What do AI support tools actually achieve? Every vendor publishes case studies; nobody lines them up. This page collects 93 named customer results from 51 vendors, pulls out the headline percentage, and keeps each metric separate, because a 70% containment rate and a 70% resolution rate are not the same claim. Across 27 published cases, the median claimed AI resolution rate is 67%, and the middle half of cases falls between 53% and 77%.
These are vendor-reported best cases, chosen for marketing. Vendors publish their happiest customers, measure on their own terms, and rarely say how long it took or which requests were in scope. Treat the medians as what a strong deployment can look like, not what you should expect in month one. Each figure links to its source so you can check the context yourself.
Customer-facing deployments only. Internal IT and HR help desks automate different requests and are listed separately. Metrics with fewer than 5 cases show the cases but no summary.
| Metric | Cases | Median | Middle half | Range |
|---|---|---|---|---|
| AI resolution rate | 27 | 67% | 53% to 77% | 26% to 91% |
| Automation rate | 25 | 70% | 50% to 80% | 34% to 90% |
| Containment rate | 7 | 74% | 64% to 78% | 45% to 84% |
| Deflection rate | 7 | 75% | 31% to 84% | 20% to 92% |
| Response time reduction | 6 | 81% | 70% to 86% | 47% to 90% |
| Resolution time reduction | 3 | — | — | — |
| Handle time reduction | 3 | — | — | — |
| Cost reduction | 7 | 64% | 40% to 69% | 25% to 95% |
Vendors use resolution, automation, containment and deflection almost interchangeably in marketing, but each counts something different, and the gap decides what you pay under per-resolution pricing. Read what counts as a resolution for how individual vendors define and bill it.
Share of conversations the AI resolved end to end, with no human involved. Median 67% across 27 cases from 19 vendors.
| Tool | Customer | Figure | What the vendor says |
|---|---|---|---|
| Posh AI | Florida Credit Union | 91% | 91% of requests resolved without a live agent Source |
| Replicant | DoorDash | 87% voice | 87% resolution rate; 350,000+ calls automated per day Source |
| HubSpot Breeze | Synergym | 85% | Customer Agent autonomously resolves 85% of support requests Source |
| Tidio Lyro | Borrowell | 83% chat | 83% resolution rate with Lyro Source |
| Sierra AI | Airtable | 80% | 80% AI resolution rate; workload equal to about 95 full-time agents Source |
| Tidio Lyro | Burker | 80% chat | 80% resolution rate with Lyro Source |
| Zendesk AI Agents | Action Property Management | 80% | 80% automated resolution rate; first response time down 81% Source |
| eesel AI | Gridwise | 73% | 73% of tier 1 tickets resolved in the first month Source |
| Fin by Intercom | Vanta | 71% | Fin resolves 71% of conversations; CSAT 96.7% Source |
| Decagon | Chime | 70% | 70% chat and voice resolution Source |
| Sierra AI | Chime | at least 70% | AI resolution rate rose from 50% to over 70% Source |
| Voiceflow | Trilogy | 70% | 70% of support tickets resolved by AI Source |
| Gladly Sidekick | The Black Tux | 68% | Gladly AI resolves 68% of inquiries overall (73% on email, 45% on SMS) Source |
| Amazon Connect | Zepz | 67% | AI agents resolve 67% of customer inquiries through self-service Source |
| Botpress | Super Dispatch | 66% | AI resolution rate tripled to 66%; cost per resolution cut to 67 cents Source |
| Parloa | GESOBAU AG | 66% voice | 66% of qualified calls resolved automatically Source |
| Fin by Intercom | Lightspeed | up to 65% | Fin resolution rate up to 65%; agents using Copilot close 31% more conversations daily Source |
| Richpanel | Aeons | 63% | 63% of conversations resolved end to end by AI Source |
| Gladly Sidekick | Ollie | 60% | AI agent resolves 60% of conversations, work equivalent to six additional full-time staff Source |
| Richpanel | Haus | 55% | 55% of conversations close with no human Source |
| Hyro | Inova Health | 50% voice | 50% of appointment management calls AI-resolved; 4,272 staff hours gained per month Source |
| Replicant | Sunrun | 50% voice | 50% of payment-related calls resolved by AI Source |
| Zowie | MediaMarkt | 50% chat | 50% chat resolution rate Source |
| Mavenoid | Stanley Black & Decker | 41% | 41% self-service resolution rate; 2,500+ conversations automated per week in peak season Source |
| Zowie | Aviva | 40% | 40% resolution rate within 2 weeks Source |
| Botpress | Extendly | at least 30% chat | AI-resolved chats rose from 5% to over 30%; 30% reduction in call centre volume Source |
| Gorgias | Psycho Bunny | 26% | AI Agent resolves 26% of customer tickets Source |
Share of total support volume the AI fully handled, so it is diluted by conversations the AI never saw. Some vendors also count conversations the AI only partly handled. Median 70% across 25 cases from 19 vendors.
