About Ram Ramanathan
I lead the operations strategy function at Kapitus, a New York based SMB lender. Previously, I was part of the corporate COO office at Cognizant and the Chief Delivery Officer’s office at TCS — across a career spanning 20 years.
My work sits at the intersection of process intelligence and AI - helping lending operations teams understand how work actually gets done before deciding where AI fits. That foundation, built across Six Sigma, Data Science, and Process Mining, is what separates AI implementations that hold up in production from those that look good in demos.
I’m a Six Sigma Master Black Belt who also codes, models, and builds the systems myself. That combination - operational depth and technical implementation - is rare in one person and is what makes the methodology in this blog practitioner-grade with sound theoretical underpinnings
What I work on
At Kapitus, I’ve built the process mining capability from scratch using Celonis, taken multiple AI systems from concept to production across lending workflows, and deployed an underwriting classification system processing ~45K contracts monthly at 90%+ accuracy.
The problems I typically get called in for:
🗹 An AI pilot in operations that doesn’t live up to expectations
🗹 Fragmented operations with no clear picture of how work actually gets done
🗹 No systematic approach to transformation — no clarity on what to fix, in what order, or how to sustain it
🗹 Operating models and workflows that need a ground-up rethink
🗹 AI investments that aren’t moving the needle on ROI
Why this blog exists
Most writing about AI in operations is either too generic to be useful or is too much in the weeds that it misses the forest for the tress. This blog exists to fill that gap. This series exists because the hardest problems in deploying AI for operational workflows — the evaluation framework, the sequencing logic, the human-in-the-loop design — are not being written about honestly by people who have actually done it.
Every post here is written from inside a real lending operation. Not from a consulting deck. Not from a conference panel. From the messy, political, data-rich reality of making AI actually work in production.
If it occasionally reads like someone thinking out loud while building something difficult - that's because they are.
You can follow my latest thinking in the articles above.
Get in touch
If something you’ve read here maps to a problem you’re dealing with - whether you’re an operations leader, a builder, or someone working on something adjacent - I’d like to hear about it.
📩 ramsthere@gmail.com 🔗 linkedin.com/in/rramramanathan
Disclaimer: Views represent my own, and none of this represents the views of any of my current or previous employers.
