Operations | Monitoring | ITSM | DevOps | Cloud

AI agents are quickly becoming part of the enterprise automation conversation because, among other things, they help teams move faster. But there is a major difference between an AI agent that sounds useful in a demo and an AI agent that is ready for production. Production agents need scope. They need to know what they own, which systems they can touch, which workflows they can run, which teams they support, and where the boundaries are.

We just added two new AI features to our app: natural-language translation for Error search and Insights queries. Honeybadger has two query languages: Error search speaks a simple token syntax in the spirit of Solr or a basic Elasticsearch query, while Insights runs on BadgerQL (BQL), our own language for digging into your event data, designed to feel familiar to CloudWatch Insights and Splunk users. Both are powerful, but sometimes you just want something that works without having to open up the docs.

You’re probably running a TLS configuration that the IETF says is “non-conformant”. But you didn’t do anything wrong. In July, the IETF published a pair of RFCs that took away three of TLS 1.2’s key exchange methods and froze the rest of it. The phrase they used is MUST NOT, the strongest thing a specification is allowed to say. Nginx, Apache, and Windows Server all ship with those key exchanges turned on by default. Nothing breaks tomorrow.

How long would it take you to get one new line of code into the container running your payment service? In a lot of organizations, the answer runs to weeks, because the change has to clear a build owner, a test cycle, and a release window that nobody wants to open early. That timeline is why so much monitoring advice fails on contact. Most of it opens by telling you to add a library, rebuild the image, and push a new version. If you could do that this afternoon, you would have done it already.

A second service call rarely looks like a strategic failure on a dashboard. It appears as another work order, another technician assignment, or another customer follow-up. That accounting view makes repeat visits look operationally normal, even when they are financially destructive. The first visit carries the visible cost of dispatch, labor, travel time, and parts handling. The second visit carries all of that again, plus customer downtime, SLA pressure, escalation risk, and lost technician capacity.

If you run Starlette or FastAPI in production with an APM tool, you should keep an eye on PR. It adds native OpenTelemetry HTTP server spans directly into the framework. No external instrumentor, no monkeypatching. Just spans emitted from the router itself. At Scout Monitoring, we instrument Starlette and FastAPI through our Python agent. A change like this touches how every APM tool in the Python ecosystem works with these frameworks, ours included.

AI coding tools were supposed to mean developers work less. On a recent webinar recorded with LeadDev, senior engineering manager Vernon put words to something a lot of teams are quietly noticing instead: “It’s concerning because it’s the opposite of what was promised. We were supposed to be working less.”

Every investigation, detection, dashboard, and AI-assisted workflow depends on one thing: data that teams can trust. But as environments grow more distributed, the data behind those experiences gets harder to manage. New applications, cloud services, security tools, infrastructure, and network devices constantly generate machine data, and each new source can introduce new formats, missing fields, inconsistent mappings, and pipeline changes that require expert attention.