The following has been condensed and lightly edited for clarity.

Fortune: There’s a lot of noise around AI. How do you decide what to pay attention to? 

Kumar: Every time I find a thesis, I experiment inside the company. It gives me this unique opportunity to test it and then reinforce my thinking or tweak my thesis. You build your thesis or your hypothesis on gut, which is a combination of experience, intuition and connecting the dots, and then you layer it with data, and you get close to 60, 70% and then you’d stop layering it with any more data, because you’d be late. And then you go back to your gut and push it through. Rethinking your assumptions in a world which is moving at a high pace is important. I think we’ll need more school graduates in the AI era.

Why more?

So many companies have a pyramid with the bottom where school graduates are. That pyramid is going to be broader and shorter, and the path to expertise is going to be faster. It’s going to be faster if you can rewire your K-12 schooling system with lifelong learners and make undergrad education much more about using AI tools and creating interdisciplinary skills. This year, we are hiring more school graduates than ever before. I can take a school graduate and give them the tooling so they can actually punch above their weight. AI is an amplifier of human potential. It’s not a displacement strategy.

What type of students are you hiring? 

I grew up thinking, the more you specialize, the more premium you get. Now that this expertise is at your fingertips, how can more specialization and more expertise be a premium anymore? If it’s faster to expertise, then expertise is not the asymmetry. Intelligence is not the asymmetry. Applying intelligence is the asymmetry.

Start to focus on interdisciplinary skills. If I’m a historian, I could blend it with computational skills and become a futurist. If I am a biology major, I could crack drug development cycles and disease using computational skills. 

A large chunk of work is problem solving, so we created these departments around problem solvers. Those departments were mostly non-STEM disciplines, and the core of the enterprise was STEM disciplines. Now, if problem solving is assisted with machines, you would find an equitable distribution of problem solvers and problem finders in an enterprise, which then means the mix of people on the core is going non-STEM disciplines like anthropologists, sociologists, psychologists, journalists; people who can be more purposeful problem finders.

You need human skills plenty at the end. The start is all about prompting, conceptualizing, finding the purposeful problem, and everything else. The middle is all there with AI, and the end of it is validation and verification by humans.

It sounds like you are capitalizing on your gut here.

It’s very similar. Everybody thinks AI is going to relegate creative skills to humans, while machines do the validation and verification. I think AI would actually be the creative thing, and humans are wired for validation and verification.

If the digital revolution was information at your fingertips, this is expertise at your fingertips. If we can wire our education system to use the tooling to increase the throughput, we will have a productivity bump. The last time the productivity bump happened was in the Internet revolution. After that, productivity has been flat, in spite of the billions of dollars we spend on technology.

Why has it been flat? 

It’s been flat because we’ve used technology to replace human work. We’ve not used technology to amplify human work. If it’s a productivity bump, it will create more distribution of wages, provided it’s not in the hands of few people and you distribute it equitably. 

We’re seeing student test scores go down in key areas; fewer people going to college. How do you imagine this next wave rolling out equitably? 

I did put a caveat on it, saying it needs to be in the hands of people and distributed. Digital skills created a divide. It really didn’t create a bridge, because the ones who had those skills were further away from the ones who did not. It covered people who produced the tool. It did not create prosperity for the people who used it. The producers made a ton of money, and the users had convenience and information at their fingertips. 

So how are you deploying this internally at Cognizant?

One of the experiments we have just kicked off with a company where I worked before is to look at mid-career shifts. We have multiple swim lanes in our company. There is a deep technology swim lane and a second swim lane of applying technology to businesses, which is not deep into technology. It’s a combo of knowing operations and knowing technology. You can land some of the mid-career people into those jobs. 

You can create upward social mobility using this tool. We are going to do an apprenticeship program where the template is work, earn and learn, and I’m starting to look to universities to partner with me to credential this work. Every technology revolution offers pathways. Here, you don’t need the skills to access the machine. We think it’s a leveler. It’s an equalizer, because the entry barriers for these jobs are much lower. The race to the top is quicker.

Are you hiring differently?

We are now going to hire non-STEM graduates. I’m going to liberal arts schools and community colleges. We have apprenticeship programs in 30 states approved and I’ve just kicked off a program with a company called Merit America, which focuses on career shifts, so people don’t leave their jobs. So we’re trying all this. The question is: Can I do this at scale? 

Can you?

We are all about rinse and repeat. 

How do you reconfigure the company itself?

The Industrial Revolution tied work, workplaces and the workforce all together. It was an integrated and hierarchical model: You go to a factory, deliver things in a time period, and then you get out. Now, the Hollywood blueprint is more viable because of AI. 

What is the Hollywood blueprint?

It used to be vertically integrated. You got a studio and six movies a year, and the movies were all similar. It had a set of directors and a set of actors locked for all the movies, and some of them also owned the movie halls and the cinema theaters. And it worked because people wanted to unleash themselves, and there was a programmed set of themes, which worked. And in the 50s, television came into picture and people wanted a variety of things. There was an unlock that led to where you have actors, directors, and technicians no longer having long-term exclusive contracts. 