| Tool | Customer | Figure | What the vendor says |
|---|---|---|---|
| Duckie | Grid | 90% | 90% ticket automation rate Source |
| Tidio Lyro | Axioma | 89% chat | 89% of conversations automated with Lyro Source |
| Yuma AI | EvryJewels | 89% | 89% automation; cost per ticket cut from $5.50 to $2 across 150,000+ tickets Source |
| Botpress | 24hassistance | 87% | 87% of support queries automated; NPS up 129% Source |
| interface.ai | Securityplus Federal Credit Union | at least 84% voice | 84%+ voice AI automation Source |
| Cognigy | BVG | 80% voice | 80% voice AI automation rate for route planning Source |
| Siena AI | Bad Birdie | 80% | 80% automation rate within one month Source |
| Zendesk AI Agents | TeamSystem | 80% | 80% automation rate Source |
| Five9 | MSU Federal Credit Union | 75% chat | AI Agents automate 75% of chat volume; call abandonment cut nearly in half Source |
| interface.ai | Del-One Federal Credit Union | 70% voice | 70% of calls automated Source |
| Kustomer | Vuori | 70% chat | 70% of chat conversations fully automated; 40% improvement in CSAT Source |
| Lorikeet | Hnry | about 70% | About 70% of conversations automated at peak tax week, from zero automation Source |
| Siena AI | Everyday Dose | at least 70% | Automates over 70% of shopper inquiries Source |
| Yuma AI | Clove | 70% | 70% automation and 3x ROI within three months Source |
| Zowie | Booksy | 70% | 70% of tickets automated, saving $600,000 a year Source |
| Zendesk AI Agents | Hello Sugar | 66% | 66% automation rate and $14k monthly savings Source |
| Plain | n8n | 60% | Handles 60% of tickets automatically Source |
| interface.ai | We Florida Financial Credit Union | 58% voice | Voice AI automates 58% of calls; staff reduced from 16 to 13 agents Source |
| boost.ai | DNB | at least 50% chat | Virtual agent automates 50-60% of incoming chat traffic and about 22% of all customer service traffic Source |
| Chatbase | Jumia | 50% | Chatbase handles 50% of all support requests Source |
| Gorgias | Orthofeet | at least 50% | Passed 50% automation after targeting 30% in 30 days Source |
| Help Scout | EmailOctopus | 50% | AI Answers handles 50% of support volume Source |
| Siena AI | Spanx | 50% | Siena handles 50% of conversations; average handle time cut in half Source |
| Parloa | MEG | 42% voice | Automates 42% of customer calls Source |
| PolyAI | Golden Nugget | 34% voice | Voice assistant handles 34% of all calls to central reservations Source |
Share of self-service contacts completed without escalation to a live agent. A contact center metric, most often quoted for voice bots and IVR. Median 74% across 7 cases from 6 vendors.
| Tool | Customer | Figure | What the vendor says |
|---|---|---|---|
| Posh AI | Citadel Credit Union | 84% chat | $663K+ annual savings and 84% chat containment Source |
| PolyAI | Carter's | 80% chat | 60% reduction in seasonal hiring, saving over $1M a year, with 80% chat containment Source |
| Amazon Connect | Centrica | 75% voice | 75% containment rate with a Nova Sonic-based voice bot Source |
| Cresta | Xanterra Travel Collection | 74% | AI agents average 74% containment and drove a $3.3M revenue lift Source |
| Posh AI | One Nevada Credit Union | 67% | 67% containment rate and $75K saved per month Source |
| Talkdesk | Quadient | 60% | 60% containment rate at launch, up from 33% with the previous system Source |
| Five9 | Exact Sciences | 45% voice | 45% containment rate and 20% overall call deflection Source |
Share of requests handled in self-service that agents would otherwise have handled. Definitions vary: some count abandoned contacts too. Median 75% across 7 cases from 6 vendors.
| Tool | Customer | Figure | What the vendor says |
|---|---|---|---|
| Duckie | Automox | 92% | 92% ticket deflection rate Source |
| Forethought | Grammarly | 87% | 87% deflection early on, implemented in 1.5 weeks Source |
| Decagon | Duolingo | 80% | 80% deflection rate Source |
| Salesforce Agentforce | Agibank | 75% | 22,500 loan status inquiries a month with a 75% deflection rate Source |
| Cresta | Snap Finance | 33% | Deflection rate rose from 6% to 33%; AHT fell 40% Source |
| Salesforce Agentforce | Trustpilot | 28% | Support backlog cut 99% within six months; 28% of support cases deflected Source |
| Five9 | Exact Sciences | 20% voice | 45% containment rate and 20% overall call deflection Source |
Before and after claims, as a percentage cut from the starting value. Figures marked * are computed from the before and after numbers the vendor states (for example, "from 7.5 hours to 4" is a 47% cut). Claims phrased as "X% faster" are left out because they do not convert cleanly into a reduction.
