Cisco Bet One Engineer and Unlimited AI Tokens on a 98% Legacy Code Rewrite — That Bet Needs an Audit Business Nobody Is Selling Yet

Cisco just made a bet most boards would call reckless: one engineer, unlimited AI tokens, and a mandate to rewrite 98% of the company's legacy code in 24 months.

That bet is a gift to anyone who can sell the risk controls nobody at Cisco has mentioned publicly yet.

The insight

Fortune reports that Cisco president and chief product officer Jeetu Patel gave Padmanabhan "Paddy" Krishnamurthy — a 27-year-old former startup founder whose translation company EzDubs Cisco acquired — unlimited AI tokens to refactor the company's legacy codebase. Patel's quote is the whole story in miniature: "I've given him unlimited tokens because he's going to refactor code. So I'm like, you can go, you can go nuts." The target: 98% of Cisco's legacy code, rewritten within 24 months, starting from a 4-week proof of concept with what Krishnamurthy pitched as a two-person team.

What's notably absent from the public framing is any mention of how Cisco is bounding the risk: no stated test-coverage threshold before a module gets rewritten, no published rollback plan, no visible ceiling on token spend despite the word "unlimited." That's not a criticism of Cisco — it's a gap. And every company that reads this story and wants to try the same thing is going to need exactly that governance layer built for them.

The money play

1. Sell an AI-rewrite readiness audit. Before a client lets a small team loose on a legacy codebase with an AI coding agent, assess what's actually being risked: test coverage, deployment blast radius, and which modules are too load-bearing to refactor first.

2. Package a token-governance retainer. "Unlimited tokens" sounds generous until finance asks what it cost per delivered feature. Sell ongoing monitoring that ties AI-assisted refactor spend to actual merged, tested output — not just tokens burned.

3. Build an AI-rewrite regression-detection service. Diff the behavior of the old module against the AI-rewritten one in a staging environment before every merge, specifically hunting for silent behavior drift that unit tests written by the same AI agent might not catch.

4. Productize the bounded pilot Cisco actually ran. Krishnamurthy's pitch wasn't "give me a huge team" — it was a two-person team and a 4-week proof of concept before the full mandate. Sell that exact structure — small team, hard deadline, checkable output — as a packaged engagement to other companies instead of a sprawling transformation program.

5. Sell the executive briefing on falsifiable AI bets. Patel didn't announce "AI-powered modernization" — he set a number (98%) and a clock (24 months). Most leadership teams don't know how to frame a bet that specific; package the workshop that teaches them to.

Who buys this

• VPs of Engineering at companies with large, aging codebases eyeing an AI-assisted rewrite after watching Cisco's bet make headlines
• CFOs and finance leads nervous about what "unlimited AI tokens" actually costs once a team is mid-rewrite
• Platform and SRE teams who will inherit whatever an AI coding agent produces, and need a way to verify it before it ships
• Boards and executive teams that want to announce an AI initiative but need help setting a number that can actually be graded in two years

Bottom line

Cisco just ran the experiment in public: one engineer, no spending limit, a hard number, a hard deadline. Every company that wants to copy it will need the governance layer Cisco hasn't described — and that's a service business sitting in plain sight.

Source: https://fortune.com/2026/10/30/cisco-president-jeetu-patel-gen-z-27-year-old-unlimited-ai-tokens-rewrite-legacy-code-entry-level-jobs/