Where the Sea Meets the Shore

How AWS guides private equity-backed enterprises into the AI era and keeps them from running aground

When it comes to navigating the capricious seas of technological change, Justin J. Leto, helps global
private equity firms sail to their ultimate destination – higher multiples.

As a Global Principal Solutions Architect for Private Equity at AWS, he’s part of a dedicated team that advises top-tier private equity firms on how to turn investment theses and operating plans into reality through cloud and AI transformation. He’s supported by Transformation and Innovation specialists and an M&A practice with tailored programs that include buy-side and sell-side due diligence and first 100-
day plans for new acquisitions.

On April 9, he’ll be honored at Penn State’s Outstanding Engineering Alumni Awards, one of 13 honorees this year from among the university’s more than 100,000 living engineering graduates. With over 20 years of experience in data engineering and AI, Leto helps them chart their cloud and AI journeys.

Reboot’s recent conversation with Leto revealed fresh insights on what PE firms should consider when
preparing for their expeditions.

There’s a longer road to full automation for most enterprises beyond deploying autonomous agents at scale. “AI agents need access to data and tools to create durable enterprise value. So you could say that the road to AI riches is paved with investments in data governance and modern data platforms,” he told us. “AI is compressing the time it takes to replicate entire software products from months to hours. But data is different,” he added. “Data is an asset. It has gravity. It incurs liability if exposed, and the
business value and differentiation AI delivers will derive from the data and what it can do with it.”

AI’s capabilities need to be perfected, and that will take time. However, recent declines in
market valuations for digital native SaaS portfolios signal that AI disruption is no longer an abstract
threat. To make the most of AI’s disruption, the time has come for companies to rethink their tech investment strategies.

“JP Morgan estimates roughly $2 trillion has been wiped off software market caps in just a few weeks, and the initial instinct is to pivot from growth to profitability. But a selloff this broad is driven by
fear: The market is overreacting, but it’s signaling that not everyone will be a winner in the AI era,
and it doesn’t know yet who the winners are. But that’s just too narrow a view: The bigger strategic
risks are in doing too little.”

He pointed to past technological shifts, and how big tech bets by early movers carried asymmetric downside risk. Cloud flips that equation by derisking AI experimentation and innovation, giving companies pay-as-you go access to highly specialized AI accelerated hardware and services.

“The risk now skews sharply to the upside. For most companies, the opportunity is in making thoughtful
cloud and data investments to stay competitive. The bigger risk is doing nothing and watching competitors use AI to reset the benchmarks in their markets.”

Builders and value creators are enthused by AI’s potential to unlock revenue that scales non-linearly
with investments in human capital.

“You hear a lot of talk about building a billion-dollar company with 100 people, but the critical path to get there is building an AI augmented data practice along the way.”

Establishing a data practice that incorporates AI and data engineering augmented with AI will position early adopters to successfully leverage AI as it continues to mature. The overriding theme around AI at
AWS’s annual re:Invent conference last December was trust: AI needs to be trusted before it can be adopted in the enterprise.

To that end, companies will need to invest in security features around AI and agents. The hardest
security problems in large-scale AI integrations show up at runtime in production: agents misusing
tools, over-permissioned identities, prompt injection risks, data leakage through retrieval and logging, and weak auditability across agent actions. Trust cannot rely on prompt engineering or model instructions alone; what’s needed is enforceable controls around what an agent can do, when, and with which data.

Leto explained that builders on AWS can now use managed services to continuously evaluate and score how well agents behave across quality, safety, and business metrics. Tracking agent-level metrics like
goal completion and tool use accuracy not only helps mitigate risk and detect drift, but it also provides
a baseline for what better looks like as new iterations are released.

“This is how I’m helping companies selling into enterprises avoid clashes with infosec teams and close deals. A novel AI feature will get a prospect to watch the demo, but it’ll be the security and trust features that get customers to sign,” he said.

AI is all about the data

Embracing a data-centric strategy means tackling the decades-old challenge of dealing with data
fragmentation. Critical signals are scattered across batch data warehouses, real time streams, and unstructured content like documents, tickets, and call transcripts.

“When those systems don’t talk to one another, you end up with partial views of customers and operations, which makes both analytics and AI far less effective,” he noted. “For years, the payoff for unifying all that data felt marginal at best, so the enthusiasm for one more transformation project quickly faded. AI agents and models can now do some of the heavy lifting to retrieve and integrate data and deliver a step function improvements in analysis, insight, and automation. AI makes the business case for connecting all that data much more compelling.”

In Leto’s view, companies meeting this challenge are doing three things well: They’re defining shared data products that span batch, streaming, and unstructured sources; they’re investing in platforms that make those sources discoverable and governable behind a single access layer that agents can access; and they’re designing AI use cases that explicitly require and reward the integration so data strategy and AI
strategy reinforce each other. Leto’s forthcoming book, “Data Engineering with Generative and
Agentic AI on AWS: Building an AI-Augmented Data Practice for the Enterprise” takes data engineering
from the traditional pre-AI practice and puts it through the AI lens.

He posits that companies will need to invest in data platforms, modern data strategy, data governance, and what he calls the “last mile” of getting that data to agents, and addresses three hard constraints to the AI revolution: power, chips, and data.

“Access to compute has become a strategic advantage, but the blind spot for technology leaders is data readiness for AI,” he asserted. “Unlike power and chips, you can’t buy clean, contextualized company
specific data off the shelf, yet every useful AI system will require it for accuracy, relevance, and trust.”

Only one of the constraints Leto mentions are within the control of enterprise and technology
leaders: deciding to invest in their own data governance to prepare themselves and the company for
the AI revolution.

“Data governance is really on the critical path to enterprise value in the AI era.”

He stressed that a new generation of AI-focused data engineers will be crucial in this new wave.

“If businesses intend to operate with fewer people because of AI, it stands to reason that the value each
person delivers must increase proportionally. It’s incumbent upon leaders to take to heart just how much
more important human capital will be in this era, despite all the automation.”

In addition to writing and spending time with his family, Leto is an accomplished sailor. For nearly two
decades he’s crewed on boats of various sizes in multi-day offshore passages, using his downtime aboard ship to explore naval history.

“I was surprised to see parallels between how they were talking about the challenges of modernizing the U.S. Navy in the early 20th century and the AI revolution of today,” he revealed, citing the words of Admiral William S. Sims, a visionary reformer who served as the Commander of U.S. Naval Forces in Europe during World War I: “The transition from the Old Navy to the new, with its enormous material development, its complication of machinery and devices, its increases in size and numbers, its watertight subdivisions and its decentralization of authority, its material expansion and contractions there has been a tendency to lose sight of the man in the machine. No error could be greater than this, no mistake so fraught with disaster.”

Fortunately, the PE operating partners Leto supports have a skilled tactician helping them
navigate the rough seas ahead.

“Data Engineering with Generative and Agentic AI on AWS: Building an AI-Augmented Data Practice for the Enterprise” hits shelves on April 19 and can be preordered through Amazon.

Dive into the chapters and learn more on the landing page here to begin applying these concepts in real-world data environments.