A high resolution rate is a tempting number to put on a page. It's also the number we'd most want a customer to read carefully rather than just repeat, because resolution rate isn't a standardized term across this industry, and a single percentage can hide as much as it reveals.
For an AI receptionist, resolution rate usually means the share of conversations that ended with the customer's need met and no person involved. That sounds simple. In practice, the definition of "met," the mix of calls a business gets, and the way the number is averaged all change what it tells you.
If you're an owner looking at your own dashboard, the most useful habit you can build is reading the number by topic. If you're comparing vendors, the most useful question is how each one defines "resolved."
Every percentage in this piece is made up to show how the number behaves. None of them is an Aiva Chat performance figure or a customer benchmark, and your own will depend on your callers and how complete your setup is.
What is resolution rate?
Resolution rate is the percentage of conversations in which the customer's need was fully met without a human stepping in.
The important word is "need." A caller asking "are you open Sundays?" is resolved the moment they get the right answer. A caller trying to book a specific slot, check a policy exception, and confirm a price in one call is resolved only if all three actually landed. That's a much higher bar than "the conversation ended without incident," and it's not the bar every vendor uses.
The idea isn't new. Contact centres have long tracked first call resolution, which HubSpot describes as a company's ability to handle a customer's issue during their first outreach, calculated as total resolved cases divided by total cases. Resolution rate for an AI receptionist carries the same trap: it's easy to quote as one clean percentage, and easy to misread if "resolved" quietly means "the call ended."
Resolution rate matters most to small businesses deciding how much of their phone work an AI can take on, and to anyone comparing vendors. It's one of several AI receptionist performance KPIs worth tracking, and it's most useful read alongside the others.
How resolution rate is calculated
Step 1: Define what "resolved" means
The definition comes first. A strict definition counts a conversation as resolved only if the caller's need was met. A loose one counts anything that didn't end in a transfer. The strict version is harder to score and far more honest.
Step 2: Classify each conversation
Each call or chat gets an outcome: resolved, handed off to a person, message taken, or abandoned. In Aiva Chat, an outcome is set on every call, and the summary written afterwards records what happened.
Step 3: Divide resolved by total
Resolution rate is resolved conversations divided by total conversations for the period. Decide whether test calls, spam, and wrong numbers belong in the total, and keep that rule consistent.
Step 4: Break it down by topic
The headline figure is an average. Splitting it by topic, such as bookings, pricing, or cancellations, shows where it's strong and where it's weak. Aiva Chat tags each call with topics automatically, which makes that breakdown possible.
Why resolution rate is useful
It shows how much phone work the AI is really taking on
A rising resolution rate means fewer calls landing on your team. That's the clearest signal of whether an AI receptionist is doing the job you set up, rather than just answering and passing everything along.
It points to gaps in your information
When a topic resolves poorly, the cause is usually missing or vague information. Resolution rate by topic turns "the AI seems weak on refunds" into a specific list of answers to rewrite.
It tracks the effect of your fixes
After you rewrite an answer or save a correction, the topic's resolution rate should move over the following weeks. It's a simple way to confirm a change helped.
It frames the cost conversation
Calls resolved without a person are calls your team didn't have to take. Paired with your staff cost, that gives you a rough sense of time saved, which Aiva Chat's Results section estimates as Hours Saved and Money Saved once you set an hourly rate.
A healthy average can sit directly on top of one badly broken category, and never show it.
Reading resolution rate in practice
Three conversations, three outcomes
Problem: Three calls all "ended normally," but they aren't equal. How to read them: A caller asks for Thursday availability, picks a slot, and gets a confirmation: resolved. A caller asks about a cancellation policy your information doesn't cover, and the AI offers to connect them to staff: an honest handoff, not a failure, and not resolved. A caller asks an unusual version of a common question and gets an answer that sounds complete but is subtly wrong. Outcome: The third is the dangerous one. A loose measurement counts it as resolved, and only reading transcripts catches it.
The healthy average with a broken category
Problem: A business sees 88% resolution for the month and assumes all is well. How to read it: The topic breakdown shows hours and pricing questions resolving above 95%, bookings around 90%, and warranty questions near 65%, dragging the blend down without standing out. Outcome: The 88% was accurate; it just averaged away the problem. Rewrite the warranty answers and watch that one topic over the next week.
The dip that isn't a problem
Problem: Resolution rate falls after a new handoff trigger is added. How to read it: More callers are being transferred because they asked for a person or sounded upset. That's the system correctly handing off, as set up in your escalation rules. Outcome: Check why calls were handed off before treating a dip as bad news.
The first few weeks
Problem: A new deployment shows a modest rate on day two. How to read it: Early conversations surface phrasing and exceptions your information didn't anticipate. That's the calibration window any pilot goes through. Outcome: Judge the trend over several weeks, not the first day.