It created an agile system where you assemble teams for a project for a broader purpose, and you dismantle it after it is done. The studio was a physical entity and everything else changed. The production houses were the capital structures and everything else was fluid. For the logistics of sourcing, onboarding, managing this high caliber, specialized talent on demand at scale, Hollywood could get it done with agencies, unions and service firms, which created a well-coordinated ecosystem. 

And that’s what the corporation needs to be because of AI? 

Corporations have evolved from the Industrial Revolution, but they didn’t go all the way to the Hollywood model. The constraint was institutional knowledge, tribal knowledge, the heritage of the company, the enabling layers of finance, HR, all of it. Also, it was hard to assemble and dismantle teams. 

What you got was this gig worker economy, which was about variable capacity, but the core piece was still very thick. Brewing coffee in Starbucks is very different to brewing coffee elsewhere. There is a hustle of a company, the tribal knowledge, the culture. We can feed that tribal knowledge in whatever form we get into the LLM to build an agent on the other side, which is very contextual.

So the Starbucks agent would act differently than one created for a competitor?

You could build context engineering in a variety of ways: feed the tribal knowledge, feed the workflows and the data flows. You could do it in pre-training or inference, where it will learn over a period of time, and then become ready. When you do that, you make the AI capital permanent, the agentic capital permanent, and you unleash people to be the variable component, which means people can go in and come out, depending on the broader purpose of the project. The fixed capacity is predominantly agentic capital that holds the heritage and tribal knowledge of a company, the culture. 

Sometimes, the culture of a company can be an impediment to making change. Take the Starbucks controversy around Charlie Kirk’s death. It made workers create a connection with customers by writing on a paper cup. Turns out, that policy had problems when people use that to make a statement. So how do you challenge the tribal knowledge that may not take you from here to there?

Great question. You want to have an organism that pivots to the future and sometimes the past is an impediment. The beauty of AI systems, unlike humans, is that you can configure it to your needs. It can dispassionately assess what needs to change.

Look, Cognizant has a rich, winning heritage. I draw from it, but I equally will change to stay relevant in the future. Now for humans, it’s harder to make that change. That’s why changing large enterprises may take more time, while the nimble companies are the new ones, which have no legacy and no heritage. The beauty of AI systems is they’re not self-aware that they’re making a mistake but they have situational and system awareness, which is much higher than humans. 

I therefore believe you’re going to see this fluid structure with agentic capital, some human capital to supervise it, and then everything else is variable. You could define an objective outcome and assemble a team for an outcome and dismantle the team for an outcome. 

So much of that Hollywood studio model relies on a certain mindset of the individual and a certain layer of security that allows them to be flexible.

When we unlocked television, we didn’t get better television. We got TikTok, YouTube and other different things. So this unlock is similar. The decoupling of work has happened with gig workers. Decoupling of the workplace has happened with pandemic. Decoupling of work will also happen. And the trigger for this is the mindset. It’s not just whether these platforms, whether these ecosystems, are available for you to express yourself. There has to be enough demand for it and that will come if an individual starts to look at the things like the 401K, health plan, and skills as things they manage. Are we ready for the Hollywood model for our professional jobs? 

There are a lot of comparisons between Gen Z and those who came of age during the Depression, a craving for stability. This generation believes that they don’t have an on-ramp to careers, and trust in institutions is going down. And so how do you then engage them with a different model? 

I think we have four generations of workers in our workforce now. Some don’t want to go with this high clock speed where you’re on your own and the economic outcome is also based on outcomes instead of the number of hours. I just believe there is an unlock to create more distribution of good work. I have other people here with two to three years’ experience, who come and tell me, ‘why are you forcing us to take the benefits and health care? Just give us the money, we will figure out.

Until they break their leg.

I think the Hollywood model is definitely applicable to project-based organizations that can operate with high clock speed, high agility, more creativity. 

We are also getting to an era where people are living longer.

Yes, they will have multiple careers in one lifetime. As people are living longer, the life of their skills is getting shorter. We have to wire the world for that future. This is a model to unlock work into modular packets so that you can access more capital, more human capital.

Where do you get the most inspiration out of your job right now?

I’m a big fan of applying the information I have to a much broader spectrum of things and generating cross functional insights, which was very difficult before. So I continue to connect the dots much better, just because there’s so much instrumentation around me to support it. I can ask an AI model something provocative, and build a hypothesis around it, and that could be interconnected between disciplines and interconnected between things which are outside my company and things inside the company.

Do you lead differently? 

I have started to believe now that you cannot, as a corporation, work in isolation to the broader environment. It’s more integrated now than before. The clock speed is much, much higher, and we should be able to recalibrate at a much quicker pace and revalidate our assumptions. I never thought that was such a big deal. I thought once you lay the foundation, set your assumptions, you kind of are on a good runway. You have to keep changing paths much, much quicker and much faster. You’re leading with four generations of people who all have their unique needs and their unique imperatives. You need that fine balance to keep an eye on the future and make changes for the future while keeping an eye on what’s current. Enterprises are the biggest platforms for societal change.