In a regulated business, the riskiest support tickets are the ones a vendor has the least incentive to handle well. Outcome-based pricing is the only model that aligns the bill with the work that actually carries compliance risk.
Outcome-based pricing for AI support charges a fee only when the AI fully resolves a customer issue, rather than per seat, per conversation, or per message. In regulated industries like fintech, financial services, healthcare, and insurance, this model matters more than in generic CX because resolution is where compliance risk concentrates. A wrongly handled KYC unlock, dispute, or claim is not a churn problem, it is a regulator problem. Pricing per outcome puts the vendor's revenue on the same line as your obligation to get those tickets right.
Human-handled tickets cost roughly $1.25 to $4 each at scale, and far more for fraud, disputes, or regulated workflows that require senior or specialist review.
Outcome-based pricing is now the dominant procurement model for AI support: Intercom Fin charges $0.99 per resolution, Zendesk $1.50 to $2.00, and Sierra and Decagon negotiate per-outcome rates inside large annual contracts.
The buyer-side risk in outcome pricing is the definition of "resolution." If the vendor decides what counts, the incentive quietly shifts toward easy tickets and away from the hard, regulated ones.
The two procurement controls that matter most for regulated buyers are a customer-held veto on what counts as a resolution, and a guarantee that escalations to a human are never charged.
Gartner predicts agentic AI will autonomously resolve 80% of common customer service issues by 2029, up from low double-digits in 2024, which makes the pricing model a long-term decision, not a pilot detail.
Last updated: June 2026
For a regulated buyer, pricing is not a finance footnote. It encodes what the vendor is paid to care about. Per-seat pricing rewards keeping headcount busy. Per-conversation pricing rewards volume, including the conversations that should never have happened. Per-message and deflection pricing reward getting the customer to go away, which is the opposite of what a regulator wants to see in a complaint file. Outcome-based pricing is the only common model that pays the vendor for the thing you are actually on the hook for: resolving the issue correctly and being able to show your work afterward. This guide explains why that alignment fits regulated buyers, what to scrutinize in procurement, and how to model ROI against the real cost of a human-handled ticket.
What Is Outcome-Based Pricing for AI Support?
Outcome-based pricing charges a fixed fee each time the AI fully resolves a customer issue, and nothing when it does not. There is no per-seat license, no per-message metering, and no charge for conversations that escalate to a human. The unit of billing is the resolved outcome, which means the vendor is paid for finishing the job rather than for attempting it.
The model splits cleanly from the alternatives that dominated the chatbot era. Per-seat pricing bills for human agent licenses and treats AI as a feature add-on. Per-conversation pricing bills every time the AI engages, resolved or not. Deflection pricing bills for "containment," meaning the customer did not reach a human, regardless of whether their problem was solved. Outcome-based pricing is the only one of these where a half-finished ticket generates no revenue for the vendor.
Resolution: A customer issue the AI handled end-to-end to a defined completion state, such as a card disputed and refunded, a KYC document re-verified, or a claim status confirmed and updated, without a human agent finishing the work.
Escalation: A ticket the AI hands to a human, either because it hit a guardrail, lacked confidence, or the customer asked. Under a sound outcome model, escalations are not billed.
Lorikeet is an AI customer support platform built for complex, regulated companies in fintech, financial services, healthcare, and insurance. It prices per resolution at roughly $0.80 for a chat, email, or SMS resolution and roughly $1.00 for a voice resolution, with its Coach QA agent at around $0.25–$0.30 per ticket. Escalations are not charged, and the customer, not Lorikeet, holds the veto on what counts as a resolution. That last detail, who defines resolution, is the part of the model that regulated buyers should weigh most heavily.
Why Outcome-Based Pricing Fits Regulated Buyers
Regulated industries have a structural feature that generic CX does not: the hardest tickets are also the highest-risk ones, and they are rare. A consumer lender might resolve thousands of balance inquiries for every one wrongful-collections complaint, but the complaint is what shows up in a CFPB file. A health platform handles routine eligibility questions all day, and then one PHI-disclosure mistake becomes a HIPAA incident. The cost of these industries is not the volume of easy tickets, it is the tail of hard ones. Outcome-based pricing matters because it determines who has the incentive to handle that tail well.
Risk Alignment Between Vendor and Buyer
Under per-seat or per-conversation pricing, the vendor is paid the same whether the AI resolves a regulated ticket correctly or fumbles it into an escalation. Their revenue is decoupled from your risk. Under outcome-based pricing, the vendor only earns when the issue is genuinely resolved, which means they are paid to build an agent that can finish the hard workflows, not just the cheap ones. This is the closest a commercial contract gets to putting the vendor and your compliance team on the same side of the table. The vendor wants more resolutions; your compliance team wants more issues correctly closed with a record. When the resolution is defined correctly, those are the same thing.
