What We Built and Learned at Pathfinder Pune

The Pathfinder Pune community got together in April for a day of demos, platform updates, and hands-on building — here's what went down.
 

Enterprise Claw Agent — Special Preview

Presented by Arjun Meda, Director of Community and Learning

This was the first public demo of the Enterprise Claw Agent, available in early access on the Agentic App Store. It's an autonomous agent that takes a plain English prompt and handles the rest — browser and desktop automation, file reading (PDFs, spreadsheets), and persistent memory across runs. Reasoning runs in the cloud (OpenAI or Anthropic) while automation executes locally via Playwright.

The live demo had the agent read expense PDFs, log into a mock SAP Concur app, and submit entries — one prompt, no predefined steps.

💡 Pro Tip: don't run Enterprise Claw Agent fully autonomously. Use it as a fallback tool inside a larger deterministic automation, with you controlling which capabilities (web, file access, Python, etc.) are enabled per run.

Product Updates

Presented by Deven Deshpande, Principal Product Manager 

The session covered a lot of ground on what's new in the platform. V40 brings meaningful updates across AI governance, developer tooling, and observability — here's the breakdown.

User Demo: Candidate Personality Insights Agent

Presented by Vatsal Mehta, Founder - HEDEHI Solutions and Most Valuable Pathfinder

Behavioral assessments are common in hiring, but interpreting them consistently across candidates takes time and introduces reviewer bias. Vatsal built an agent that automates the interpretation step while keeping humans in the loop. The agent reads trait outcomes (Strong / Solid / Developing / Risk Area) from a candidate's Excel assessment — no names, no PII — then uses an AI Skill to generate strengths, development areas, and suggested interview questions. A reviewer approves the output via an AA Form before a final PDF report is generated.

Things that came up:

• Descriptions on API tasks and variables are critical — that's how the agent decides which tool to use
• LLM output is probabilistic; grounding on your own data and prompt engineering improve consistency
• Human-in-the-loop is easy to add — just surface output to a review form before any action is committed

The Hands-On Session — Building a GTM Agent

Led by Shekhar Sharma, Sales Engineer and Arjun Meda, Director of Community and Learning

Using Automation Anywhere's AI Agent Studio, attendees built an AI agent to automate a real sales workflow end-to-end:

• Lead Capture — Extract lead data from emails or web forms and auto-populate the CRM
• Deduplication — Built-in checks identify and merge duplicate records
• Follow-Up Reminders — The agent schedules nudges so no prospect slips through
• Process Standardization — Consistent workflows mean every lead is handled the same way, every time

Low-code tools, drag-and-drop components, and prebuilt templates meant attendees could build, test, and deploy a working agent in the session. The takeaway: with AI Agent Studio and the Mozart Orchestrator, the agent development lifecycle has shortened considerably — prototype fast, ship quickly.

Wondering what’s next? Join us for Imagine India 2026 on 2nd July, in Bengaluru. This time - it's bigger, bolder, and built for what’s next. No hype, no fluff— just enterprise-grade AI agents, customer-validated outcomes, and hands-on experiences designed for real business impact. More than a conference, this is the moment India’s automation community comes together to build, learn, and lead.