What happens when you multiply a single number, $0.99, by thousands of customer conversations every month? For teams evaluating Intercom’s Fin AI agent, that question sits at the center of every budget meeting. The headline price looks simple enough. The reality is messier, and the gap between the sticker price and the actual bill is where most planning mistakes happen.
This breakdown walks through how the per-resolution model works, why the math depends on a metric that different vendors define differently, and what finance and support leaders should model before signing anything for 2027.
Why the $0.99 Figure Deserves a Second Look
Intercom launched Fin in 2023 as an AI customer-service agent, and its launch pricing was widely understood as $0.99 per resolution. A resolution, in Intercom’s framing, means a conversation Fin handles end-to-end without handing off to a human. That definition is the entire foundation of the pricing model, and it is also the first place budgets go sideways.
Here is the uncomfortable part: the $0.99 figure comes from launch-era pricing. Pricing structures in the AI support space have shifted since then, with tiered and outcome-based options appearing across the market. Anyone building a 2027 budget on a 2023 number is making an assumption, not a calculation. Treat $0.99 as a starting reference point and confirm the current published rate with a dated source before it goes into a spreadsheet.
That caution is not cynicism. Intercom is a private company founded in 2011 and headquartered in San Francisco. It does not file audited financials, so every pricing claim in circulation traces back to the company’s own marketing pages or press announcements rather than regulatory documents. When a number changes, it changes quietly on a pricing page, not in a filing.
The Resolution Math, Step by Step
The core arithmetic is refreshingly simple. If Fin resolves a share of your conversations at a fixed price per resolution, your effective cost per conversation is the price multiplied by the resolution rate.
| Resolution rate | Cost per conversation at $0.99 | Cost per 10,000 conversations |
|---|---|---|
| 20% | $0.198 | $1,980 |
| 30% | $0.297 | $2,970 |
| 50% | $0.495 | $4,950 |
| 70% | $0.693 | $6,930 |
| 90% | $0.891 | $8,910 |
Those figures are arithmetic, not published rates. They show the shape of the model: cost scales with how much work the agent actually completes. A low resolution rate keeps the bill small but leaves most conversations for humans. A high rate shifts volume onto the AI line item.
The practical takeaway is that resolution rate is the real price lever, not the per-resolution fee. A team negotiating hard on $0.99 versus $1.10 while ignoring whether its resolution rate is 35% or 65% is optimizing the wrong variable.
A Worked Example
Picture a support desk handling 10,000 conversations a month. At a 50% resolution rate, Fin closes 5,000 of them. At $0.99 each, that is $4,950 in AI resolution costs, and 5,000 conversations still reach human agents.
Now suppose a configuration change, better knowledge-base content, or a narrower scope pushes the resolution rate to 70%. Fin now closes 7,000 conversations, costing $6,930, while human volume drops to 3,000. The AI line item grew by roughly $2,000, but 2,000 fewer conversations landed on the human team. Whether that trade is good depends entirely on loaded human cost per contact, which is usually far higher than a dollar.
Run that comparison before renewing. The number that matters is total cost per conversation across both channels, not the AI line item in isolation.
Where the Definition of “Resolution” Changes the Bill
The single biggest source of misleading math is the word resolution itself. Intercom’s definition, a conversation the agent resolves end-to-end without human handoff, is narrower than some competitors’ definitions, which may count tickets, sessions, or partial assists. Two vendors quoting similar per-unit prices can produce wildly different invoices because they are counting different things.
Before comparing any two proposals, get the definition in writing and ask three questions:
- Does a conversation count as resolved if the customer reopens it within 24 hours?
- Does a partial answer that ends in a human handoff count at all?
- Are billing disputes resolved by the vendor’s own classification system, and can you audit it?
That third question matters more than it sounds. If the vendor’s dashboard is the only source of truth for what counts as resolved, your finance team is auditing a black box. Ask for exportable resolution logs with timestamps.
This is the same discipline that applies when comparing AI output quality across tools. Teams that benchmark models without pinning down the scoring criteria end up with numbers that look precise and mean nothing. The same trap waits in AI support pricing. A related comparison of how vendors evaluate output quality, such as this look at DeepSeek vs ChatGPT vs Gemini judge output quality, shows how much the evaluation method shapes the conclusion.
The Hidden Costs Around the Per-Resolution Fee
Fin does not exist in a vacuum. The per-resolution charge sits on top of the platform costs that make the agent work. Budget for these alongside the headline rate:
- Seat costs. Human agents and admins typically still need platform seats. The AI fee is additive, not a replacement for the base subscription.
- Implementation and configuration. Knowledge-base cleanup, workflow design, and testing consume internal time or paid onboarding.
- Knowledge maintenance. Stale help content quietly drags resolution rates down, which paradoxically lowers the AI bill while raising human workload. Someone has to own content quality.
- Escalation handling. Every unresolved conversation still costs human time. A rising resolution rate is good; a falling one is a warning sign dressed up as savings.
The last point trips up a lot of teams. A dropping AI bill can look like a win in a monthly review when it actually signals that the agent is failing more often and humans are absorbing the overflow. Watch resolution rate and human volume together, never the AI line item alone.
Why Outcome-Based Pricing Feels Different
Per-resolution pricing belongs to a broader category called outcome-based pricing, where the vendor charges for completed results rather than access. It aligns incentives on paper: the vendor earns when the agent works. In practice, it shifts risk onto the definition of the outcome and onto the quality of the vendor’s reporting.
The practical tip here is unglamorous but effective: pilot for one full billing cycle before committing to annual volume. One month of real resolution data from your own conversation mix beats any vendor case study. Ask for the pilot to run on your actual traffic, not a curated demo set.
