The AI Messenger
The CEO of a new unit of a Japanese data giant lands in Silicon Valley in hopes of showing firms that AI “playtime” is over—now. Meet Bratin Saha.
July 22, 2026
Bratin Saha has a blunt message for companies still treating artificial intelligence as an experiment: Playtime is over. For Saha, who was appointed CEO of NTT DATA AIVista about a year ago, it’s game on.
NTT DATA, the Tokyo-based IT-services provider with about $30 billion in annual revenue, launched AIVista in December 2025 and installed Saha as its leader with a mandate to help firms operationalize AI at scale. And it’s not as if it placed him far from the action—the company is smack dab in the middle of Silicon Valley, the heart of artificial intelligence.
As Saha sees it, his mission is to help move firms “from the pre-AI era into the post-AI era.” It’s a tall order for the most seasoned leaders, but for a first-time CEO like Saha, it’s a career-defining challenge. But he arrives with some critical experience, including helping to build Amazon’s AI services at AWS and the software infrastructure at Nvidia, and creating the AI business at DigitalOcean, a cloud firm in Colorado.
“AI is ultimately a people thing. It changes how companies work, how employees work, how the workforce looks, and how humans interact with technology.”
To be sure, Saha’s background as a product leader and operator is what differentiates him from the classic services-company CEO. Put another way, he has serious credibility with the engineers, coders, and builders in Silicon Valley. The goal is to commercialize AI at scale for firms, as he has done with other new technologies for the 75 percent of Fortune 100 companies that NTT DATA counts as clients.
In one of his first interviews as CEO, Saha spoke to Korn Ferry about the challenges firms have with scaling AI, his experience leading a company for the first time, and his goal of democratizing the tech. Below is an edited version of the conversation.
A new global report from NTT DATA proclaims that “playtime for AI is over.” What do you mean by that?
For the past few years, enterprises have been experimenting with AI and running pilots. The problem is that many of those pilots do not deliver business value. But the ones that do create value can be transformational. Companies now have to get serious about deploying AI at scale. The bottleneck is not the technology itself. It is operationalizing the technology: moving from prototype to production value. Companies that do not get serious over the next 18 to 24 months risk being left behind.
What is it about the next 18 to 24 months that makes them so critical?
In the technology world, we are used to Moore’s Law, where computing power doubles every two years. AI’s progress is an order of magnitude higher. Over the last two or three years, it’s been doubling roughly every six months. A 24-month window could mean the technology gets about 10 times better. Leaders cannot think in terms of five-year planning cycles anymore.
The NTT DATA report also claims that only 15 percent of firms are positioned to be AI-ready leaders. Can you explain what that means and why so few firms are at that level?
Enterprise AI is hard; I’m not going to sugarcoat that. But it can be made easier. AI is trying to automate human work, and humans operate with company-specific knowledge. A graduate who joins JPMorgan and one who joins Goldman Sachs may have similar educations, but after a year they operate differently because they have learned the company’s proprietary processes, controls, customers, and priorities. AI models have learned what is public. What they do not have is that “last mile” knowledge. Bridging that gap is critical.
What do you mean by “last mile” knowledge?
The last mile is the difference between what the model knows and what it needs to know to be effective inside a company. To make AI work in production, you have to capture that knowledge in a form agents and models can use.
Get to Know Bratin Saha
- Alumnus of Harvard Business School, PhD in computer science from Yale University, B.Tech in Computer Science from IIT Kharagpur.
- Holds more than 70 patents and has published more than 30 papers.
- As VP of Software Infrastructure at Nvidia, led global software infrastructure for product lines.
- Helped scale Al services at Amazon Web Services from zero to a multibillion-dollar business as well as one of the fastest growing businesses in AWS history; work was the subject of Harvard Business School case studies.
- As chief product and technology officer and Section 16 Officer at DigitalOcean, launched DigitalOcean's Al products and other innovations, contributing to a 30 percent increase in market capitalization.
You’ve said that the main bottleneck for firms with AI is moving from proof of concept to production at scale. How do you break that bottleneck?
