Operations | Monitoring | ITSM | DevOps | Cloud

You're already using AI without even thinking about it. That's the realization that kicks off this ShipTalk moment: Apple Maps quietly using on-device machine learning to learn your routes and driving habits — complete trust, zero thought. Which raises the real question: why aren't we there yet with AI in software delivery? The answer comes down to one word: guardrails. Consumer AI earned invisible trust. Shipping software hasn't — not until the guardrails catch up.

In this Harness University Tidbit, learn how to use a custom plugin step in a CD stage. You'll walk through using a plugin to update a task on a self-hosted ITSM service inside your private Kubernetes cluster.

The U.S. just barred foreign nationals from accessing two advanced AI models — Fable and Mythos — citing national security. Around the same time, the Five Eyes intelligence alliance warned that AI-enabled cyberattacks are "months, not years" away. In Season 5 of ShipTalk, host Adam and co-host Martin dig into whether that warning is already overdue — and what it means for the people actually defending software.

Today, we're launching Autonomous Worker Agents, AI agents that run as governed pipeline steps inside Harness. They inherit OPA policies, RBAC, audit trails, and scoped credentials from the first run. And because they live inside your Harness pipelines, they reason using the Harness Knowledge Graph: your services, deployments, incidents, and policies.

Engineering teams are burning through AI budgets with nothing to show for it — $100M across 10,000 engineers and no cost per run, no cost per outcome, just a number that keeps climbing. When it runs dry, your infrastructure upgrade gets cut. Harness ties every AI token to the outcome it created: cost per run, cost per resolved ticket, and anomaly detection before the invoice hits. One customer went from a $28,000 black box bill to $0.60 per ticket.

The era of token maxing is over. When Claude Fable 5 launched last week at $10/$50 per million tokens - double the price of Opus 4.8 - it was a clear reminder that the most powerful model isn't always the right model. Not every task needs the Ferrari. The fastest way to burn your Al budget is sending every request to the most expensive model by default. The real question for the next phase of Al cost management isn't "can this model do the job?" — it's "is it the right model for the job?".

Harness has been named a Leader in the 2026 Gartner Magic Quadrant for DevSecOps Platforms — for the third year in a row. Here's what stands out: almost every company in the Leaders quadrant is missing a piece. Some have Dev, some have Ops, but not the security. Harness is the one platform that brings Dev, Sec, AND Ops together — with AI built in. AI is changing how software gets built: more code, more automation, more agents — but also more complexity. Teams need delivery, security, testing, reliability, and cost in one place. That's what we're building every day.

Will AI replace engineers? Hugo Charré from Cursor argues the opposite. In this clip from the upcoming summit, Hugo Charré (Cursor) explains why the number of engineers is going to grow as we scale agentic systems — and why AI raises the bar for who's on the team.

Your entire AI stack ran on a model that disappeared in three days. The US government issued a directive suspending all access — a few hours' notice, no deprecation window, no roadmap. Launched Tuesday. Gone by Friday. And every enterprise that had built workflows on top of it just found out what the real risk was: not the model itself, but the absence of a governance layer underneath it.

Your best engineer spent 500,000 tokens last week. Nothing shipped. There's a name for it now: tokenmaxxing. Failed prompts, dead PRs, code that never reaches production — it looks like productivity, but it isn't. Most engineering leaders can't tell you what percentage of AI-generated code actually ships, or where the budget went. You should be able to say "that bug cost me $2,700 in tokens to fix.".