30-second reduction in human chat handling time; deployed via WhatsApp on Service Cloud knowledge baseAs of April 2026
Q3AutomatorWiley customer service automation"AI-powered agents help customers resolve common issues"✅213% ROI; $230K savings; 50% faster onboarding for seasonal agents; >40% case resolution improvement vs prior chatbot; deployed during back-to-school peakAs of April 2026
Q3AutomatorReddit support deflection"Reddit deflected 46% of support cases"✅46% case deflection; 84% reduction in resolution time (8.9 to 1.4 min); 20% advertiser satisfaction liftAs of April 2026
Q1OperatorSaks "Sophie" — flagship launch demoAt launch: personalised concierge "powered by Agentforce"✅One year later: hotline has robotic IVR with basic transfer to human; new "Sophie" announced August 2025 in partnership with Amazon and NLX, not Agentforce aloneAs of April 2026
Q3AutomatorMid-market Agentforce deployment (Kristi Valente case)Implied "clicks not code"; easy out-of-box deployment🔶Purchased April 2025; not yet operational by November 2025; consultants themselves "confused"; mid-market Salesforce admins finding Agentforce too complex without specialist consultingAs of April 2026
Q4WeaverAgentExchange / Partner Network — third-party agent ecosystem"World's first agent ecosystem"; "premier marketplace"✅55 apps March 2025 to ~83 by May 2025 (5 net additions in two months — 14 added, 9 removed); ~200 partner solutions claimed at launch; required additional $50M partner investment in 2026 to drive volume; on a base of ~6,000 AppExchange appsAs of April 2026
Q4WeaverHyperscaler / model partnerships (AWS Bedrock-Anthropic, Google Gemini, OpenAI)"Open agent ecosystem"✅Announced as customer model choice inside Agentforce; same hyperscalers also support Microsoft Copilot Studio, Workday Agent Partner Network, ServiceNow agentic stackAs of April 2026
Q4WeaverSlack as "agentic operating system""The conversational interface for humans and agents to work together"; agentic OS framing✅Slack remains messaging-first product; native Agentforce in Slack and AgentExchange surfaced in Slack are real but Slack is still a channel, not an operating systemAs of April 2026
Q4WeaverAtlas Reasoning Engine / hybrid reasoning / Agent Script"Atlas Reasoning Engine"; "hybrid reasoning"; deterministic Agent Script✅Reasoning layer that classifies intent, retrieves data via RAG, calls LLMs, orchestrates Topics and Actions; Agent Script (Dreamforce '25) introduced as deterministic control layer above LLM after model-trust degraded; hybrid reasoning = LLM + deterministic rulesAs of April 2026
Q4WeaverAgentforce Builder / low-code natural-language tooling"Anyone can build an agent in minutes" / "clicks not code"🔶Used internally and by flagship customers; mid-market admins find configuration complex; 20-active-agent / 15-topics / 15-actions / 60-second-action-timeout architectural limits; no bring-your-own-model; no version control for agentsAs of April 2026
Q3AutomatorPricing architecture"Per-conversation," then "Flex Credits per action," then "Agentforce 1 bundled editions"✅$2/conversation at launch to $0.10/action via Flex Credits May 2025 to ~$550/user/month bundled "Agentforce 1" editions Oct 2025; pricing churned twice in 13 monthsAs of April 2026
Q3AutomatorTotal deal numbers / ARR / adoption depth"Fastest-growing product ever"; "one billion agents by end of 2025"✅18,500 cumulative deals; 9,500 paid; $540M Agentforce ARR (+330% Y/Y); $1.4B combined Agentforce + Data 360 ARR (+114% Y/Y); 3.2T tokens processed; <2% of customers having >50 conversations/week (summer 2025 internal data); 12,000 of ~150,000 customers (~6%) have any Agentforce deal; "one billion agents" target quietly retiredAs of April 2026
Published record

The lines this reading rests on. A date is the source date, when the page or the line names one.

Today, Salesforce (NYSE: CRM), the world's #1 AI CRM, announced the general availability of Agentforce 360, the culmination of a year of transformation, redefining how work gets done in the age of AI.

October 2025salesforce.com ↗
Strength
Salesforce's own internal customer-service deployment and named customer case studies (Heathrow, Wiley, Reddit) show real, measurable Q3 Automator ROI on structured-ticket workloads.
Fragility
The platform-coordination layer, AgentExchange and third-party agent volume, remains small and requires direct subsidy, while non-exclusive hyperscaler partnerships do not create the asymmetric dependency a genuine platform Star requires.
Failure mode
The company narrated 'world's first agent ecosystem' and 'agentic enterprise' language a full platform-maturity cycle ahead of the evidence, with flagship demos like Saks 'Sophie' failing to survive production on Agentforce alone.
Governance ceiling
There is no external regulator or market check on the platform-coordination claim itself; the correction so far has come from a quiet internal architectural pivot toward deterministic rules rather than any customer- or partner-forced reckoning.
← Back to the Case Library