Research
Carrier Portal Automation Benchmark
Relay scored head-to-head against leading general-purpose AI agents on real HO-3 carrier portal work.
Scorecard
What we measured.
The run tested source-backed field fill, replacement cost, and rate completion on a live HO-3 carrier portal in June 2026.
91.9%
field accuracy on live portal work
3.6 min
median time to quote
1st
of 4 AI agents tested head-to-head
Field accuracy
Correct-fill rate averaged with error-free rate. Higher is better.
Time to quote
Median wall-clock per lead, steady state. Lower is better.
Output tokens per run
Tokens the model generates per run. Lower is better.
HO-3 new business quoting on a live carrier portal, June 2026. Source-backed field fill, replacement cost, and rate, scored head-to-head.
*Gemini CLI never completed a run, so its token count reflects unfinished work.
Methodology
How the run worked.
- Run the same HO-3 new business quote task against a live carrier portal.
- Score source-backed field fill, replacement cost, and rate completion.
- Measure median wall-clock time per lead after steady-state setup.
- Track output tokens because lower token volume means lower cost and less wasted model work.
- Keep the carrier and portal vendor anonymous on the public site.
Limits
What this does not claim.
- This is one carrier-portal workflow, not a claim that every carrier or line behaves the same way.
- The benchmark measures portal execution, not coverage judgment or binding authority.
- A human still reviews intake and returned quotes before anything goes to a client or carrier.
- AI Quoting remains a preview while we add carrier coverage and keep publishing measured runs.
AI Quoting is a Research Preview.
The benchmark is why we are building it in public. Agencies can start with Relay today, then work with us on the quoting preview when the carrier mix is a fit.
Get started
Get on the Summer ’26 list.
Spring ’26 is full. Get on the Summer ’26 list and we will be in touch as spots open.