Deflection rate is the share of support conversations that ended without a human agent. Resolution rate is the share where the customer's problem was actually solved. Deflection measures the absence of a person; resolution measures the presence of an outcome. A conversation can be deflected and still unresolved, so read the two together: the gap between them is volume that left without a fix.
| | Deflection rate | Resolution rate | |---|---|---| | Formula | Conversations closed without a human ÷ all AI-handled conversations × 100 | Conversations with a verified fix and no re-contact within 48 hours ÷ all AI-handled conversations × 100 | | Counts | Conversations where no human joined | Conversations where the problem was fixed | | Easy to inflate? | Yes, in several ways | Hard, the outcome is real or it is not | | Tells you | How much volume the AI absorbed | How much volume the AI actually closed | | Read alone | Misleading | Incomplete but honest |
What is deflection rate?
Deflection rate is the percentage of conversations that ended without a human agent getting involved. The AI replied, the customer did not reach a person, the conversation closed. That is a deflection.
Notice what the definition leaves out: whether the customer's problem was solved. Deflection is about the absence of a human, nothing more. Some teams use a stricter version that also excludes customers who come back about the same issue, which is the definition in our glossary entry on deflection rate. The stricter the definition, the closer deflection moves to resolution. Most dashboards use the loose one.
What is resolution rate?
Resolution rate is the percentage of conversations where the customer's actual problem was fixed. The refund was issued. The address was changed. The order status was confirmed against live data and the customer left with the answer.
Resolution needs evidence. Either an action was completed in a connected system, or the customer confirmed they got what they needed, and in both cases they did not come back about the same issue within a set window. We use 48 hours.
Worked example: 1,000 AI conversations
Here is the calculation on an illustrative month of 1,000 conversations handled first by an AI agent.
| Outcome | Conversations | |---|---| | Handed to a human | 180 | | Closed without a human, verified fix, no re-contact | 610 | | Closed without a human, customer re-contacted within 48 hours | 90 | | Closed without a human, customer left with no confirmation | 120 | | Total | 1,000 |
Deflection rate = (610 + 90 + 120) / 1,000 = 820 / 1,000 = 82%
Resolution rate = 610 / 1,000 = 61%
Gap = 82% − 61% = 21 points, or 210 conversations that ended without a human and without a verified fix.
A useful third number falls out of the same table: resolution within deflected conversations = 610 / 820 = 74%. It answers a simple question: when the AI kept a conversation away from a person, how often did the customer actually get helped?
Now look at what the 210 conversations in the gap are. Ninety customers came back, so they already cost you a second ticket. The other 120 are unknown. Some got their answer and did not say so. Some gave up. You only find out which by sampling them, and that sample is the most valuable QA work you can do this month.
Why deflection rate alone misleads
Deflection rate is easy to move in the wrong direction. You can raise it without helping a single extra customer:
- Let the AI answer questions it should escalate.
- Hide the request-a-human option one click deeper.
- Count a conversation as deflected when the customer does not write back within 48 hours, which also counts everyone who gave up.
Every one of those moves makes the deflection chart go up. None of them solves anything. In the worked example, hiding the handoff button might move 60 of the 180 escalations into the "left with no confirmation" bucket. Deflection climbs to 88%. Resolution stays at 61%. The dashboard looks better and the customer is worse off.
Resolution rate resists this. You cannot fake a refund into a customer's account or an order that actually shipped. To move resolution rate you have to do the real work, which is why it is the harder, more honest number, and why fewer dashboards lead with it.
We make the opinionated case against leading with deflection in deflection rate is the wrong metric. This guide is the reference companion: definitions, formulas, and how to read the pair.
How to read deflection and resolution together
The signal is the gap.
- Deflection 80%, resolution 75%. The AI is doing close to real work. The 5-point gap is conversations that ended without a clear fix. Sample them, do not panic.
- Deflection 80%, resolution 40%. This is not an automation win. It is a backlog of unhappy customers who have not contacted you again yet. They will.
- Deflection falling, resolution rising. Often a good sign. It usually means handoff got easier and the AI stopped guessing on hard questions.
Track the gap over time. If deflection climbs while resolution stays flat, someone has tuned the AI to answer more and help the same. That is exactly the failure a deflection-only dashboard is built not to show you.
How to measure resolution rate honestly
A loose definition turns resolution rate back into deflection with a nicer name. Three rules keep it strict.
- Tie it to an outcome, not a reply. A resolved conversation has a verifiable result: an action in a connected system, or explicit customer confirmation. "The AI sent a plausible answer" is not resolution.
- Fold in re-contact. A conversation that looks resolved but produces a new contact about the same issue within 48 hours was postponed, not resolved.
- Split it by topic. A blended rate hides what the AI is good at. Order status, store policy, and account questions usually resolve cleanly. Complex returns and exceptions usually do not.
Resolution also depends on whether the AI can act, not only talk. An agent that reads live order data and completes the task will resolve things a text-only bot can only describe. Grounding answers in your real systems, which our grounding glossary entry covers, is what turns a confident reply into an actual resolution.
Where CSAT fits
Neither rate tells you how the customer felt. That is what CSAT is for, and it belongs next to the pair. Compare CSAT for AI-resolved conversations with CSAT for human-resolved ones. If they sit close together, the AI is resolving well. If AI-resolved CSAT trails badly, some of your "resolutions" are technically complete but unhelpful. Our guide to measuring CSAT for AI-handled tickets covers how to survey both groups fairly.
What good looks like
There is no universal target, because resolution rate depends on which questions you route to the AI. A team that sends only order-status questions should expect high resolution. A team that sends everything, including disputes, will see lower resolution, and that can still be the right setup.
So judge the pair, not the absolute number. Healthy looks like this: deflection and resolution move together, the gap is small and stable, resolution holds up when you slice by topic, and CSAT for AI-resolved conversations sits close to human-resolved CSAT.
The industry ambition is framed the right way. Gartner predicts that by 2029 agentic AI will autonomously resolve 80% of common customer service issues without human intervention. Note the verb. Gartner said resolve, not deflect. The forecast everyone quotes is a resolution forecast, and that is the number worth being held to.
How Keloa approaches this
We report both numbers and show the gap on purpose. Keloa's AI agents resolve by acting: they read live order and account data and complete the task, so a resolution is an event that happened, not a sentence that sounded right. When the AI cannot resolve something, it hands off rather than padding the deflection number with a guess.
Our pricing is per AI reply, not per claimed resolution, so we have no reason to inflate either number. The metric we want you watching is the gap.
Frequently asked questions
What is the difference between deflection rate and resolution rate? Deflection rate counts conversations that ended without a human. Resolution rate counts conversations where the customer's problem was actually solved. A conversation can be deflected without being resolved, which is why the two numbers can tell opposite stories.
How do you calculate deflection rate and resolution rate? Deflection rate = conversations closed without a human ÷ all AI-handled conversations × 100. Resolution rate = conversations with a verified fix and no re-contact within 48 hours ÷ all AI-handled conversations × 100. Use the same denominator for both so the gap is meaningful.
Is a high deflection rate good? Only if resolution rate is high alongside it. On its own, a high deflection rate can mean the AI answers well, or that customers give up. The number cannot tell you which.
What is a good resolution rate? There is no single target. It depends on which questions you route to the AI. Judge resolution next to deflection and by topic, not as one headline figure.
Should I stop tracking deflection rate? No. Track it, but never alone. Next to resolution rate, the gap shows how much volume left without a fix. By itself, deflection rate is the number that lets a problem hide.