PEX Network’s key takeaways:
- Rahul Chawla discusses the shift from robotic process automation (RPA) to process autonomy at tech giant google.
- Process autonomy, powered by agentic artificial intelligence (AI), mimics human judgment.
- The modern process excellence leader must evolve into the ‘architect of agency.’
Enterprises are moving beyond basic automation and into an era defined by intelligent, self-directed systems. In turn, the conversation is shifting from efficiency to autonomy.
Traditional tools like RPA have long promised speed and consistency, but their limitations become clear in complex, data-rich environments that demand judgment, adaptability, and trust. A new wave of agentic AI is now reframing what’s possible, enabling systems not just to execute tasks, but to reason through ambiguity and own outcomes end to end.
In this interview, Rahul Chawla, lead business intelligence data warehouse engineer at Google, unpacks what this shift really means in practice. He draws a clear line between rule-based automation and process autonomy, illustrating how agentic AI can transform everything from financial reporting to business decision-making.
Through real-world examples from Google and his prior work on Amazon Alexa, Chawla explains how autonomous agents are redefining speed, accuracy, and strategic impact.
The discussion also goes deeper into the often-overlooked foundations of autonomy: data architecture, governance, and trust. Chawla argues that sustainable autonomy depends less on flashy models and more on robust data pipelines, and that this reality is forcing a fundamental evolution in the role of process excellence leaders. As organizations edge closer to the autonomous enterprise, this conversation offers a grounded, practical view of what it takes to get there.
PEX Network: How do you distinguish between traditional process automation (like RPA) and the new wave of process autonomy driven by agentic AI?
Rahul Chawla: The distinction lies in the shift from ‘execution’ to ‘reasoning.’ Traditional automation (RPA) mimics human hands. It is deterministic: ‘If X happens, do Y.’ It is incredibly efficient at repetitive tasks, but it is brittle – if the data schema changes or an edge case appears, the bot breaks.
Process autonomy, powered by agentic AI, mimics human judgment. Instead of following a linear script, we assign an agent a high-level goal (e.g. reconcile the North American product-level profit and loss). The agent uses reasoning capabilities to inspect the data, adapt to changes, and determine the optimal path to achieve that goal. At Google, we view this as the evolution from systems that simply execute instructions to systems that navigate ambiguity, enabling true ‘lights-out’ operations for complex workflows.
PEX Network: Can you share a real-world example of this shift at Google?
RC: A prime example is the transformation of our internal P&L reporting. Historically, this was a heavy, 24-hour cycle involving significant manual reconciliation by our finance teams to aggregate disparate global datasets.
To solve this, we didn’t just automate the spreadsheet, we architected a ‘financial agent’ on Google Cloud Platform. Unlike a standard script, this agent autonomously ingests transaction data, validates it against dynamic accounting rules, and executes complex logic to derive the final P&L.
The impact was immediate: we reduced the reporting latency from 24 hours to under 15 minutes. By shifting from human-led reconciliation to agentic execution, we effectively eliminated 80 percent of the manual effort, allowing our finance leadership to move from reactive reporting to proactive strategy.
Register for All Access: AI in PEX 2026!
PEX Network: Beyond just speed, how does process autonomy transform business intelligence into decision intelligence?
RC: We are exiting the era of the passive dashboard. For the last decade, business intelligence has been about showing humans what happened yesterday so they can make a decision today. Decision intelligence flips this model. When you have a trusted autonomous agent, the system doesn’t just display the data; it analyzes the ‘why’ and recommends the ‘what next.’
For example, in my previous work on Amazon Alexa’s data pipelines, we didn’t just track error rates on a dashboard. We built feedback loops that identified ‘why’ an interaction failed and autonomously fed that data back for model retraining. That is the difference: business intelligence is a rear-view mirror, decision intelligence is a self-driving navigation system. It turns data into an active asset that improves the business without constant human intervention.
PEX Network: You recently wrote about data foundations. Why is the underlying data architecture often the biggest hurdle to sustaining autonomous systems?
RC: Put simply: algorithms don’ t fail, pipelines do. Everyone wants to build the ‘brain’ (the AI model), but they forget about the ‘nervous system’ (the data architecture). If you feed an autonomous agent data that is stale, un-governed, or lacks lineage, it will confidently make the wrong decision – what we call a ‘hallucination’ in the enterprise.
The biggest hurdle I see is that companies try to layer generative AI on top of fragmented legacy data silos. For an agent to be autonomous, it needs a unified, governed view of the truth. Without strict data governance and lineage, you cannot audit the agent’s decisions. If you can’t audit it, you can’t trust it. Trust is the currency of autonomy.
PEX Network: As we move toward the ‘autonomous enterprise,’ how must the role of the process excellence leader evolve?
RC: The process excellence leader must evolve into the ‘architect of agency.’ In the past, our job was to map processes and remove waste (Lean/Six Sigma). In the future, our job is to define the ‘rules of engagement’ for digital workers. Leaders need to stop asking ‘how do I automate this task?’ and start asking ‘what data and guardrails does an agent need to own this outcome?’
This requires a new skill set that blends business logic with technical architecture. We are no longer just optimizing workflows; we are designing the governance structures that allow humans and AI agents to collaborate safely at scale. The leaders who master data strategy will be the ones who successfully make this transition.