Box CEO Aaron Levie and Rahul Pathak, VP of Data and AI GTM at AWS, recently sat down to talk about what it really takes to make enterprise AI work. Pathak asserts that success in turning proof-of-concept AI experiments into mature programs hinges on connecting advanced models to enterprise data, operational processes, and production-ready infrastructure — with Amazon Bedrock, open standards, and a pragmatic modernization strategy sitting at the center of that vision.
Key takeaways
- AWS sees 2026 as a breakout year for agentic AI in the enterprise
- Enterprise AI success depends on connecting models to enterprise data
- Enterprises should innovate while they modernize rather than wait for perfect data
- AWS and Box work together to help connect AI systems to the content and workflows businesses already run on
How can enterprise data become an AI differentiator?
The differentiator between customers that make it into production with AI versus those that stay at the proof-of-concept phase is the integration of enterprise data. Pathak says, “It's the combination of enterprise data about the business, about customers, and all of that embedded knowledge — coupled with state-of-the-art AI — that allows customers to build unique, differentiated experiences.”
Enterprise AI becomes more valuable when it’s grounded in business data rather than generic prompts alone.
That insight leads to a shift in perspective for many enterprise leaders. Advanced models are becoming more widely available, but proprietary business context is not. Internal documents, customer records, workflows, knowledge bases, images, and operational docs are what make enterprise AI useful in practice.
Box helps organizations unlock value from all of their unstructured enterprise content across documents, files, and images. Combined with AWS infrastructure and AI services, that content layer gives enterprises a way to connect AI systems to the information that powers their businesses.
Enterprise AI becomes more valuable when it’s grounded in business data rather than generic prompts alone.
How does Amazon Bedrock help enterprises scale AI?
AWS positions Amazon Bedrock as a key part of its enterprise AI strategy because it gives organizations model optionality without forcing them to rebuild their systems every time the market changes.
As Pathak explains, Amazon Bedrock allows customers to maintain stable APIs while choosing among different models. That matters in a fast-moving AI environment. Enterprises want the freedom to adopt new model capabilities, but they also need consistency, governance, and maintainability.
Amazon Bedrock allows customers to maintain stable APIs while choosing among different models.
Amazon Bedrock gives teams a stable platform for both experimentation and production. Instead of locking into a single model approach, organizations can evolve their AI stack over time while preserving the surrounding infrastructure. That flexibility is a strategic advantage. It helps reduce operational friction and makes it easier to move from pilots to production-scale AI workflows.
When’s the right time to integrate AI into legacy systems?
A major theme in the Box and AWS conversation is that enterprises should not wait for perfect data or fully modernized systems before starting serious AI work.
Enterprises don’t need perfect conditions to begin. They need a clear business objective, access to the most relevant enterprise data, a platform to get their legacy content onto the cloud, and an architecture that can improve over time. According to Pathak, "You have to innovate while you modernize. Don't wait for clean data. First, there's no such thing.”
I think you have to innovate while you modernize. Don't wait for clean data. First, there's no such thing.
He continues, “Today's technology with agents and models allows us to make sense of messy data in ways that we couldn't before. The role that [Box] plays in managing enterprise data, coupled with the AI and infrastructure that we can bring to bear, is a great unlock for customers.”
Pathak maintains that enterprise leaders should treat modernization and AI adoption as parallel efforts. That approach can help organizations build momentum sooner and avoid delaying value creation.
How should enterprises approach AI adoption?
Pathak recommends starting with business objectives rather than AI for its own sake. That means identifying the outcome the organization wants to improve, determining the minimum data needed to support the use case, and putting the right guardrails in place.
Those guardrails include security, responsible AI practices, and financial controls. Once those foundations are established, teams can move faster with greater confidence.
For organizations early in their AI journey, Pathak recommends practical use cases such as:
- Coding and agentic assistance for employees
- Customer support
- Customer relationship management
For more mature organizations, the opportunity expands to workflow redesign, higher throughput, lower operational costs, and new forms of business value.
Why do open standards matter for enterprise AI?
Enterprises need confidence that the systems they build today will remain useful even as models, tools, and orchestration frameworks continue to change. Open approaches can reduce lock-in and make it easier to adapt over time.
In the conversation with Box, Pathak pointed to standards such as Model Context Protocol (MCP) as part of that foundation. For enterprise buyers, this matters because long-term AI adoption depends on durable architecture, not just short-term experimentation.
AWS combines flexibility with stable foundations, a combination that can help enterprises scale AI confidently across changing technologies and business needs.
What’s the next frontier for enterprise AI?
Pathak’s view of the future goes beyond task automation. The next frontier is building systems that can improve workflows over time.
Pathak described customers moving toward systems where outputs feed back into processes and help improve future performance. That points to a more advanced model of enterprise AI: not just assistance, but compounding operational improvement.
This is where agentic AI becomes especially powerful. When organizations combine automation, evaluation, feedback loops, and trusted guardrails, they can create systems that become more effective the more they are used.
The implications extend to productivity, decision-making, and the redesign of work itself. From Pathak’s perspective, the goal is to provide the infrastructure, model access, and flexibility needed to support increasingly adaptive enterprise systems.
Why are AWS and Box a strong enterprise AI combination?
Enterprise AI depends on both intelligence and context.
AWS brings scalable cloud infrastructure, AI services (including Amazon Bedrock, Amazon Bedrock AgentCore), model optionality, and the architectural foundation for agentic AI. Box brings the secure content layer that helps organizations unlock value from unstructured data across documents, files, and images.
Together, AWS and Box help solve one of the biggest customer challenges in enterprise AI: connecting powerful models to the content, knowledge, and workflows businesses already rely on.
That makes the collaboration more than a standard integration story. It’s a practical answer to a real enterprise problem. Organizations need a way to connect AI to their data, modernize without stalling innovation, and build systems that scale securely over time. AWS and Box are positioned around exactly that need.
FAQ
What is agentic AI in the enterprise?
Agentic AI in the enterprise refers to AI systems that can do more than answer questions. These systems can participate in workflows, take action across business processes, and help teams execute tasks using enterprise data and operational context.
Why is enterprise data important for AI adoption?
Enterprise data gives AI systems the context they need to produce useful, accurate, and business-specific outputs. Without enterprise data, AI often remains generic and disconnected from real workflows.
How do AWS and Box work together for enterprise AI?
AWS provides cloud infrastructure, AI services, and Amazon Bedrock. Box provides access to enterprise content and unstructured data. Together, they help organizations connect AI systems to the information and workflows that drive real business operations.
Final takeaway
AWS's enterprise AI strategy centers on helping organizations operationalize agentic AI by combining model optionality through Amazon Bedrock with enterprise data, open standards, governance, and production-ready infrastructure. Enterprise AI success will be defined not by model access alone, but by the ability to connect AI to real workflows and real business context.
For enterprise leaders, the message is straightforward: start with the business objective, connect the right data, establish guardrails, and move now. The organizations that do this will be best positioned to turn AI from a promising technology into a significant operational advantage.