| Tool | Customer | Figure | What the vendor says |
|---|---|---|---|
| Decagon | ClassPass | 95%Cost reduction | 95% cost reduction Source |
| Richpanel | Ridge | 70%Cost reduction | 70% lower cost per ticket; CSAT up from 88% to 96% Source |
| Quiq | Brinks Home | 67%Cost reduction | Calls to human agents cut by two-thirds; cost per contact down 67% Source |
| Yuma AI | EvryJewels | 64%*Cost reduction | 89% automation; cost per ticket cut from $5.50 to $2 across 150,000+ tickets Source |
| Retell AI | SWTCH | at least 50%Cost reduction | Support costs cut by over 50% Source |
| Genesys Cloud CX | RedSalud | 30%Cost reduction | 20% more unique patient bookings and 30% lower contact center costs Source |
| Cognigy | Škoda Auto | 25%Cost reduction | 25% reduction in operational costs Source |
| Siena AI | Spanx | 50%Handle time reduction | Siena handles 50% of conversations; average handle time cut in half Source |
| Cresta | Snap Finance | 40%Handle time reduction | Deflection rate rose from 6% to 33%; AHT fell 40% Source |
| Capacity | DSW | about 20%Handle time reduction | Average handle time down nearly 20%; CSAT up 30% Source |
| Gorgias | Mybacs | 53%Resolution time reduction | Resolution time down 53% (from 15 hours to under 7 hours) Source |
| SupportLogic | Coveo | 53%Resolution time reduction | MTTR down 53% and escalation requests down 56% in the first six months Source |
| Forethought | Upwork | 50%Resolution time reduction | Time to resolution reduced by 50% Source |
| Sprinklr Service | 3M | 90%Response time reduction | Case response time reduced by 90% Source |
| Yuma AI | Glossier | 87%Response time reduction | 91% accuracy on complex shipping tickets; 87% drop in response times Source |
| Zendesk AI Agents | Action Property Management | 81%Response time reduction | 80% automated resolution rate; first response time down 81% Source |
| Plain | Tinybird | 80%*Response time reduction | Cut Enterprise first response time from 1 hour to 12 minutes Source |
| Plain | Sourcegraph | 67%Response time reduction | Cut first response time by 67% Source |
| HubSpot Breeze | ISSA | 47%*Response time reduction | First response time cut from 7.5 hours to 4 Source |
Employee-facing service desks resolve password resets, access requests and policy questions, which are more repetitive than customer requests, so their numbers are not mixed into the figures above.
| Tool | Customer | Figure | What the vendor says |
|---|---|---|---|
| Moveworks | Broadcom | 88%AI resolution rate | 88% autonomous resolution rate Source |
| Workativ | GoTo | 84%AI resolution rate | 84% of employee requests auto-resolved by AI across 5,000 employees Source |
| Moveworks | Mercari | 75%AI resolution rate | 75% of tickets solved autonomously Source |
| Unthread | Lemonade | about 40%AI resolution rate | AI automatically resolves about 40% of all internal tickets Source |
| Siit | Monzo | 60%Automation rate | 60% of inbound support requests automated; 2x faster first response Source |
| Moveworks | West Monroe | 40%Cost reduction | 40% cost reduction in one year, saving $1.4M in ticket services Source |
| Espressive Barista | Dexcom | 90%Deflection rate | 90% ticket deflection and 43% fewer phone calls to the help desk Source |
| Espressive Barista | Trimble | 73%Deflection rate | 73% ticket deflection Source |
| Freshservice | Five9 | 65%Deflection rate | Freddy AI virtual bot deflected 65% of tickets Source |
| Kore.ai | AMD | 50%Deflection rate | 80% reduction in HR resolution time; 50% self-service deflection Source |
| Siit | Qonto | 28%Deflection rate | 28% of IT requests deflected Source |
| Leena AI | Coca-Cola | 88%*Resolution time reduction | HR, IT and finance ticket turnaround cut from 2 days to 6 hours Source |
| Kore.ai | AMD | 80%Resolution time reduction | 80% reduction in HR resolution time; 50% self-service deflection Source |
| Atomicwork | Pepper Money | 75%Resolution time reduction | 75% reduction in mean time to resolution Source |
Results come from vendor case studies, product pages and press releases, recorded with their source and the date we checked them. We keep the vendor's own hedging ("over", "about", "up to") and use the lower bound of any range. Results without a clean percentage, such as volume shifts, absolute counts or "X% faster", are kept on the vendor's listing but left out of these statistics. Medians and quartiles use linear interpolation (spreadsheet PERCENTILE.INC) and are rounded to whole percentages, so you can reproduce them from the CSV. Sources last checked September 2026.
99 rows, one per metric. Free to reuse under CC BY 4.0 with a link back to this page.
Cite this page
supportrequest.ai, "AI Customer Support Benchmarks (2026)", https://www.supportrequest.ai/benchmarks, data as of September 2026.
Are you a vendor with a published case study we have missed? Send us the link; we add results that name the customer and state a number.