AI agents are no longer futuristic prototypes — they’re integrated, functional teammates embedded in customer experience, revenue operations, and support teams across industries. Whether it’s resolving Tier 1 queries, routing leads, or chasing unpaid invoices, the role of AI in operations is no longer “if,” it’s “how fast.”
But one piece of the puzzle still lags behind: pricing clarity.
As AI adoption accelerates, more buyers are asking not just how to deploy AI, but how to price it fairly — and how to avoid scaling success into a financial liability.
At the core of this debate lie two models:
- Resolution-based pricing
- Conversation-based pricing
Both sound reasonable. But when you dig deeper, their implications for performance, ROI, and operational strategy diverge sharply.
Let’s break down how each model works, what to watch out for, and why choosing the right one shapes your long-term automation success.
Why AI Pricing Still Feels Confusing
The confusion isn’t accidental — it’s a hangover from human agent pricing models, where value was tied to seats, shifts, or hours worked. You paid for time.
AI flips that logic: now, you pay for outcomes.
But many pricing models still carry legacy logic in disguise. Some vendors pad metrics with unclear definitions of “success,” while others use attractive terminology to hide unpredictable billing patterns. For buyers, it creates an apples-to-oranges environment where comparing vendors or forecasting costs becomes a full-time job.
The Two Pricing Models That Matter
You’ll hear different labels — session-based, event-based, usage-based — but nearly all AI pricing fits into two models:
1. Resolution-Based Pricing
You’re charged whenever the AI “resolves” a conversation. The intent sounds logical: only pay when the AI delivers value. But the devil’s in the definition. What counts as a resolution?
If one vendor defines it as “no human escalation” and another as “customer drops off after 5 mins,” are those both truly ‘resolved’ interactions? Or did the customer disengage?
2. Conversation-Based Pricing
You’re charged for each conversation the AI handles, regardless of whether it resolves or escalates. It’s a clean, volume-driven model. No fuzzy thresholds. Just simple metering.
The 4 Critical Evaluation Criteria
Let’s evaluate both models across the four questions that most impact buyer decisions.
1. Definitions and Mechanics
Resolution-Based:
No consistent benchmark. One vendor may claim a conversation is resolved if it doesn’t escalate. Another might base it on silence. This leads to the “containment trap,” where unresolved or abandoned queries are counted as successful.
Conversation-Based:
Transparent by nature. A conversation is a conversation. No guessing. No inference. That makes comparisons across vendors fair and billing more honest.
Bottom Line: Resolution-based is subjective. Conversation-based is standardized.
2. Ease of Charge Validation
Resolution-Based:
Teams often have to audit transcripts to ensure claims match reality. Mislabelled resolutions result in double-spend: once for the AI’s claim, again for human intervention later. The cost of accuracy becomes part of your ops load.
Conversation-Based:
Invoices match reality. Each handled chat or voice interaction is billable. No disputes. No reconciliation effort.
Bottom Line: Resolution-based demands effort to verify. Conversation-based is transparent by design.
3. Predictability of Cost
Resolution-Based:
Your AI improves → your bill increases. As resolution rates climb, you paradoxically spend more — even if total volume remains stable. This unpredictability complicates budgeting and penalizes performance.
Conversation-Based:
Cost is based on volume. If you know how many conversations you’ll handle, you can model spend down to the penny. Your AI gets better? Great. Same cost, more value.
Bottom Line: Resolution-based pricing introduces volatility. Conversation-based enables financial forecasting.
4. Incentive Alignment for Long-Term Scale
Resolution-Based:
This model misaligns incentives. Why improve AI if success triggers cost spikes? It turns automation into a growth tax — especially in high-volume environments like telco or eCommerce.
Conversation-Based:
This model rewards performance. Better AI = more deflection, faster resolution, reduced human handover — all at stable cost. Incentives stay aligned with business outcomes.
Bottom Line: Resolution punishes maturity. Conversation fuels growth.
A Real-World Example
Imagine you’re running an AI agent that handles:
- 500,000 conversations in Year 1
- Grows volume by 10% annually
- Starts with a 25% resolution rate, improving to 75% by Year 3
Pricing comparison:
- Resolution-Based: £1.25 per resolution
- Conversation-Based: £0.35 per conversation
By Year 3, your AI resolves far more — but your bill triples under resolution-based pricing. That’s the cost of doing better.
Is Resolution-Based Ever Useful?
Absolutely. Resolution is a fantastic performance metric. You should track it.
It helps answer:
- Are we deflecting the right volume?
- Is the AI answering clearly?
- Are customers satisfied without escalation?
But using it as a pricing mechanism mixes operational KPIs with financial models. And that introduces hidden risk, confusion, and billing friction.
Measure resolution. Don’t use it to drive cost.
Choose a Model That Works for Your Growth, Not Against It
AI isn’t static. Agents improve. Data gets cleaner. Workflows become more complex. Pricing should reflect that evolution — not penalize it.
The ideal pricing model:
- Scales predictably
- Is easy to validate
- Doesn’t punish performance
- Keeps incentives aligned across teams
That’s what conversation-based pricing does best.
Final Thoughts
You shouldn’t need a spreadsheet army to understand your AI agent bill. Nor should your CFO question why better performance led to surprise costs.
AI is meant to bring clarity, not confusion. And as adoption matures, pricing models must evolve to match — not mirror outdated paradigms.
Conversation-based pricing:
- Keeps operations honest
- Keeps finance predictable
- Keeps your AI agent on track to become your top employee
If you’re serious about AI maturity, don’t just compare features. Compare incentives. Choose pricing that scales with you — not against you.
