Agentforce Pricing 2026: What Changed, What It Costs | Ksolves

Ask any Salesforce admin, RevOps lead, or IT buyer who has sat through an Agentforce pitch in the last two years, and you’ll hear some version of the same complaint. Ever since Salesforce first introduced Salesforce Agentforce pricing at Dreamforce ’24, the debate around it has never really settled. The benefits of Agentforce for businesses were never really in question; what businesses couldn’t agree on was whether the billing model behind those benefits made sense. That’s finally changing.

Salesforce has rolled out a new autonomous service agent, Agentforce Help Agent, paired with a pricing model that flips the old “pay for activity” logic on its head: pay-per-resolution. Instead of being billed every time a customer interacts with the AI, businesses are now billed only when the issue is actually solved, start to finish, without human help. Agentforce Help Agent and its pay-per-resolution pricing became generally available in July 2026.

If you’re evaluating Agentforce pricing models for your service desk, sales pipeline, or customer support channel, here’s what changed, why it matters, and how different industries can put it to work.

The Problem With the Old Agentforce Pay Per Conversation Model

If you’ve ever had to explain to a CFO why a chatbot cost the same whether it solved anything or not, you already know why this pricing model wore out its welcome. When Agentforce launched at Dreamforce ’24, Salesforce kept the pricing story simple: $2 per conversation. On paper, it sounded fair. You only pay when the AI actually talks to someone.

But the flat fee didn’t match how support conversations actually go. It billed a customer who got frustrated and demanded a human the same $2 as a customer whose problem got solved in one exchange, and enough customers pushed back that it forced Salesforce to rework the approach.

Some of the recurring complaints from customers and partners:

• Billed Regardless of Outcome: Whether the AI solved the problem, failed, or handed it off to a human, the meter still ran.
• Pay for Escalations: If a customer got frustrated and asked for a human agent, that failed interaction still cost $2.
• Unpredictable Data 360 Consumption: Every conversation could trigger unmetered Data 360 (formerly Data Cloud) consumption behind the scenes, creating overage risk that was hard to forecast.
• Weeks-to-Months Setup Time: Teams had to connect knowledge sources, define custom actions, and wire up every channel individually: voice, web, portal, messaging.

Only around 5,000 Agentforce deals closed in the model’s first two quarters, and just 3,000 of those were actually paid. Adoption was tepid at best. Analysts and partners were blunt about why.

As Ian Gotts of Elements.cloud put it at the time, companies didn’t know how to price the tool, and nobody wanted to sign a blank check without caps or predictable ROI.

Agentforce Pay Per Resolution vs Pay Per Conversation

Put the two models side by side and the difference stops being an abstract billing debate and starts looking like the exact conversation most procurement teams have been having for two years. The new Agentforce Help Agent, running on Agentforce pay per resolution pricing, addresses nearly every one of those pain points directly:

Salesforce’s own framing is direct: if a customer asks for a human or walks away unhappy, there’s no charge. Cost is tied to outcomes, not activity. It moves the bill off the interaction itself and onto whether the interaction actually solved anything, which is a different question than the one the old model was answering.

Salesforce is backing the claim with its own usage data: the company says Agentforce has handled 4.3 million inquiries on help.salesforce.com and autonomously resolved 70 percent of them, with everything learned from that deployment built into the new Help Agent.

How Does Agentforce Pay Per Resolution Work?

Ask three different Salesforce partners how a resolution actually gets counted and you’ll likely get three slightly different answers, which is exactly why it’s worth going straight to what Salesforce has published. The mechanics are simpler than the old model, even though the underlying agent is doing more work per interaction. Understanding how Agentforce pay per resolution works comes down to three conditions a resolution has to clear before it’s billable:

1. The agent handles the issue autonomously, start to finish, without a human stepping in.
2. The customer doesn’t ask to be escalated to a human agent at any point in the conversation.
3. The customer doesn’t indicate they were left unhappy with the outcome (through explicit negative feedback or by abandoning the interaction).

If any one of those three conditions fails, the interaction costs nothing. Salesforce collects nothing for an attempt that didn’t land, only for one that did. Data 360 (formerly Data Cloud) queries and Agentforce processing are both unmetered during the interaction itself, so there’s no separate consumption bill layered on top of the per-resolution fee.

Agentforce Pricing Models Compared

Anyone who has sat through a Salesforce quote knows the real question is rarely “what does Agentforce cost.” It’s which of the Agentforce pricing models actually applies to your rollout, and what that means for your total Agentforce license cost. Pay-per-resolution isn’t Salesforce’s only option anymore:

Conversations Pricing

The original $2-per-conversation model, billed on activity rather than outcome.

Help Agent Pay-Per-Resolution Pricing

$2 per resolution, billed only when the agent resolves an issue autonomously, with no charge for failed attempts or human escalations.

Flex Credits Pricing

Roughly $0.10 per action, sold in blocks of 100,000 credits for $500. Better suited to high-volume, simple-query use cases where per-resolution billing doesn’t fit as cleanly.

Per-User Licensing

The Agentforce license cost here starts around $125 to $150 per user per month for add-ons, or $550-plus per user per month for bundled Agentforce 1 Editions.

Agentforce has reportedly crossed $540M in ARR by Q3 FY2026, growing 330 percent year-over-year across roughly 18,500 total deals, with about 9,500 of those paid. The multi-model approach looks less like indecision and more like Salesforce letting different buyer types self-select the billing structure that fits how they actually plan to use AI agents.

