Cisco president and chief product officer Jeetu Patel has staked a legacy code rewrite on a 27-year-old founder he found buried in the organization, with unlimited AI tokens to refactor 98% of the company’s old software within 24 months. Padmanabhan Krishnamurthy has given himself four weeks to prove the approach before that 24-month refactor. Separately, Patel has described a small team as a catalyst so core engineering owns the final product. For CxOs, it is a test of token governance and of who stays accountable when generated code reaches production.
Key Takeaways
- Patel gave Krishnamurthy unlimited AI tokens to refactor 98% of legacy code in 24 months.
- Krishnamurthy set four weeks to prove the approach before the 24-month refactor.
- A two- to three-person team is described as a catalyst for core engineering ownership.
- Patel said the approach could help more migrations to memory-safe languages.
A Legacy Code Rewrite With No Token Ceiling
Krishnamurthy cofounded the real-time translation startup EzDubs in 2023 and joined Cisco in 2025, when the $429 billion company acquired it, Fortune reported. Patel said that, despite founder and CEO experience, he was “buried away in the organization doing something unstrategic,” five levels down.
In Cisco’s San Francisco office, Krishnamurthy said a two-person team could rewrite the legacy estate. Patel told Fortune that even the most AI-forward people on his own team were not thinking that way.
“I’ve given him unlimited tokens because he’s going to refactor code,” Patel told Fortune at a small roundtable dinner. “So I’m like, you can go, you can go nuts.”
On this job, unlimited tokens are the constraint Patel removed. The 98% target and the 24-month clock remain. Read against Amazon’s tokenmaxxing memo, which puts outcomes over usage, the grant is a contrasting control, not a new standard.
Four Weeks, and a Separate Handoff
Krishnamurthy has given himself four weeks to prove the approach before the 24-month refactor, Fortune and India Today’s report said. He proposed two people. That clock is not described as the point when core engineering takes the result.
“The thing that we have done here is thought of this 2 to 3 person team as someone that catalyzes the movement and gets the core team to own the final product,” Patel wrote.
“These experiments are super important to sponsor,” he wrote, and added that “it’s definitely worth the pursuit,” even though not every experiment will work. The short proof and the catalyst handoff appear as separate descriptions, not one sequence ending at four weeks.
Hiring for Friction, Not Data Entry
Patel is also making a workforce bet. Fortune reported that some 40% of bosses say they plan to hire fewer graduates because AI can do the job cheaper. His question ran the other way: “If you don’t hire the early-in-career [workers], and you just let your workforce continue the way that they were, where do fresh ideas get injected? How do you challenge the status quo?”
He wants people who will “challenge the system and the status quo” and act as “an energizer bunny that just goes out and kind of revs everyone up.”
“Tension is good,” he said. “I actually try to create tension in the teams because you want to have enough friction on the ideas. You don’t want to have friction with people and personalities.”
| Cohort | Share | Role Patel described |
|---|---|---|
| Scaled operators | Roughly one-third | Understand how the Cisco engine works |
| Systems thinkers from the market | Second third | Outside-in perspective |
| Founders, CEOs, and AI-native engineers | Final third | Operate in founder mode |
He told Fortune that either group alone is not “nearly as potent” as when all three work together. “It’s what creates magic.” The useful reading for other companies is narrower: an AI-native third still has to satisfy the operators who understand the Cisco engine.
Security Upside and the Verification Gap
Patel has also framed a security benefit. As India Today reported, he said the approach could help Cisco carry out more migrations to memory-safe programming languages, which can reduce vulnerabilities linked to memory-related errors. “Every engineer at Cisco will benefit from this experiment,” he said. He believes this kind of experimentation could eventually lead to software that is “infinitely more secure, more innovative and filled with a lot more possibilities.”
That is intent, not a measured result. The accounts do not say how much code has moved, which models are in use, or how refactored software will be tested.
As analysis, not as a Cisco forecast, large organizations can pressure-test four points:
- Who may spend without a token cap, and against which outcome.
- Whether core engineering accepts the handoff Patel described.
- Whether memory-safe migration changes vulnerability counts.
- Whether a four-week proof can stop a thin result.
Unlimited tokens authorize generation. They do not price verification, the planning issue in Forrester’s AI cost squeeze. The accounts include no token budget, defect target, or rollback rule. Until those appear, the question is whether operators who know the engine will still recognize the software after the handoff.
