Salesforce Charges $4 Per Resolved Ticket Now. What Happens to the Ones AI Can’t Solve?

Salesforce Agentforce just repriced the AI support market. Two to four dollars per resolved conversation. Per resolution. Not per seat, not per month, not per user. Per ticket your AI actually closes. Understanding the true AI support per resolution cost means looking beyond the sticker price to what happens at the edges of what AI can handle.

On paper, that sounds like a bargain. And it is, for the easy stuff. FAQ deflection, order status lookups, password resets. Your AI handles a thousand of those for four thousand dollars and your human support team never has to touch them.

The problem is what happens to the ones it can’t solve.

The AI Support Per Resolution Cost Nobody Is Talking About

A meaningful slice of enterprise and field support tickets can’t be deflected because they require someone to see something. Not read something, not search for something. See it. The technician in the field who needs to know if a valve is installed correctly. The customer whose equipment display shows an error code that isn’t in any manual. The installer who needs a visual confirmation that the connection is right before powering up. These aren’t edge cases. In field service, manufacturing support, and technical hardware, they’re closer to 30% of your volume.

Per-resolution pricing doesn’t make those tickets cheaper. It makes the gap between “deflectable” and “not deflectable” impossible to ignore.

What Happens When AI Loads Your Queue With Only Hard Cases

Here’s the math that support leaders are now staring at. If AI resolves 70% of your tickets at $4 each, great. But the remaining 30% still fall back to human agents. Except now your human queue isn’t handling routine stuff anymore. The AI already screened that out. What’s left is harder. The escalations are messier, the context is scattered across bot conversations and knowledge base dead ends, and your agents are handling a volume of genuinely difficult cases back to back with no easy wins to break the tension.

You didn’t make support cheaper. You made it more expensive per hard ticket, because you loaded your human queue with nothing but hard tickets.

The per-resolution model implicitly assumes a clean handoff. AI resolves what it can, humans handle the rest, everyone wins. But the “rest” isn’t a uniform category. Some of what AI can’t handle is just process friction – forms that haven’t been digitized, approval loops that haven’t been automated, workflows that a few engineering hours would fix. Those are solvable. The genuinely intractable ones are the physical, visual, in-the-field problems where no amount of AI chat can substitute for someone actually looking at the thing.

The Non-Deflectable Tier: Where Visual Support Becomes Infrastructure

I’ve been thinking about this a lot at Viewabo, where we build remote visual support for exactly these situations. What we see in the market is that the companies most affected by per-resolution pricing aren’t the ones with bad support operations. They’re often the ones with sophisticated support operations that already deflected everything deflectable years ago. The stuff left in their queue was already the hard visual and field stuff. Now they’re being asked to pay per resolution on tickets that were never going to resolve without a human with eyes on the problem.

The Agentforce pricing shift is going to accelerate a segmentation in the support software market that was already happening slowly. On one side, you have “deflectable support” – the knowledge base queries, the status lookups, the password resets. AI owns this category now and per-resolution pricing is a reasonable model for it. On the other side, you have “non-deflectable support” – the physical, visual, field, and technical escalations that require presence, context, and judgment. These tickets are getting more expensive to handle not because AI failed, but because AI success concentrated the hard cases.

What this means practically: if you’re running a support operation with meaningful field or technical support volume, you need a cost model for your non-deflectable tier, not just your deflectable tier. The instinct is to throw more agents at it. The better answer is to make each non-deflectable interaction faster and more effective – which is where visual support, screen sharing, and remote assistance tools start looking less like nice-to-haves and more like cost reduction infrastructure.

Rethinking Self-Service When Some Problems Can’t Be Self-Served

There’s a related shift worth noting in how support leaders think about self-service more broadly. Self-service deflection, like AI resolution, is only as good as your ability to identify which issues can actually be self-served. Visual and physical problems usually can’t be. When you push all your support volume toward self-service or AI resolution, you don’t eliminate the non-deflectable tier – you isolate it, and then you have to figure out what it actually costs.

Salesforce didn’t invent this problem. They just priced the market in a way that makes it visible. Four dollars per resolved ticket is a very clear number. It’s easy to build a business case around. What’s harder to price is the ticket that costs forty dollars in agent time because it required three escalations, a video call, and a field visit before anyone understood what was actually wrong.

Those tickets aren’t going away. They’re just getting more expensive to ignore.