Real-World Impact by Industry

None of this matters much on a pricing page. It matters in whatever queue your support team is actually staring at on a Monday morning. Here’s how the shift toward outcome-based, minutes-to-deploy agents plays out across different sectors:

Retail and E-commerce

High order volumes mean high interaction counts: “Where’s my order,” returns, sizing questions. Under the old model, every one of those queries was billable even if the customer bounced. With pay-per-resolution, a retailer only pays when the agent actually resolves the “damaged item, needs a replacement” case end to end. Reservation platform OpenTable is a useful reference point here. Its AI agent reportedly handled 73% of restaurant web queries within three weeks of launch, a 50% improvement over its previous tool, the kind of high-volume, repetitive query pattern that benefits most from outcome-based billing.

Financial Services and Insurance

Account balance checks, claims status, and card replacement requests are high-frequency, often simple queries, but compliance requirements mean escalation to a human is common for anything sensitive. The old model penalized that caution by charging for every escalated case. Under pay-per-resolution, insurers and banks can let the agent attempt straightforward requests, like a card freeze or claim status update, without worrying that a necessary human handoff turns into wasted spend.

Healthcare and Life Sciences

Appointment scheduling, prescription refill status, and insurance verification are strong candidates for autonomous resolution, but healthcare is also an environment where “when in doubt, escalate to a person” is often the right call. A no-charge-on-escalation model removes the financial disincentive to route sensitive cases to a human, which matters in a regulated environment. Publisher Wiley’s Agentforce deployment offers a directional data point outside healthcare itself: a reported 213% ROI and roughly $230,000 in savings, with case resolution improving over 40% in the first weeks.

Manufacturing and Field Service

Field service and B2B manufacturing support tend to involve multi-step troubleshooting: checking warranty status, diagnosing an equipment fault, scheduling a technician. These are exactly the higher-action, multi-turn interactions where the old $2-per-conversation flat fee could feel disconnected from actual complexity. Faster setup (minutes instead of months) also matters for manufacturers who want to pilot an agent on a single product line before rolling it out company-wide.

Hospitality and Travel

Booking modifications, refund requests, and itinerary questions are high-volume and time-sensitive. Engine, a hospitality tech platform, reportedly launched an Agentforce agent in 14 days and estimated $2 million in annual cost savings, a good illustration of how fast time-to-deploy translates directly into fast time-to-value in a seasonal, high-turnover industry.

Telecom and Utilities

Bill disputes, outage reporting, and plan changes generate enormous call and chat volume. Voice support in particular benefits from the Help Agent’s unified, prebuilt approach across web, portal, and voice channels, rather than configuring each channel as a separate project.

What Early Adopters Are Reporting

Salesforce’s own launch announcement names several organizations already deploying Help Agent, including PenFed Credit Union, Fisher & Paykel, and PowerSchool, spanning financial services, manufacturing, and education. A credit union, an appliance maker, and a K-12 software provider have little in common operationally, but all three had enough resolvable, repeatable support volume, balance checks, appliance troubleshooting, account questions, for the pay-per-resolution math to make sense.

Worth noting separately: Salesforce has also signed a definitive agreement to acquire Fin, a customer agent platform built for small and medium-sized businesses, with the deal expected to close in Q4 of Salesforce’s fiscal year 2027. Help Agent is built and priced for enterprise-scale deployments; Fin’s SMB customer base fills in the smaller end of the market Help Agent doesn’t currently reach.

What This Means If You’re Evaluating Agentforce

None of the above is worth much until someone on your team actually runs the numbers against your own ticket volume. A few practical takeaways for teams weighing this decision:

Resolution Rate Determines Your ROI

Pay-per-resolution is a better deal the higher your autonomous resolution rate is. If your knowledge base is messy or your use case is genuinely complex, you may resolve fewer cases autonomously. That’s fine, since you’re not billed for those anyway, but it does affect your ROI math.

Data 360 Readiness Is a Prerequisite

Salesforce’s own guidance is clear that Agentforce runs on Data 360 (formerly Data Cloud) for unified customer data, and a well-structured knowledge base is a prerequisite for good agent responses. Pricing flexibility doesn’t remove the need for clean data.

No Single Pricing Model Fits Every Use Case

A high-volume, simple-query deflection bot may still be cheaper on Flex Credits than on a per-resolution basis. A complex, multi-action support flow may favor the new Help Agent, and a broader rollout across departments might land better on per-user licensing instead. Your Agentforce implementation partners can help you model your total Agentforce license cost against your actual use case mix rather than defaulting to whichever model is marketed first.

Fast Deployment Still Requires a Ready Knowledge Base

The “minutes to deploy” promise assumes your Salesforce Knowledge content is already structured. Businesses without that in place will still need a setup phase, and your Agentforce implementation partners can help get that content ready before go-live.

Where Ksolves Fits In

None of the pricing flexibility above matters if the underlying implementation is sloppy. A cheap per-resolution rate doesn’t help if your knowledge base can’t get the agent to a resolution in the first place. Ksolves’ Agentforce implementation services help businesses:

• Assess which Agentforce pricing model (Conversations, Flex Credits, or the new Help Agent) actually fits their support volume, use case mix, and budget
• Structure and clean up Salesforce Knowledge and Data 360 (formerly Data Cloud) so autonomous resolution rates are as high as possible
• Configure and extend prebuilt Help Agent actions to match industry-specific workflows across service, sales, and field service
• Set up cross-channel deployment (voice, web, portal, and messaging) from a single, well-governed configuration

Getting paid only for resolutions puts a number on what used to be a vague promise: a support desk that actually finishes the job. Getting the setup right from day one determines whether that outcome-based pricing actually works in your favor.

Frequently Asked Questions