In the old world, “pricing” was a thing you could point at.
A phone system cost this much. A contact center seat cost that much. A migration cost whatever your patience could tolerate, multiplied by an hourly rate and a vague apology.
Now we’re pricing work performed by systems that don’t clock in, don’t ask for training, and don’t politely underperform until the third quarter. So we fall back to what procurement understands: seats, tokens, minutes, credits. Units that are easy to invoice, even when they’re hard to believe in.
That’s the quiet shift Mark Vange, Founder and CTO of Autom8ly, is describing: buyers don’t purchase “AI.” They purchase the moment the work is finished.
What Autom8ly Is Challenging About AI Pricing (and Why It Matters)
Autom8ly’s Mark Vange says token-based AI pricing breaks down in real buying cycles because it’s hard to see, hard to forecast, and difficult to package for customers.
He argues that as AI moves from “features” to completed tasks, pricing will shift toward resolution-based units that customers can predict, approve, and budget—especially in service provider channels.
“The final consumer pays for it by outcome… it doesn’t matter how many tokens it takes.”
Why Autom8ly Thinks Outcome-Based AI Pricing Beats Tokens for MSPs
Token pricing is rational if you’re building an internal model. It’s even rational if you’re an AI team trying to measure costs precisely.
It becomes irrational the moment you need to sell.
Because the unit you’re selling is not “inference.” It’s not “GPU time.” It’s not “tokens.” The buyer is not walking into a budget meeting asking for approval to purchase an unknown quantity of invisible math.
They’re asking for approval to eliminate a problem.
Vange’s framing is blunt: customers buy finished work, and anything else is vendor-centric accounting dressed up as transparency.
“If I come in and measure it in tokens… it’s nothing to you.”
What’s happening here is a collision between two worlds:
- The AI vendor world, where cost is granular, variable, and measured in micro-units.
- The service provider world, where value must be packaged, repeatable, and defensible to a customer who does not want surprise invoices.
That mismatch is why outcome-based pricing is attractive to MSPs and UC providers. It allows them to productize AI into something sellable. It turns “AI capability” into a line item.
And crucially, it makes the conversation boring again—which is exactly what procurement wants.
10DLC Onboarding as a Priced Outcome
Autom8ly’s example is 10DLC campaign registration in the US—something that’s mandatory for A2P SMS, complex for small businesses, and labor-heavy for service providers.
The claim is that MSP onboarding for 10DLC can take 4–6 hours of hand-holding, whereas their AI approach reduces that to a ~5-minute review by the MSP, with ~95% first-time submission success.
Instead of charging “for AI,” Autom8ly charges per successful campaign application. A fixed unit. A predictable SKU.
The deeper point isn’t just that they automated a form. It’s that they found a unit a customer can agree to without needing to understand how AI works.
“The price is based on this form getting submitted and passed.”
Related Story on CX Today
- The $50 Billion Question: How Do You Charge for an AI Agent Seat That Doesn’t Exist? (CX Today)
- Why Wall Street is Suddenly Nervous About Your Contact Center Vendor
Why Token-Based AI Pricing Fails at the Buying Table (Even When It’s “Fair”)
Token pricing has a seductive pitch: you pay for what you use. It’s measurable. It maps to cost.
That’s also its problem.
In most real-world buying committees, “fair” is not the top requirement. Controllable is.
Finance teams want:
- Forecasting accuracy
- Spend governance
- Contract clarity
- The ability to compare vendors without learning a new unit of physics
Token pricing struggles on all four.
It’s also psychologically backwards. Tokens ask the buyer to care about the vendor’s internal mechanics. But buyers don’t want to sponsor your infrastructure decisions. They want to pay for their own outcomes.
Even when vendors attempt to fix this with dashboards, the dashboards become part of the tax: another system someone has to watch to prevent budget surprises.
Outcome pricing is not perfect, but it has one massive advantage: it attaches spend to something legible. Something a line-of-business owner can point at.
If your AI can’t be pointed at, it will be treated like a variable risk.
“Being able to relate that operational cost… to outcomes… makes it that much easier for you to adopt this.”
The UC Angle: Why Outcome Pricing Fits Service Providers Better Than Consumption
UCaaS channels live and die by packaging.
They’re not just selling capabilities. They’re selling deployment, change management, compliance, adoption, and support. That’s why the “unit” matters more for them than it might for a hyperscaler.
The minute a UC provider tries to resell token-based AI, three uncomfortable questions appear:
- Who carries the overage risk?
If usage spikes, does the provider eat the cost, or does the customer get the bill shock? - Who can explain the bill?
Not “what caused it technically,” but who can explain it in a way that doesn’t sound like a shrug. - Who owns the outcome?
If the AI fails, do you refund tokens, credits, minutes, or something else?
Outcome-based pricing simplifies all three. It doesn’t eliminate complexity under the hood—but it prevents that complexity from leaking into the customer relationship.