The Customer Veto on Resolution
The single most important control in an outcome contract is who decides what counts as a resolution. If the vendor defines it, "resolution" can quietly mean "the customer stopped replying" or "the AI sent a plausible answer." For a regulated buyer that is unacceptable, because a plausible-but-wrong answer to a dispute or a claim is exactly the failure mode that draws regulator attention. The model only works in your favor when the customer holds the veto: you define the completion state for each workflow, and you can reject a billed resolution that did not meet it. Lorikeet structures its pricing this way, with the customer holding the veto on what counts as resolved. Without that veto, outcome pricing can become deflection pricing with a friendlier name.
Escalations Should Be Free
In regulated support, escalation is not a failure, it is often the correct and compliant action. An AI that recognizes it is out of its depth on a fraud case and routes to a human specialist is doing exactly what a regulator would want. A pricing model that charges for that escalation punishes the safe behavior and pressures the vendor, or your own team, to let the AI push through tickets it should have handed off. When escalations are not billed, the AI is free to be conservative on the tickets that warrant it. This is why "escalations are not charged" belongs in the contract, not just the sales deck. It removes the financial incentive to over-automate the riskiest interactions.
A Record You Can Defend
Outcome-based pricing also tends to pair with a model architecture regulated buyers need anyway. To bill a resolution honestly, the vendor has to know precisely what the AI did, which pushes toward complete, replayable logs of every tool call and reasoning step. That same audit trail is what supports your obligations during an examination or a complaint review. The pricing model and the compliance artifact reinforce each other: you cannot fairly bill a resolution you cannot describe, and you cannot defend a resolution you cannot replay. Buyers should confirm that the resolution log and the audit log are the same record, not two systems that can disagree.
Procurement Considerations for Regulated Buyers
Outcome-based pricing is an improvement on per-seat and deflection models, but it is not automatically buyer-friendly. The terms around the headline price decide whether the model protects you or quietly works against you. The considerations below are the ones a regulated procurement team should pin down before signing.
Define Resolution Per Workflow, In Writing
A "resolution" for a password reset and a "resolution" for a chargeback dispute are not the same event, and they should not bill the same way without scrutiny. Write a completion definition for each high-risk workflow: what state the ticket must reach, what record must exist, and what counts as a failed attempt that should not be billed. Make the customer the final arbiter. If a vendor resists letting you define and veto resolutions, treat that as the most important signal in the evaluation, because it tells you the model is built for their margin, not your risk.
Confirm Escalations and Retries Are Not Billed
Ask directly: if the AI engages a ticket, fails, and escalates to a human, is anything charged? If a customer comes back three times on the same unresolved issue, is that one outcome or three? A sound model bills nothing for escalations and does not let repeat contacts on the same unresolved issue inflate the count. Get the answer in the contract, not the demo.
Separate the QA Layer From the Resolution Fee
Quality assurance in a regulated business cannot be a sample. You need to know that every resolved ticket was reviewed, not one in twenty. Some platforms price an automated QA agent separately and cheaply, which lets you run 100% QA without it distorting the per-resolution economics. Lorikeet, for example, prices its Coach QA agent at around $0.25–$0.30 per ticket and can deploy it standalone for root-cause analysis and resolution verification. Treat the QA layer as a line item you want, not an upsell to resist, because in a regulated context proving correctness is the point.
Model the Total Cost, Not the Sticker Price
The lowest per-resolution number is not always the lowest total cost. A $0.99 resolution that only handles simple tickets, plus a per-seat helpdesk fee, plus a human team for everything hard, can cost more than a slightly higher per-resolution price that actually closes the regulated workflows. Build the model on your real ticket mix: what fraction of your volume is the hard, regulated tail, and which vendor can resolve it without escalating. The price that matters is cost-per-resolved-regulated-ticket, not the headline rate on the easy ones.
Check the Compliance Posture Behind the Price
Outcome pricing is necessary but not sufficient. Confirm the platform supports your obligations: SOC 2, BAA availability for HIPAA, GDPR alignment, PII redaction, role-based access, and data residency in the regions you operate. Confirm there are contractual no-train terms with the underlying model providers. A cheap resolution on a platform your compliance team cannot approve is not a saving, it is a stalled deployment.
