Risk Frameworks Nobody Updated Are Deciding Which Enterprises Survive The Next AI Cycle

Large companies used to decide whether to build software or buy it. Now some are building it with partners who already have the pieces they need. The tradeoff is access: the partner gets close to the data, workflows, and decisions the company used to keep inside. Plenty of companies can afford the engineering, and far fewer are willing to open the door.

Kunal Nagpal is Chief Business Officer at InMobi, where his remit spans InMobi Advertising and Glance. A former McKinsey consultant, he spends much of his week across the table from the company's advertisers and platform partners, and for two years those meetings have turned on AI. Early on, the executives were worried about their own jobs; lately, they want to know how much exposure they can live with.

"I'll obviously have my guardrails, but I'm going to bias towards the side of 'let's be okay with some risk' versus saying no risk," he says.

Ten arms and ten feet

Executives who take that position stop trying to build everything in-house. "It can't be winner takes all, but it definitely will be that the people who work with an ecosystem of partners are going to create a competitive advantage, because suddenly you have 10 arms and 10 feet working in conjunction with your own aspirations," he explains. Partnering at that depth means handing over more than a contract. His proposal is a long way from the clean rooms enterprises built over the past decade. "We all have data pools. Let's remove interfaces, let the data connect to each other through an LLM that we train together," Nagpal says. The first client he tried it on stopped him to ask who would control the resulting model. "You keep the governance. I have my data, you have your data, we'll keep it where it is, and then let the LLM travel."

One of those clients is a CPG company with billions to spend and decades of moving slowly. The company keeps that pace on the things it treats as core and abandons it everywhere else, rebuilding how it works with finance controllership, media partners, and its CRM stack as a system that keeps changing rather than a procurement cycle. "Even if I have oodles of money and I throw everything at it, it would be stupid of me to try and make this all on my own, because suddenly I'm not a tech company, I'm still a CPG company," he says.

What counts as proprietary now

Handing over that much forces a question most enterprises have never had to answer: once the tools are installed, what do you still own that nobody can copy? InMobi ran that cycle on itself first, throwing out the SaaS platforms it used for collateral sharing, training, and transcription, and putting agents on Salesforce. "That was a good shift. It needed to happen, because it opens people's minds on how you can think," he says.

"Is my IP the fact that I can write code that nobody else can write, which seems a bizarre statement to make in the world of Cursor and Codex and Claude Code? Or is my IP the fact that I'm going to create these interfaces that are uniquely adapted to me and to my clients, so that my competitive advantage becomes purely my ability to drive results for them?" When anyone can rent the engineering, what's left is context, customer relationships, and owning the workflows that create value. Those are the assets an enterprise guards most closely and the ones co-building requires it to open, which is a large part of why so many organizations are still short of enterprise-wide impact.

Risk appetite becomes the limiting factor

Opening that door is where most enterprises stop, and Nagpal has watched the argument play out from the other side of the table. In the deals that get done, a sponsor drags the conversation away from the lawyers and hands it to the product and technology people, and in the ones that stall, nobody does. "Most companies have not updated their risk appetite, and it's because of that they struggle to see beyond the slow evolutionary aspect of their growth," he says.

Nagpal doesn't claim to know the timing. "All else equal, sitting still is equal to dying," he says. A slow company can stay slow in most places, as long as it picks where a faster partner sets the pace. "If we don't move, there's somebody else moving. And that somebody else, as you can bet, is from my competitor base." Everyone can license the same models. The difference is how quickly they put them to work.

From the Journal test to the Reddit post

Risk appetite gets tested on a second front, and the ground under reputation has shifted just as fast. Nagpal learned the old standard early in his career. "When I was at McKinsey, we used to call it the Journal test. If your name showed up at the Wall Street Journal front page, how would you react?" He doubts many executives lose sleep over the front page now. "What they care about is some LLM picking up somebody's post on Reddit and you getting ruined on your brand reputation."

The old guardrail was media training for the ten people allowed to speak publicly, which Nagpal considers close to irrelevant. "My risk now lies with the people who just came out of undergrad who might tweet something about my strategy, my financial numbers, my relationship with a client," he says, or who might burn the house down with a tweet on the way out. Companies have been tightening how they handle reputational risk, and what a company publishes can shape what AI systems say about it later.

Co-building strains those frameworks from the other direction. A breach is one traumatic event on one day, while the questions partners now ask each other cover the whole training setup, the pools feeding it, and the controls around it. The change shows in who turns up. "Suddenly, the lawyer walked out of the room and the product manager and the techie came into the room."

Who moves first

The cost of launching a new brand, in his reading, has fallen far enough that his clients' competitor lists are growing so fast that it can be difficult to keep up. "This is not a point of 'let me improve 5 bips of EBITDA for Wall Street,'" he says. He came back from a recent trip to China with the same read. Chinese brands are moving upmarket at prices that would hurt a Fortune 500 competitor, and assembling the pieces faster than the incumbents they chase. "What are you going to compete on? You're going to compete on your brand recall, your emotional value," he says. Turning brand affection into revenue means first-party data and partners who can build against it.

The advantage goes to the companies that have already settled what exposure they can tolerate, what they're willing to share, and what they need partners to build alongside them, which is a different question from what they can afford. He expects companies out of Asia, Eastern Europe, and Africa to arrive in the next few years looking and operating like Western incumbents, with one difference. "They'll be more nimble, because they'll have done all the things faster or better."