AI is fundamentally transforming organizations

The ongoing AI transformation is radically changing how enterprises run almost every aspect of their business.

But the transformation is just in its early stages, and enterprises must contend with new roles, processes and growing costs.

Peter Day, CTO at Checksum.ai and general partner at Super{set}, is a machine learning and AI veteran who has seen the technologies grow from his start in academia, through enterprise financials to the current AI startup world. He holds a Ph.D. in machine learning from the University of Liverpool, worked for UBS Investment Bank and is now general partner at Super{set}, a data and AI venture studio based in San Francisco.

Super{set} provides startup founders with resources to build their companies, including services such as bookkeeping and accounting, legal, recruiting and marketing that enable founders to focus on development.

Informa TechTarget spoke with Day at the AI Agent Conference 2026 held in New York City from May 4-5, where he discusses some of the thorny issues that CIOs need to consider as the AI transformation takes root in organizations.

Editor's note: The following transcript was edited for length and clarity.

AI is described as a transformative technology, but many companies are struggling with the transformation and finding it difficult to realize ROI. Why is this?

Peter Day: We see this across [early stage] companies all the time, which are relatively speaking, unencumbered. We don't have these large organizations to transform, and we're still struggling with that. Because we're having to throw away a lot of what we've learned over the last 20 years of building tech companies in Silicon Valley. When I speak to enterprise leaders, they're scared right now.

Why?

Day: They're being shouted at by their boards and their employees to adopt AI. So they do their due diligence to bring in a Gemini or Claude, but they're not seeing returns on that, because this is an empowering technology that changes who does what and how decisions are made. That's a much harder thing to change in a company than just bringing in a subscription to Gemini. The way you used to build tech companies is to have specialized roles around things like product manager, front-end engineer, and back-end engineer or database administrators. When we build tech companies now, we don't have those roles. We have builders, and we expect them to understand problems and build solutions, ship them and see if they work well. So, the level of personal leverage and autonomy has never been higher, but who gets to make the decisions is a bit chaotic.

Peter Day

Is the struggle that enterprises have with AI now primarily a tech problem or an organizational business problem?

Day: It's a mixed bag at the moment. My general advice to enterprise vendors is that this is not a technology problem. Too many enterprises early on thought that they had to think about AI and they tried to solve it by allocating capital to their IT teams to solve a problem that [they considered to be] a technology problem. It's similar to what happened to Blockbuster video when the internet revolution hit. They spent a fortune on technology, not realizing that the internet companies would come along and transform who makes decisions and how fast decisions could be made. So, they got massively outflanked by the Netflixes of the world, because organizational decision-making was different.

How does that compare to the current AI transformation?

Day: The same mistake is being made by some enterprises now. I spent time in a large consulting firm, and they spent hundreds of millions of dollars on infrastructure to enable AI, but it wasn't manifesting to the people who really needed it -- their frontline consultants. I talked to a Fortune 500 company recently that thought AI should make their finance processes go faster and be more predictable. My advice to them wasn't that they should use a startup or go to IT, they should find a smart kid who deeply understands their finance processes today and enable them to work out how to get the monthly close done in two days -- not two weeks. If you remove the roadblocks and incentivize them by saying, if you get this done in 6 months you'll get $100,000, that's going to give a much higher return on adopting AI technology.

Where is AI getting embedded in the enterprise, and what are the roles that it's taking?

Day: The flagship for this right now is in engineering, and the best agents now are the coding agents. Having these on engineering teams is a sign of things to come for every function. Companies used to have DBAs, front-end and back-end engineers, but they now will have one role, which is builder. They need to be able to code, but they also need to be able to direct coding agents. They need to be able to architect systems, but they also need to be able to build solutions that humans can use. The same is going to start happening in other functions. In sales, for example, organizations often have sales enablement, product marketing, SDRs and account executives, that probably becomes one role. That's one person who's surrounded by technology that understands what they're trying to do, everything from automating outbound follow-ups, to meeting recording, to the random email you have to send to the CFO each week. People who can do a bit of all these things are going to win.

Should AI be blamed for recent tech layoffs, and will it consume more jobs going forward?

Day: One of the challenges for any business is that cost is the only thing that they can really control. So if you want to prove that you've got an ROI on something, the easiest way to do that is to reduce cost. It's inevitable that some of this stuff will demonstrate ROI with cost reduction. For example, one Super{set} company does automated testing for software. One of their customers used to have 40 people doing that job, but they've now got zero because that role got fully automated. Not all of those 40 people got laid off -- many of them became engineers who use agents to build new solutions, so roles will change.

What are top issues CIOs should consider as they look at transforming climate?

Day: It's going to be really messy for them for a few years. We're already seeing a change from the way they used to buy software. [Before] you had to put everything through the CIO organization, it was procurement driven, and it might take six months to make a decision, another six to 12 months to do an installation and you had the software for another five years. In the SaaS era, it changed a little bit, but not that much. AI is starting to empower individual owners to adopt their own technology. That's the right call because it's all moving so fast, and trying to slow down decision-making is tough. Most things are moving to token-based pricing. It has to because that's where the cost structure is. It also reduces the risk for a business adopting this technology, but there's likely to be a few shocks when it comes to bills. The CIO and CFO are going to have to start keeping an eye on what technology's being used, how much they're using and how they're getting value from it. So, when they see the technology is adopted, they can turn that into something that is more integrated, controlled and predictable in terms of costs.

Do enterprises fully understand all the costs of the transformation?

Day: No, not really. These LLMs are expensive relative to the technology that we're used to, particularly in the cloud era with cheap databases. LLMs are expensive to run -- they're the single most expensive piece of delivering a solution for LLM-oriented software. That's going to be tough for enterprises, because whether they're adopting the LLMs directly via APIs and building their own solutions or having new software on top of them, it's all being metered. It's the right call right now because there's a lot of budget available for testing, but it's definitely something that's going to have to be answered over the next few years.

Much of this is still in the early innings, can we predict where the game is headed?

Day: It is early, we're still working out how to build on top of these LLMs and what the next generation of UI looks like. There are a lot of questions around exactly where the foundation model sits in terms of the value, what the pricing models are going forward, what margins we should expect and what our roles are. It's inevitable that software will become ubiquitous and that we'll have software that doesn't have a single pane of glass that we can speak to in very human way. It’s going to start absorbing tasks from us rather than giving us more tasks to do, all those things are inevitable, but I don’t know over what time frame and exactly what those things look like in the future.

Jim O'Donnell is a news director for TechTarget, where he covers IT strategy and enterprise ESG.