Agentforce Help Agent: Pay-Per-Resolution AI Comes to Salesforce

For two years, the biggest objection we've heard from mid-market leaders about AI agents hasn't been "does it work?" It's been "what will it actually cost me?" Consumption-based pricing — tokens, credits, conversations — made budgeting feel like guesswork, and guesswork is exactly what a lean B2B operation can't afford. This month, Salesforce answered that objection in the most direct way possible: an AI service agent you only pay for when it finishes the job. Agentforce Help Agent, announced June 25 and generally available as of July 2026, charges per resolution — not per conversation, not per token. If the agent doesn't fully resolve the issue, or the customer asks for a human, you pay nothing. Here's what that means in practice, why it matters more for mid-market companies than almost anyone else, and how to figure out whether your org is actually ready for it.

1. What Agentforce Help Agent Actually Is

Help Agent is a prepackaged, autonomous customer service agent built on the Agentforce 360 Platform. The key word is prepackaged . Until now, getting real value from Agentforce generally meant designing your own agents: defining topics, writing instructions, building actions, and testing behavior — a meaningful project even for well-staffed teams. Help Agent flips that model. It ships with guided setup that Salesforce says takes minutes rather than months, grounds itself automatically on your existing Salesforce Knowledge articles, and comes with prebuilt actions for the workhorses of everyday service: case management, appointment scheduling, and order updates.

It's also channel-agnostic out of the box. One agent, configured from a single control screen, can handle conversations across web, customer portal, messaging, and voice. That's a quiet but important detail for mid-market teams, because deploying a bot per channel — each with its own logic and its own maintenance burden — is precisely the kind of sprawl that turns a promising pilot into an unowned mess eighteen months later.

Salesforce has been running a version of this on its own help portal, and the numbers it reports are worth noting: 4.3 million inquiries handled on help.salesforce.com with roughly a 70% autonomous resolution rate. Vendor-reported figures always deserve healthy skepticism, but as a proof point for "prepackaged agent grounded on a good knowledge base," it's a substantial one.

2. The Pricing Shift: From Credits to Outcomes

To appreciate why pay-per-resolution is a big deal, it helps to remember the road that led here. When Agentforce launched at Dreamforce 2024, pricing was $2 per conversation — whether or not that conversation went anywhere useful. In May 2025, Salesforce moved to Flex Credits, a consumption model that gave customers more granularity but still left many finance teams squinting at usage dashboards trying to forecast spend. One consultant memorably compared it to a 1991 cell phone plan, where every month's bill was an adventure.

Pay-per-resolution is the third act, and it's philosophically different. Under the new model — reported at a flat $2 per resolved issue — you're charged only when Help Agent takes an issue from start to finish on its own. If the customer escalates to a human, no charge. If the customer leaves negative feedback, no charge. Underlying consumption — the Data 360 queries and Agentforce processing behind each interaction — isn't metered separately during those conversations. Salesforce's own product leadership framed the goal plainly: let customers attribute spend to value.

Why outcome pricing favors the mid-market

Enterprises can absorb a fuzzy AI bill as an R&D line item. A 150-person B2B company can't. Outcome-based pricing does three things for a mid-market budget: it makes the unit economics legible (cost per resolved case is a number your CFO already tracks), it shifts risk onto the vendor (a badly performing agent costs you little), and it makes the build-vs-buy conversation honest, because you can compare $2 per resolution directly against your current fully loaded cost per case. For many service teams, that internal number lands somewhere between $8 and $25 once you account for salaries, tooling, and overhead — which is exactly why this pricing model is getting attention.

3. A Timely Signal, Not Just a Product Launch

Help Agent's general availability isn't happening in a vacuum. On July 24, Salesforce announced a $1.6 billion contract with the U.S. Department of Veterans Affairs to modernize veteran services — one of the largest signals to date that agentic AI on the Salesforce platform is moving from demos into large-scale, high-stakes production. When procurement processes that cautious commit at that scale, it usually means the technology has crossed a credibility threshold. And with Dreamforce coming September 15–17 in San Francisco, it's a safe bet that prepackaged agents and outcome-based pricing will be central themes — Help Agent looks like the first of a family, not a one-off.

For mid-market B2B leaders, the takeaway isn't "buy this today." It's that the era of AI agents as a science project is ending, and the era of AI agents as a line item with a measurable ROI is starting. That changes the questions you should be asking.

4. The Honest Caveats

Before anyone gets carried away, a few things deserve clear eyes.

• "Resolution" is defined by the vendor. A customer who gives up and closes the chat without escalating or complaining may still count as resolved. Watch your own CSAT and reopen rates, not just the resolution counter.
• The agent is only as good as your knowledge base. Help Agent grounds on Salesforce Knowledge. If your articles are thin, outdated, or contradictory, you'll get confident answers to the wrong questions — at $2 apiece.
• B2B service is not B2C service. Mid-market B2B cases often involve contracts, entitlements, and relationships. The right early scope is the high-volume, low-complexity tier: order status, scheduling, password and access issues, documentation questions.
• Pricing details will evolve. This is the third pricing model in under two years. Build your business case on ranges, not single points, and revisit it after Dreamforce.

5. A 6-Step Readiness Checklist for Mid-Market Teams

Whether you adopt Help Agent this quarter or just want to be ready when outcome-priced agents mature, the preparation work is the same — and none of it is wasted, because every step below also improves your human-powered service today.

1. Audit your knowledge base first. Pull your top 50 case drivers from the last 12 months. For each, ask: is there a current, accurate Knowledge article a machine could act on? The gap between that list and your article library is your real project plan.
2. Clean the data the agent will touch. Duplicate accounts, stale contacts, and inconsistent case categorization all degrade agent performance. An agent scheduling an appointment against the wrong account record isn't a productivity gain — it's a new problem.
3. Map your escalation paths deliberately. The moments Help Agent hands off to a human are where trust is won or lost. Decide who receives escalations, with what context, and how fast, before the agent goes live.
4. Model the economics honestly. Calculate your current fully loaded cost per resolved case. Compare it to $2 per autonomous resolution at a conservative resolution rate — assume 30–40% for your use cases, not the 70% Salesforce reports for its own portal.
5. Pilot narrow, measure wide. Start with one channel and two or three case types. But measure everything: resolution rate, CSAT, reopens, escalation quality, and agent-handled volume versus your baseline.
6. Assign a human owner. Autonomous doesn't mean unmanaged. Someone on your team should review agent conversations weekly in the early months, feeding corrections back into Knowledge. This is the single most predictive factor we see in whether AI service tools compound in value or quietly decay.

Where This Is Heading

Pay-per-resolution pricing is Salesforce betting that its agents are good enough to be paid like employees: for finished work. That confidence will spread — expect prepackaged, outcome-priced agents for sales development, scheduling, and collections to follow. Mid-market companies that spend the next two quarters getting their knowledge, data, and processes agent-ready will be able to adopt each of these in weeks. Companies that don't will keep funding pilots that stall.

The honest first step isn't buying anything — it's knowing whether your org is ready. That's what we do at 360 Quantum Solutions: fixed-fee projects, certified consultants you talk to directly, and a go-live guarantee we put in writing. Start with our free 2-minute org health check at 360quantumsolutions.com, or book a 5-day org audit and we'll tell you exactly where your Salesforce org stands — and what an agent-ready roadmap looks like for your team. Your Salesforce should be your #1 sales tool. Let's make it one.