Back Office BeatSubscribe

Published by J9 Systems. What's new in business tech, and what it means for yours.

From the brief of Wednesday, October 28, 2026 · No. 15

From the daily brief · AI

Claude Sonnet 5.5 errors hit claude.ai and the API for 35 minutes on Oct. 27

A short Anthropic outage is no reason to switch vendors, but any automation that called Claude this morning may have failed without telling anyone.

By Carter Josephson · · 1 min read

What happened

Anthropic reported elevated errors on requests to Claude Sonnet 5.5 on Oct. 27, 2026, affecting claude.ai, the Claude API, Claude Code and Claude Cowork. The company said impact ran from 07:42 to 08:17 PT (14:42 to 15:17 UTC) and marked the incident resolved at 15:50 UTC.

What it means for a 10-to-50-person business

For staff typing into claude.ai, this was a 35-minute annoyance, from 8:42 to 9:17 a.m. Mountain. They retried and moved on. The risk sits in unattended automations. Take a 16-person accounting firm with a Zapier or Make workflow that sends each inbound client email to the Claude API for a summary and a tag. Assume 30 emails an hour arrive mid-morning. That is about 17 runs that could have errored in the window. Each one may now be an untagged email nobody has looked at. That matters in the weeks before year-end. Anthropic's status page gives no cause and says nothing about credits or lost requests. Care about this if you run Claude through Zapier, Make, n8n or custom code with no error alerts. You can ignore it if your team only uses Claude in the browser. One 35-minute incident says little about reliability. Look again only if these become regular.

What to do
Ignore

Make no plan or vendor changes, but if any Zapier, Make or n8n workflow calls Claude, open its run history for 7:42 to 8:17 a.m. PT on Oct. 27 and replay any failed runs.

What it costs
What it costs
Staff paused on claude.ai, assume 5 people × 35 min
≈ 3 hrs (≈ $150 at $50/hr), already spent
Checking run history and replaying failed runs
15–30 min (≈ $25 at $50/hr)
Adding an error alert to each Claude step, assume 3 workflows
1 hr (≈ $50 at $50/hr)
Anthropic credits for the outage
none stated on the status page
Doing nothing
$0 now; any failed runs stay unprocessed until someone notices
Sources
status.claude.com (primary)

Questions owners ask

Will Anthropic credit my account for the outage?

The status page does not mention credits. If you have an enterprise or committed-spend contract, check its service-level terms or ask your account rep.

Were requests or data lost?

Anthropic only describes elevated errors on requests to Claude Sonnet 5.5. Treat any request that failed in the window as not processed and resend it.

Should we add a backup AI provider?

Not over a single 35-minute incident. If Claude sits in a workflow that cannot wait half an hour, an error alert and a retry step cost less than running a second vendor.

Three to five of these, every weekday morning.

Related