Delhi NCR Pathfinder Meetup #2: Agentic Process Automation in Practice

Our second NCR Pathfinder meetup picked up where the first one left off, with more product updates and a lot more hands-on building. The room was full of practitioners who came to see what agentic process automation looks like in practice, and we spent the day watching AI agents do real work. This recap is a little late, but the timing works out well: we are hosting our next meetup in December, and this should give you a good flavour of what to expect.

Here’s what went down.

Product Session: Plan Smarter, Build Better, Scale Faster

We opened with the latest in Agentic Process Automation (APA) and the top features.

We began with definitions. “Agent” means slightly different things to different people, so it helped to settle on shared terms before going further. Automations are discrete, rules-based tasks that run the same way every time. Processes are orchestrated workflows that connect multiple automations, people, and systems from start to finish. AI agents are goal-based systems that reason, use tools, and adapt to the situation. Agentic Process Automation brings all three together, combining the reliability of deterministic automation with the flexibility of AI.

That idea ran through the whole session: each step of a process works best with the level of autonomy that suits it. The top features discussed:

• Mozart Orchestrator: event-driven execution, dynamic routing of human tasks, and an improved design canvas, so automations start the moment a business event happens and work reaches the right person automatically.
• AI Evaluation and Agent Logs: out-of-the-box scoring of agents, skills, and models before deployment, plus full traceability of agent activity for compliance.
• Enterprise UI Agents: browser-based agents that navigate through natural language, useful for systems that have no clean API.
• Aisera Integrated Model Connections: agents grounded in your own enterprise data, with verifiable citations.
• OAuth Support for MCP Inbound: secure authentication for third-party AI assistants that trigger automations remotely, and the ability for them to invoke goal-based agents directly.

With evaluations, observability, and governance built in, teams have a clearer path from experiments to mission-critical work.

Special Preview: The Enterprise Claw Agent

We gave attendees a look at the Enterprise Claw Agent, available on the Agentic App Store.

You give it a plain English prompt and it handles the rest. It can automate browsers and desktops, read files across PDFs and spreadsheets, and keep memory across runs. Reasoning runs in the cloud through OpenAI or Anthropic, while the automation itself executes locally through Playwright.

In our demo, the agent read expense PDFs, logged into a mock SAP Concur environment, and submitted the entries end to end, all from a single prompt with no predefined steps. The use case resonated with a lot of people in the room.

A useful design principle came up during Q&A. Enterprise Claw Agent fits best as a fallback tool inside a larger deterministic automation, with explicit control over which capabilities (web, file access, Python execution) are enabled for each run. Autonomy works best when it comes with guardrails, and the same point came up in the product session.

Hands-On Session: Building a GTM Agent

Attendees then got to building an AI agent that automates a real sales workflow end to end, using Automation Anywhere’s AI Agent Studio.

The agent captures lead data from emails or web forms and fills in the CRM automatically. Built-in checks find and merge duplicate records, and the agent schedules follow-up reminders so no prospect slips through. Because the workflow is standardized, every lead is handled the same way.

Low-code tools, drag-and-drop components, and prebuilt templates meant attendees could build, test, and deploy a working agent within the session. Together with the Mozart Orchestrator, the agent development lifecycle has become much shorter, which makes it easier to prototype quickly and ship.

Wrapping Up

Across all three sessions, one idea kept coming back: pair dependable, deterministic automation with AI where it adds the most value, and keep governance in place so you can see and audit what your agents do.

Thank you to everyone who joined, asked great questions, and built something along the way. We look forward to seeing you at Pathfinder #3 in December. Stay tuned!