Resolution rate vs. other call metrics
| Metric | What it measures | What it misses |
|---|---|---|
| Resolution rate | Needs met without a person | Whether the answer was right or the caller satisfied |
| Handoff rate | Share of calls transferred to staff | Whether those transfers were necessary |
| First call resolution | Issues solved on the first contact | Satisfaction; also depends on a strict definition |
| Repeat contacts | Callers who come back about the same thing | Why they came back |
| Quality and sentiment | How the call went and how the caller sounded | Whether the need was met |
| Bookings and leads | Concrete outcomes from calls | Non-booking questions that were handled well |
Resolution rate and handoff rate are two views of the same split. A rise in handoffs where callers asked for a person shows up as a dip in resolution that's working as intended. A rise in handoffs because the AI couldn't handle a request looks identical in the top-line number but is a gap worth fixing. Only the reasons behind the handoffs tell them apart.
First call resolution is the contact-centre cousin of resolution rate. It's useful for comparing with a human team, but it measures whether an issue was solved on the first contact, not whether a person was involved.
Quality scores and caller sentiment catch what resolution rate can't: the call that technically answered the question but left the caller frustrated. In Aiva Chat, every call is rated for quality and caller sentiment, and calls that score poorly are flagged as Needs review.
How to track and improve resolution rate in Aiva Chat
Step 1: Set your business hours and staff cost
In AI Training, under Business Settings, set your opening hours and staff cost per hour. Results use these for after-hours counts and time-saved estimates.
Step 2: Watch Resolved Without a Person
The Results section of the analytics dashboard shows Resolved Without a Person next to Handoff Rate and Appointments Booked. Note your starting point.
Step 3: Tidy your topics
Topics are tagged automatically. Merge and rename them so the breakdown matches how you think about your calls, such as "bookings," "pricing," and "cancellations."
Step 4: Work through Needs review
Filter call logs to Needs review, read why each was flagged, and look for patterns in the topics that resolve poorly. The guide to reviewing calls and results covers each filter.
Step 5: Fix answers at the source
When the AI said something wrong, use "Fix this answer" on the transcript line to save a correction or add knowledge. A correction is used on the next matching call.
Step 6: Ask your calls
Use "Ask your calls" with questions like "Why did callers ask for a person last week?" The answer links to the calls it's based on.
Common resolution rate mistakes
Chasing 100%
A small share of calls genuinely needs a person. A system reporting 100% either isn't seeing those calls or is handling things it shouldn't. The goal is the right handoffs, not zero handoffs.
Reading only the average
The blended number hides the categories underneath. Always check the topic breakdown before deciding things are fine or broken.
Treating resolved as satisfied
A complete answer can still leave a caller unsure, so they call back. Watch repeat contacts and sentiment alongside resolution.
Judging on day one
Early numbers reflect gaps you haven't found yet. Give it several weeks of fixes before deciding.
Comparing vendors on headline numbers
Vendors define "resolved" differently, and some measure on curated demo traffic. Without the definition, the comparison is meaningless.
Resolution rate best practices
- Use the strict definition. Count a conversation as resolved only when the need was met.
- Read by topic, not just in total. The weakest category is where the work is.
- Pair it with handoff reasons. Separate intended handoffs from capability gaps.
- Spot-check transcripts weekly. Catch confident wrong answers the percentage can't see.
- Track the trend. Compare week to week, not day to day.
- Ask vendors for their definition. And whether the number comes from real traffic.
- Learn how the dashboard works. Our walkthrough of the resolution rate dashboard explains each figure.
Conclusion
Our position: resolution rate is worth watching, and worth trusting only after you've looked underneath it. The headline figure is a blend, and the topic that's quietly failing is almost never the one that moves it.
The caveat that matters most is the confident wrong answer. No percentage catches a call that sounded complete and wasn't, so a weekly spot-check of transcripts stays necessary however good the number looks. When you compare vendors, treat their headline figure as the start of a conversation about definitions.
This week, find the lowest-resolving topic in your calls and rewrite the answers behind it. Watching that one topic move will teach you more than the blended figure.
Ready to see your own resolution rate?
Numbers from someone else's calls won't tell you much. Yours will, usually within the first couple of weeks.
Start a 14-day free trial with no card, point some real calls at the assistant, and check Resolved Without a Person by topic at the end of each week. The Results dashboard, topic tagging, and Ask your calls come with every plan. If you'd like someone to walk through reading the breakdown with you, book a demo, or call +1 (970) 972-0974 to hear a call first.
Posts from the people building Aiva Chat in Rajkot — how the product works, what we have learned building it, and guides for getting the most out of an AI receptionist.