Comparing the Model Against Alternatives
Per-resolution pricing competes with per-seat, per-conversation, and flat-fee models. Each has a different risk profile.
| Model | What you pay for | Main risk | Best fit |
|---|---|---|---|
| Per resolution | Completed AI-handled conversations | Definition disputes, rate drift | High-volume, well-scoped support |
| Per seat | Human and admin access | Overpaying as AI reduces human need | Small teams, stable headcount |
| Per conversation | Every AI interaction | Paying for failed attempts | Early pilots, unpredictable volume |
| Flat fee | Predictable access | Overpaying at low volume | Budgets that need certainty |
There is no universally correct answer. A team with a narrow, repetitive ticket mix and a strong knowledge base usually benefits from per-resolution pricing because the agent resolves a lot. A team with complex, long-tail issues may find the resolution rate too low for the model to pay off, and a flat fee may be safer.
The comparison logic mirrors other hardware and infrastructure buying decisions, where the sticker price hides compatibility and workflow costs. Anyone weighing a platform migration knows the pattern, whether it is checking whether an AMD vs Intel CPU will fit an existing build or deciding between CMR and SMR NAS drives for a storage array. The listed price is the beginning of the analysis, not the end.
Building a Defensible 2027 Budget
Because no verified 2027 pricing exists in public circulation, any 2027 figure in a budget deck should be labeled a projection, not a quote. Here is a structure that survives scrutiny:
- Anchor on the last confirmed rate and note the date it was published.
- Model three resolution rates (conservative, expected, optimistic) rather than one.
- Add platform seats and implementation as separate lines.
- Include a sensitivity band of plus or minus 30% on the per-resolution fee to absorb rate changes.
- Define the audit method for resolution counts before signing.
That fifth step is the one most teams skip, and it is the one that saves the most money. A contract with a clear, auditable resolution definition is worth more than a small discount on the headline rate.
Common Mistakes That Inflate the Bill
Three patterns show up again and again in AI support budgeting:
- Forecasting on demo resolution rates. Vendor demos use curated conversations. Your traffic is messier. Discount the demo rate substantially.
- Ignoring reopens. A conversation that reopens in three days was not really resolved. If the contract counts it twice, the effective rate climbs fast.
- Treating the AI fee as the total cost. Seats, implementation, and content maintenance often rival the per-resolution spend in year one.
Each of these is fixable with a clause or a spreadsheet column. None of them require a legal team, just attention.
How to Pressure-Test Any Vendor Quote
When a proposal lands on the desk, a short checklist separates a solid deal from an expensive surprise. Confirm the per-resolution rate and the date it was last updated. Get the resolution definition in writing, including reopen windows and handoff rules. Ask whether the rate can change mid-contract and how much notice applies. Request sample resolution logs to see what the reporting actually looks like. Finally, ask what happens to pricing if your volume doubles or halves, because support volume is rarely flat.
Teams that run this checklist tend to discover the same thing: the per-resolution fee is the easy number, and the definition is the hard one. The $0.99 headline is a useful entry point into the conversation, but the contract lives or dies on how a resolution is counted.
For readers comparing AI evaluation approaches more broadly, the same principle applies across tools. A side-by-side of Gemini vs ChatGPT for evaluating output quality makes the point that scoring criteria drive results as much as the models do. Pricing works the same way: the metric drives the bill.
Conclusion
The $0.99 per-resolution figure is a clean, memorable number that hides a genuinely complicated calculation. Intercom’s Fin charges for completed AI resolutions, and the real cost depends on a resolution rate that varies by traffic, knowledge quality, and how the vendor defines a “resolution” in the first place. Multiply the fee by the rate, add seats and implementation, and audit the count before signing.
For 2027 planning, the honest position is that the headline rate should be treated as a reference point pending a dated confirmation, and every projection should carry a sensitivity band. Teams that model three resolution rates, demand a written definition, and pilot on real traffic will land closer to the truth than teams that multiply $0.99 by last month’s conversation count and call it a budget. The math is simple. The assumptions are where the money lives.
FAQ
Is $0.99 per resolution still the current Fin price?
That figure traces to Fin’s 2023 launch pricing. Pricing structures in the AI support market have shifted since, with tiered and outcome-based options appearing. Anyone budgeting for 2027 should confirm the current published rate and the date it was published rather than relying on the launch number.
What exactly counts as a resolution?
In Intercom’s framing, a resolution is a conversation Fin handles end-to-end without handing off to a human. Competitors define the unit differently, sometimes counting tickets or partial assists. The definition is the single biggest factor in whether a pricing comparison is fair.
How do I calculate my expected monthly cost?
Multiply your monthly conversation volume by the resolution rate, then multiply by the per-resolution fee. At 10,000 conversations, a 50% resolution rate, and $0.99 per resolution, the AI cost is $4,950. Run the same math at a low and high resolution rate to see the range.
Why does a lower AI bill sometimes signal a problem?
If the AI bill drops because the resolution rate fell, the agent is failing more often and humans are absorbing the overflow. Watch resolution rate and human conversation volume together. A shrinking AI line item paired with rising human workload is a warning, not a saving.
Should I choose per-resolution pricing or a flat fee?
It depends on your ticket mix. Narrow, repetitive issues with a strong knowledge base usually favor per-resolution pricing because the agent resolves a lot. Complex, long-tail issues may produce resolution rates too low for the model to pay off, making a flat fee safer.
Can the per-resolution rate change during a contract?
Possibly, depending on the agreement. Ask specifically about rate-change clauses, notice periods, and whether volume commitments lock the price. Getting this in writing before signing is far easier than renegotiating after the first surprising invoice.