It also aligns with how UC providers already sell: bundles, per-site fees, per-location pricing, per-activation services, onboarding packages, managed services tiers.
Outcome pricing is simply that mindset applied to AI execution.
“Understanding the use case… includes understanding the economics and the economic levers of that use case.”
What Outcome-Based AI Pricing Really Forces Vendors to Admit
Outcome pricing sounds like a pricing tactic. It’s actually a product maturity test.
Because once you charge for outcomes, you’re no longer charging for effort. You’re charging for reliability. That forces uncomfortable discipline:
- You need clearer definitions of “done.”
- You need tighter guardrails on edge cases.
- You need operational workflows that reduce exception handling.
- You need a failure policy that doesn’t become a margin crater.
Token pricing lets vendors hide behind variability. If the customer complains, you can say: usage went up.
Outcome pricing removes that excuse. If the customer complains, the implied question is: why didn’t the system complete the work as expected?
That’s why outcome pricing, when it works, is so compelling for channel partners. It turns AI into a sellable product instead of an experimental capability with a variable meter running in the background.
It also explains why the market keeps circling back to the same tension: buyers want predictable spend, vendors have variable compute, and neither side wants to hold the risk alone.
A recent CX Today piece framed the parallel problem in contact centers: pricing “seats” for agents that don’t exist breaks down as autonomous agents do more work, forcing pricing teams to rethink the unit of value. The unit is drifting away from people, and toward usage and outcomes—because that’s where the work is moving. Source.
“It doesn’t matter how many tokens it takes… the price is based on this form getting submitted and passed.”
The Plausible Future Drift: “Resolution Units” Become the New SKU
Here’s the likely drift, if Vange is right, and if buyers keep behaving like buyers.
First, the customer stops asking “how many tokens” and starts asking “how many resolutions.” Not because they’ve become more sophisticated, but because they’ve become more tired.
Then the provider packages AI the way they package everything else: standard units, predictable tiers, tight boundaries.
A resolution unit becomes a SKU:
- One submitted and accepted 10DLC campaign.
- One compliant collections call completed with required disclosures.
- One support case resolved without escalation.
- One onboarding workflow completed and verified.
This won’t be marketed as “resolution units,” at least not at first. It’ll be marketed as managed automation, or AI-enabled services, or “agents.”
But the commercial logic will be the same: outcome pricing quietly replaces compute pricing, because compute pricing makes the buyer feel like they’re paying for your uncertainty.
The darker drift isn’t dystopian; it’s administrative.
We’ll end up with contracts that look less like software licenses and more like manufacturing agreements. Defined outputs, acceptance criteria, service credits, and endless negotiation over what counts as “success.”
And once that happens, the question won’t be “how smart is the model?” It’ll be “how defensible is the invoice?”
Not because buyers dislike AI.
Because buyers dislike ambiguity with a payment schedule.
Keep up with UC Today
Join the UC Today Newsletter: https://www.uctoday.com/sign-up/
Join the UC Today LinkedIn Community: https://www.linkedin.com/groups/9551264/
If you’re seeing this pricing shift in buying cycles right now, share what questions your CFO or procurement team is asking—and what answers they’ll actually accept.
Follow the author on LinkedIn
FAQs
1) What is outcome-based pricing for AI?
Outcome-based pricing charges for a completed task or verified result rather than for compute metrics like tokens, minutes, or GPU hours. The goal is to align pricing with what buyers actually value: finished work. It’s often packaged as a fixed fee per resolution, submission, or completed workflow.
2) Why do buyers struggle to approve token-based AI pricing?
Tokens are hard to forecast and difficult to map to business value in budget conversations. Even if token pricing is “fair” from a technical standpoint, it can feel uncontrollable to finance teams. That creates friction in procurement and slows adoption.
3) How does outcome pricing help MSPs and UC providers sell AI?
Service providers need repeatable, packaged offers they can quote, deliver, and support. Outcome pricing turns AI into a predictable SKU that can be bundled into managed services. It also reduces bill-shock risk and simplifies customer explanations.
4) What’s an example of outcome-based AI pricing in practice?
Autom8ly prices its 10DLC onboarding automation by the campaign application outcome—effectively a fixed fee per successfully submitted and accepted registration. That’s easier for an MSP to resell than variable token consumption. The buyer pays for completion, not model effort.
5) Is outcome-based AI pricing always better than usage-based pricing?
Not always. Outcome pricing requires clear definitions of “success,” strong reliability, and agreement on edge cases and exceptions. Usage-based models can be simpler to implement for vendors, but they often create forecasting and governance issues for buyers.
6) How far could outcome-based pricing realistically go if left unchecked?
If buyers standardize on outcome units, AI contracts may start resembling operational service agreements with acceptance criteria, credits, and audited definitions of “done.” That could reduce ambiguity, but it could also create a new layer of contract complexity around measurement and attribution. The “seat” may survive as a legacy wrapper, while real value shifts into priced outcomes.