ROI Versus the Human Cost Baseline
The case for outcome-based AI support rests on a simple comparison: what does a resolution cost a human team, and what does it cost the AI. The human baseline at scale runs roughly $1.25 to $4 per handled ticket, depending on complexity, geography, and how much specialist or compliance review the ticket type requires. Regulated tickets sit at the high end and beyond, because fraud, disputes, and claims pull in senior staff and longer handle times.
Against that baseline, a per-resolution price of roughly $0.80–$0.95 for chat, email, or SMS and roughly $1.20–$1.50 for voice changes the unit economics on the tickets the AI can close. The saving is not just the price gap per ticket. It is the gap multiplied by the volume of regulated tickets that previously required a human, plus the reduction in variance: an AI resolution that is logged and QA-checked has a more predictable compliance profile than a busy human team working a queue at 2 a.m. To make the ROI concrete is $1.00 per resolution all-in, against a human cost that, even at the conservative $1.25 end, would run $60,000 for the same volume, and far more for regulated ticket types.
The honest version of the ROI case includes what does not get automated. No regulated buyer should plan for 100% automation on day one, and a credible vendor will not promise it. The realistic model is that the AI resolves the high-volume and mid-complexity tickets, escalates the genuinely hard tail at no charge, and that tail stays with your human team. The ROI comes from moving the resolvable majority to a fraction of the human cost while keeping a clean audit trail, not from eliminating the human team. Buyers who model it as full replacement will be disappointed; buyers who model it as a shift in the cost curve, with the riskiest tickets still escalated, will find the math holds.
How Lorikeet Approaches Outcome Pricing in Regulated Support
Lorikeet was built for regulated buyers, and its pricing reflects the controls those buyers need. It charges per resolution, roughly $0.80–$0.95 for chat, email, or SMS and roughly $1.20–$1.50 for voice, with the customer holding the veto on what counts as a resolution and escalations never charged. The Coach QA agent runs at around $0.25–$0.30 per ticket, which makes 100% automated quality assurance affordable rather than a sampling exercise. The point of these specifics is not the numbers themselves but what they encode: the vendor is paid for finished, correct, reviewable work, and not for attempts, escalations, or deflection.
The pricing is backed by an architecture regulated teams can approve before launch rather than apologize for after. Lorikeet pairs natural-language and deterministic structured workflows, runs adversarial simulations before go-live, applies inbound message checks and outbound guardrails at runtime, and keeps replayable logs that support audit obligations. It resolves across chat, email, voice with sub-one-second latency, SMS, and WhatsApp on one engine, and it carries SOC 2, BAA availability for HIPAA, GDPR alignment, and US, UK, and AU data residency.
The honest limitation: this depth is not the fastest path to a live chatbot. A team that wants a simple FAQ deflector live this afternoon will find a drop-in tool quicker to stand up. Lorikeet's model assumes you have regulated workflows worth defining carefully and a compliance stakeholder who needs to sign off, and the configuration and validation that serve those needs take longer than a templated bot. For a regulated buyer that is the right trade. For a team without regulated risk, it may be more rigor than the use case requires.
Key Takeaways
Outcome-based pricing aligns the vendor's revenue with the resolution work that carries the most compliance risk, which is why it fits regulated buyers better than per-seat, per-conversation, or deflection models.
The model only protects you when the customer holds the veto on what counts as a resolution and escalations are never charged. Without those two controls, outcome pricing can drift into deflection pricing.
Define resolution per workflow in writing, confirm retries and escalations are not billed, and run 100% QA as a separate, cheap line item rather than a sample.
Model cost-per-resolved-regulated-ticket, not the headline rate. Against a human baseline of roughly $1.25 to $4 per ticket, a per-resolution price near $0.80 to $1.00 changes the unit economics on the tickets the AI can close.
Plan for a shift in the cost curve, not full replacement: the AI resolves the resolvable majority with an audit trail, while the genuinely hard tail still escalates to humans at no charge.
Conclusion
For regulated buyers, the pricing model is a statement about what the vendor is paid to care about. Per-seat and deflection models pay for activity and avoidance. Outcome-based pricing pays for the one thing a regulator will judge you on: resolving the issue correctly and being able to prove it. That alignment is the reason the model fits fintech, healthcare, insurance, and financial services better than it fits generic CX, where the stakes on a single mishandled ticket are lower.
The model is not self-protecting, though. The terms around the headline price, who defines resolution, whether escalations are free, whether QA is real, decide whether it serves your risk or the vendor's margin. A regulated procurement team that pins down those controls and models the total cost against the real human baseline will find outcome-based pricing is both cheaper and safer on the tickets that matter. If you are evaluating AI support for a regulated business, bring your hardest workflows and ask how each vendor defines and bills a resolution.
