How CTOs Use AI

Adithya Bellathur is the co-founder and CTO of Notey, a platform that turns music practice into a video game. He built the first version of the audio engine himself in college. Since then, Notey has grown to more than 100,000 users and 60 schools, and the team just closed its seed round.

In this episode, Adithya explains how Notey went from hand-uploading sheet music into config-driven mini-games to an agentic World Builder. The World Builder turns sheet music, YouTube videos, or even a hummed melody into personalized, playable lessons. He also covers:

• Why a three-person engineering team can now ship in a week what used to take Bloomberg six to twelve months
• Why AI-assisted engineers have to run two or three tasks in parallel, and why siloed engineers no longer work
• How non-technical teammates like artists and musicians ship real changes, with an engineer still making the final call
• Why human playtesting still matters most for quality in a game
• How Notey got to 46% day-360 retention for subscribers, in a field where most guitar learners quit within 30 days
• How LLMs changed ML from specialized models per problem to one general tool, from someone who did ML research before the shift
• The tools he uses, including Codex computer use, Warp, and AI audio splitters for music production

In this episode of How CTOs Use AI, Neel Jain talks with Joydeep Roy, Global CTO at Source Intelligence, about why his team skipped bolting AI onto existing processes and rebuilt the platform, and the SDLC itself, from the ground up. They did it without giving up security, architecture, or operational discipline.

In this episode:

• Why Source Intelligence unified five legacy companies into one AI-native compliance platform, and why MCP is becoming table stakes
• How hackathons and PRD-driven iteration won over the engineering team
• Why PRDs now start as markdown files an agent can consume, and how an intake package of requirements, guardrails, and golden rules drives the build
• The agent that asks what, why, and how, and writes no code until every question is answered
• Cognitive debt: why it’s the new technical debt, and how documented plans fight it
• Micropods: replacing two-pizza teams with one engineer and one product owner, each owning a PRD and its end-to-end flows
• Dropping two-week sprints for daily iterations and weekly demos
• Measuring ROI by how fast reliable features reach customers, not by lines of code or velocity
• Moving QA to risk-based testing and agentic coverage
• The unsolved PR review bottleneck, and the idea of routing PRs by blast radius to humans or agents
• Why open-source models and orchestration may be the future, plus Joydeep’s AI-powered wildlife photography habit (and its toll on his camera gear budget)

Guest: Joydeep Roy, Global CTO, Source Intelligence
Host: Neel Jain

Ruby Agarwal, VP of Engineering at Avaya, shares the framework she uses to decide when AI belongs in a workflow, when regular automation is enough, and how to keep costs, reliability, and governance under control.If you’ve ever been told to “just throw AI at it,” this conversation offers a practical reality check. Ruby explains why AI decisions should start with business outcomes, how much intelligence a task truly needs, and how much authority an AI system should have before it reaches production.You’ll discover:

• The three questions to ask before using AI
• Why AI is best suited to the intelligence layer—not always routine execution
• How caps, budgets, alerts, and model defaults help control AI spend
• Why context length, retry limits, and routing rules matter
• The production guardrails every team should consider, including observability, audit logging, access boundaries, and a kill switch
  

Ruby also shares practical examples from engineering workflows, including when AI makes sense for test generation and troubleshooting—and when it becomes an expensive overreach.This episode is especially useful for CTOs, engineering leaders, platform teams, and anyone responsible for moving AI from demo to dependable production.

AI is already reshaping how CTOs build software, but the biggest wins are not where most teams expect. Rajiv Jain and AJ reveal where productivity jumps by 30% to 150%, why new builds are getting faster while legacy code still creates friction, and how the real bottleneck is shifting from writing code to judging its quality.

Rajiv Jain, founder of Thinksys, has spent 14 years scaling a 500-person engineering company that supports software development, QA, DevOps, and security. AJ, founder of Pulso People and a software engineer with 20+ years of experience, brings the operator's view from the staffing and delivery side. Together they break down how AI is changing the software development lifecycle from the inside out.

You'll discover:

• Why AI is delivering the strongest gains on greenfield development, QA automation, and maintenance of test scripts
• How teams are using story points, Jira tickets, and token budgets to measure real ROI instead of AI hype
• Why guardrails matter more in 2026 as unrestricted AI spend gives way to disciplined usage and proven return
• How agentic AI is starting to handle manual QA by reading Jira tickets, running tests, and logging bugs automatically
• Why the hardest new challenge is testing non-deterministic, LLM-powered applications
• The everyday tools both leaders actually use, from Claude workflows and Chrome extensions to Granola for meeting notes

The episode also exposes a new management problem: when AI makes it easy to add code, bloated applications and unreviewed outputs become real risks. Rajiv and AJ explain why engineering teams now need to get better at saying no, enforcing quality gates, and training people to verify AI-generated work instead of blindly trusting it.

If you're leading engineers, managing AI budgets, or trying to figure out where AI creates measurable leverage versus hidden cost, this conversation gives you the practical lens you need. Essential listening for CTOs, founders, engineering leaders, and anyone trying to turn AI from a buzzword into a durable productivity advantage.