To me, there are a few ways. The first is last-mile specialization. You cannot take AI out of the box and expect it to work inside an enterprise. You have to train it with the same proprietary knowledge that you train a human with. Second is change management. Third is business-domain knowledge. You also need good data rigor and a clear use case. You cannot boil the ocean with your first deployment, but you also cannot pick something so trivial that you fail to learn how to deploy AI at scale.
What did you want to accomplish in your first year?
The first priority was to build the organization. The second was to set the strategy by understanding key customer pain points. The third was to deliver production outcomes—not pilots or proof points, but something that works in production. We have hired the leadership team, developed a differentiated strategy, and started building products with milestones toward production with multiple clients.
What motivated you to move from the product/operator side into the CEO role?
I helped build models and the plumbing around models. But if 80 percent or 95 percent of pilots are not getting to production value, you have to ask what is going on. I realized that to democratize this technology, I needed to be on the side that does the last-mile deployment. AI is ultimately a people thing. It changes how companies work, how employees work, how the workforce looks, and how humans interact with technology.
Your public comments suggest that retrofitting AI into legacy infrastructure at big firms is inherently challenging. Does that mean legacy firms are at a disadvantage versus AI-native firms? Or can they be restructured?
An AI-native way of operating is fundamentally different. Take a word-processing application. Today, you create a document, edit it, and finish it. In an AI-native world, that document becomes dynamic. You could tell agents to update it with new information as it becomes relevant. What used to be a static process becomes an always-changing process.
Products become living things that are always connected and constantly refreshed. So you have to approach and think about AI native-products differently.
Both sides have advantages and disadvantages. Existing businesses have domain knowledge, business knowledge, customer data, and an understanding of how clients use their products. AI-native companies do not carry the burden of legacy systems, but they may lack the domain expertise and customer data. It is a race. The winners will be the ones that move fast enough, with urgency and flexibility.
You’re far from new to the challenge of building a new enterprise on an enormous scale, having led Amazon Web Services’s AI products into a multibillion-dollar business. What lessons did you learn there that will help in this new venture?
The work at AWS is the subject of three Harvard Business School case studies. Looking back, there were some key lessons that mattered. The first is customer obsession. Stay relentlessly focused on the pain point and how to solve it. The second is “nail before you scale.” In a fast-moving market, there is a temptation to move fast, win a few customers, and pile up pilots. But you have to get it right before you scale it. The third is knowing your durable advantages and building on them. At NTT DATA, those advantages include global reach, client relationships, domain knowledge, and change-management expertise.
Are there any parallels from the transition to cloud computing that can be applied to AI transformation?
Cloud made IT faster and more efficient. It transformed business models, but it was still taking an existing thing and improving it. AI is different. It is the first technology where we are saying it can analyze, think, and act like humans do, which is not an improvement of an existing technology but a completely different ball game. The opportunity is 100 times greater.
There are a record number of first-time CEOs in the US. As a first-time CEO at a new NTT DATA company, what’s been your most unexpected challenge?
The increase in demands on your time. I have grown from engineer to manager to director to VP to CXO, and the demands on time always go up. But the CEO transition drove a step-function change in the demands on time. It requires ruthless prioritization, cutting out noise, and exercising judgment about what to do and what not to do—especially in an environment with so much AI hype.
How has your view of leadership changed now that you are a CEO?
You become aware that you represent the company in every room, every conversation, and every statement. A misplaced word or false step reflects on the organization and everyone in it. That weight is real, and you have to stay conscious of it all the time.
With all the challenges CEOs face, what makes stepping into the role attractive for you right now?
We are at a truly transformational point. There is an opportunity to do work you can be proud of and leave a legacy. Helping NTT DATA move from being a leader in the pre-AI era to a leader in the post-AI era is hugely motivating. I want to help democratize enterprise AI for organizations that have the ambition but have not yet gotten there. So, I think the opportunity to leave a mark, do real transformation in the industry, and democratize one of the most important technologies humankind has invented makes the hard parts worth it.
Image credits: Ookawa/iStock/Getty Images; Yuichiro Chino/Getty Images; Robson 90/Alamy; Andriy Onufriyenko/Getty Images
