By: Kandice Vincent
Rajeev Jaswal once signed a renewal he did not want, at a price he did not like, with a vendor who knew exactly why he was signing it.
He was a CIO at the time, and the system in question was one everybody in the organization complained about. So he ran the numbers on replacing it. The license turned out to be the small part. The real cost was retraining several thousand people, rewriting integrations nobody had documented, and a long stretch of the entire company working slower while everyone relearned their jobs.
They renewed. At a higher price. “That is not a negotiation,” he says.
That renewal is the closest thing Argo has to an origin story. Jaswal is now the CEO and founder of Graytitude, a mainframe modernization firm that serves large enterprises as a channel partner to IBM, Rocket Software, and Broadcom, and Argo is the enterprise AI platform his team built out of it. The design principle he keeps returning to comes straight from that conversation across the table: “The invoice is the smallest number in the conversation. The number that matters is the exit cost.”
Twenty-Five Years of Saying No
Before Graytitude, Jaswal spent roughly twenty-five years inside enterprise technology, including CIO roles at Rapid7 and Red Hat and technology leadership at GlaxoSmithKline. Ask him what pushed him out of that world and the answer has an edge to it.
“Twenty-five years of watching good ideas die somewhere between approval and production,” he says. “I would sign off on something in March and see a partial version of it in November, by which point the problem it was meant to solve had moved.”
He is blunter still about the job itself. A CIO spends most of the day saying no. No, that does not meet our security requirements. No, we cannot integrate that this quarter. Most of those calls were correct, and he got tired of being the person who made them. “At some point I wanted to be on the other side of the table, building the thing that was almost right into something that worked.”
Graytitude was the version he started with. Argo, he says, is the thing he actually wanted to build.
Where Enterprise AI Projects Die
Jaswal’s view on why so much enterprise AI never reaches production is specific, and it has very little to do with the models.
“They stall at the point where the system has to write something rather than read something,” he says. “Reading is a demo. Writing is a risk decision, and the person carrying that risk is a CIO who will be asked, ” What happened, and who approved it. Most AI tools have no answer to that question.”
So his team built the answer first. Every action Argo takes outside itself runs the same path: who is permitted, does a person approve it, did the system actually do what it reports it did, and is there a record that cannot be edited afterward. He describes that work as unglamorous, and then explains why it was the priority anyway. In his view, that groundwork is what decides whether a product reaches production or stalls in a pilot.
The problem it solves for the people using it is more ordinary. A finance leader wants last week’s booked revenue, and that question crosses an ERP, a data warehouse, somebody’s spreadsheet, and two people’s calendars. None of it is hard. It is scattered, and the scattering is what costs the week.
“More often than I would like to admit, I decided before the answer arrived, because the decision could not wait,” he says. “That is the thing I think about most.”
What Years Inside the Mainframe Taught Them
Argo’s connection to Graytitude shows up as an assumption the team refuses to make: that anything can simply be rewritten.
Graytitude has spent years doing mainframe DevOps for large enterprises, moving legacy source control onto Git and building delivery pipelines around systems that process critical transactions overnight and have change windows measured in minutes. That work happens in production, with the consequences attached. So Argo was built to work across a mainframe without asking anyone to modernize it first.
“That is a constraint we have lived inside on every project for years, turned into a product decision,” Jaswal says. He is direct about what that responsibility means when the systems in question move money and pay claims. “When one fails, somebody does not get paid, and it is rarely the person who can absorb it.” In practice, the team defaults to reversible, and nothing touches a system of record without a person approving it and a durable record afterward.
The Decision That Cost Them Months
Early on, the team could have shipped Argo tied to a single model provider. It would have put them in market months sooner, and the commercial terms were good.
They said no, and built the model layer so a customer can use what ships with the product, bring their own model, or point Argo at their own compute.
“It cost us time and resources,” Jaswal says. “But the entire premise of the product is that you should not be dependent on one vendor, and I was not going to spend twenty-five years complaining about lock-in and then build it into our own product on day one. If the principle only holds when it is convenient, it is not a principle.”
The same test gets applied inward. What would it cost a customer to walk away from Argo? If the answer is significant, they built it wrong.
The Only Adoption Signal He Trusts
The moment that convinced Jaswal the product worked was small and internal. Somebody on his own team stopped opening a system they had opened every day for years. No announcement, no write-up. They just stopped, because asking Argo was quicker than logging in.
“That is the only signal I trust,” he says. “People do not adopt a tool because it impresses them in a demo. They adopt it because at some point the old way became annoying by comparison, and they never went back. Everything else is enthusiasm, and enthusiasm has a short shelf life inside a large company.”
What he wants a CIO to feel after a month with it is less about savings and more about position. “I want their next renewal conversation to feel different. Not because they have saved money yet, but because for the first time they have a credible alternative to renewing, and the person across the table knows it.”
Share the Fame, Take the Blame
Jaswal credits a handful of mentors with how he thinks about building teams, and one piece of their advice has stayed with him the longest. A spider dies when you remove the head. A starfish regenerates from almost any part of it. Most technology organizations are held together by three or four heroes who know how everything really works, and everyone admires that arrangement until one of them leaves.
“I have inherited those teams,” he says. “The heroism is genuine, and the design is negligent.”
His own leadership rules are short. Give people the outcome and the context, then stay out of how. Say what you think in the room rather than in the hallway afterward, which he admits is harder than it sounds. And share the fame, take the blame. “If it went well, it was the team. If it went badly, I approved it.”
He also holds a position on high standards that he thinks gets misread as coldness. “The kindest thing you can do for someone is tell them clearly and early that the work is not where it needs to be, while there is still time to fix it. The unkind version is being pleasant about it for six months and then having a very different conversation.”
What He Expects Next
Jaswal will make two predictions about enterprise AI.
The first is that the AI project stops being a category. Right now most large companies have an AI budget, an AI steering committee, and someone whose title is head of AI. He lived through the era when serious companies had an e-business division, and within a few years that looked absurd, because the thing had stopped being a department and become a property of everything.
The second is that sovereignty moves from a preference to a procurement requirement. Most enterprises today cannot answer three questions with confidence: where does our data actually go, which models have seen it, and what would it cost us to move. Today a few careful CIOs ask them. Within five years, he expects regulators, auditors, and boards to ask them, and being unsure will not be an acceptable answer.
Away from all of it, he takes long walks at very high altitude. He has hiked to Everest Base Camp and completed the Mt. Kailash trek, and he likes that neither one has a shortcut. No amount of wanting to be further along gets you there faster, which he calls a useful thing to be reminded of when you run a company. There is also golf, less often than he would like, and two dogs, Skip and Yogi, who have no interest in how the week went and want to play the moment he walks in.
Argo is currently running early pilots with enterprise teams, two weeks and one process at a time, read-only to start. Companies interested in a walkthrough can learn more at argointelligence.ai.
