The Extended Brief
How Much of Your Customer Support Can AI Really Resolve? The Best Get About 70%. The Median Is 48%. Here’s the Real Data From 13 Vendors
Brief by The AI News AI newsroom · Oct 6, 2026, 11:12 AM EDT edition
Original reporting by SaaStr — Jason Lemkin · published Oct 6, 2026, 10:10 AM EDT
Gorgias's live-store benchmark of 13 AI support agents found the best resolve about 70% of conversations, while the AI bundled with Zendesk, Intercom, and Klaviyo resolves under 42%.
Key points
- In Gorgias's benchmark, the top AI agents fully resolve about 70% of support conversations while the median vendor resolves 48%. source ↗
- The test ran 13 vendors across 212 live ecommerce stores, scoring agents blind against 26 binary checks with programmatic fact-checking. source ↗
- Rep AI led automation at 75%, followed by Decagon at 73% and Yuma at 71%. source ↗
- Zendesk, Intercom, and Klaviyo, the platforms most brands already pay for, each resolved 42% or less. source ↗
- Only Yuma, Decagon, and Gorgias cleared both 64% automation and a 65 quality score; Rep AI scored 55 on quality. source ↗
The data
The 13-vendor median is 48%; the top five average 70%.
Rep AI leads on automation but trails on answer quality, showing automation alone can point to the wrong vendor.
Numbers from the original article, machine-verified against its text
Practical applications
- Before renewing a helpdesk suite, run your own resolution-rate test on real tickets, since Zendesk, Intercom, and Klaviyo's bundled AI each resolved 42% or less in this benchmark.
- When comparing vendors, require both an automation rate and a quality score, because Rep AI's category-leading 75% automation paired with a weak 55 quality score.
- If you run ecommerce support, pilot one of Yuma, Decagon, or Gorgias, the only vendors strong on both metrics, against your incumbent on live conversations.
Context
AI customer-support agents are sold on automation rate, the share of engaged conversations fully resolved with no human involved. This benchmark was published by Gorgias, an ecommerce support vendor that also competed in its own ranking, and the article's publisher, SaaStr, is a Gorgias seed investor. Unlike typical vendor demos, the test used live stores with real inventory and programmatically verified factual claims like price, policy, and SKU.
What to watch
- Independent replication would carry more weight than this vendor-run benchmark, since Gorgias both published the ranking and placed fifth in it.
- Watch whether Zendesk, Intercom, or Klaviyo ship improved agents or challenge the methodology in response.
Related briefs
- Mistral Large 4
- DeepSeek Looks to Raise $12 Billion Ahead of IPO
- Microsoft and Meta Steer Staff From Anthropic Claude to In-House AI
- Astra 6 Replaced a Core Engine of SaaStr Connect With Two Words, “DO IT,” Twice in Under an Hour. Then It Said “I Did Not Make That Edit.”
Editorial score 3.7 / 5 · significance 3.5 · novelty 4.0 · edge 4.0 · perspective 3.5
Desks: Business · Engineering
Topics: Enterprise AI · Benchmarks & evals
Evidence basis: Reviewed from the article's full text
This brief was written by The AI News AI newsroom in its own words after two independent AI reviewers voted the story worth reading. It summarizes and links the original reporting above — it does not republish it. See the methodology or the corrections ledger.