April's .40 Innovation Showcase Recap

Welcome to the .40 Release Innovation Showcase Recap!
Once a quarter, we get with the product team to highlight our top new features. The .40 Innovation Showcase did not disappoint, and if you missed it, this recap is your next best option. (Key word: next best.)

To watch the full recap, sign in here.

Where Are You on the Journey?
Before diving into features, Micah grounded the whole session with something worth sitting with: a map of what the path from task bots to full agentic process automation actually looks like. The short version: task bots aren't going anywhere. They're the foundation. But as you layer in intelligent automations, processes, and AI agents, you stop thinking about cost savings alone and start thinking about how automation can actually move business levers: revenue, strategic differentiation, the stuff that gets leadership excited. A customer Micah met with the week before was saving $2M a year with task bots only. The projection for when they start layering in AI? Double that within a year and a half.

We polled the audience on where their programs sit today. The result was, as expected: all over the map. Which is exactly why this release had something for everyone.

The Demo: One Request, Five Products, Zero Manual Steps
The centerpiece of the session was a live end-to-end demo built around an IT hardware request, and it was a masterclass in how all of these capabilities connect.

Here's the flow. An employee opens Microsoft Teams, has a conversation with an Aisera AI agent, and requests a new laptop. Aisera already knows who they are, what department they're in, who they report to, and what they're eligible for, so the conversation is short and smart. No logging into ServiceNow. No form hunting. Just a chat that ends with a ticket automatically created and submitted.

That ServiceNow ticket creation is what Mozart Orchestrator was listening for. The moment it fired, a process kicked off inside A360. From there, goal-based agents took over: one for request routing, one for policy review. The policy agent was grounded using the new Grounded by Aisera model connection, which meant it wasn't just running a prompt against a generic model. It was working with the actual org policies, approval thresholds by job function, hardware categories by department, the real stuff. The difference between a grounded response and an ungrounded one was visible in the demo, and it was stark.

From there, the process split. Auto-approve if everything checks out. Escalate to a manager if it doesn't. Parallel branches. Merge point. Asset check. Device fulfillment. Done.

The Feature That Might Change How You Build
AI Evaluation, or evals, is new in .40 and it's worth paying attention to. Rinku walked through what it actually does: automated testing pipelines for your skills and agents before they go to production, and ongoing monitoring after they do.

For skills, you're looking at accuracy scores and LLM-as-a-judge reasoning that tells you not just whether something passed or failed, but why. One test in the demo caught a hallucinated cost value that would have sent a perfectly valid request down the wrong approval path.

For agents, it goes deeper. Goal completion, tool selection, trajectory, which is a session-level metric that checks whether the agent followed the right steps in the right order. One iteration of the agent failed trajectory. Rinku refined the action plan, re-ran the eval, and watched the scores climb. That feedback loop is the whole point.

As Micah put it: this is how you open the black box.

Top .40 Features at a Glance
The session covered five headline features from this release: web triggers in Mozart Orchestrator so external systems like Salesforce or SAP can kick off processes directly; Agent Skills and Evals for testing and validating AI skills and agents at scale; Enterprise UI Agents, a new agent type that manipulates web UIs via prompt without any recording; the Grounded by Aisera model connection in AI Agent Studio; and OAuth support for MCP inbound, which means the automations, agents, and processes you've already built can now be exposed as tools to external systems like Microsoft Copilot, with enterprise-grade authentication.

What's Coming Up

• Imagine 2026 is happening May 18–20 at the Omni Dallas Hotel. Product announcements, hands-on sessions, the AI Automation Engineer certification track. If you're going, you already know.
• Product Club continues in May, June, and July covering COE Manager, AI Evaluation and Audit Logs, and Mozart Orchestrator.
• Developer Meetups are tackling Automation Cloud Service, using agents in the cloud, and a live build with Mozart Orchestrator.
• Listening on the go? Check out the Agentic Edge podcast.
• Want a tailored follow-up session on anything covered in this recap? Reach out. community@automationanywhere.com