01-08-2025
How Data Infrastructure Quietly Drives The AI Engine
Michel Tricot is Cofounder and CEO of Airbyte.
AI cannot operate effectively without context, and that requires being fed by reliable data flows.
This is where I've watched most AI initiatives fall apart. Companies get excited about the potential of AI technology, but they forget about the unglamorous work of building data infrastructure. You can't just plug an AI model into your existing systems and expect magic to happen.
Reliable data pipelines aren't just nice to have; they are the foundation that makes everything else possible. You can't build a city without roads, plumbing or electricity, especially if they are not integrated in a well-organized central system. The same is true for AI applications.
To make reasoned decisions, AI technology must have the full context and information, and that can only happen by bringing all the right data together. This is why Salesforce is spending $8 billion to acquire Informatica, a legacy mainstay company in data integration technology. While this may not seem like an AI play, it is shaping up to be an essential part of Salesforce's AI strategy.
Salesforce's Agentforce aims to transform how businesses use AI agents to handle everyday operations. But there is a catch that I've seen trip up countless companies: AI agents can't function effectively without access to complete, well-organized data. Most enterprises still struggle with fragmented systems—data trapped in silos, scattered across platforms that don't communicate—making it difficult for AI to operate with full context.
This moment highlights a broader shift in enterprise AI strategy. Success increasingly hinges not on the model itself, but on the ecosystem that supports it. Forward-looking companies are beginning to realize that AI agents are only as effective as the infrastructure surrounding them, particularly the ability to unify and orchestrate data from across the organization.
Rather than treating data integration as a separate IT concern, the most strategic moves we're seeing involve embedding it directly into AI planning. This helps reduce friction and accelerate deployment while enabling AI to work with complete, real-time context.
However, it also introduces a new kind of trade-off: the more seamless the stack, the more locked-in an organization may become. Businesses must now weigh the benefits of speed and simplicity against the risks of vendor dependency and long-term flexibility.
I've always believed that open-source software fuels innovation, lets you customize and expand your tools, and most importantly, helps avoid being tied down to a single vendor. The Salesforce-Informatica deal marks a big change in how we think about AI in businesses. It confirms something I've long evangelized: the real advantage doesn't come from AI models themselves, since everyone will have access to similar ones. Instead, the true strength lies in how well a company can organize, protect and use its own unique data to power AI.
To do this, I recommend that organizations build their infrastructure to be scalable, portable and sovereign. Leverage open source and flexible data formats for your destinations, and ensure sensitive data is under your control, not being hosted and parsed by external parties. This blend of flexibility and sovereignty ultimately offers the best approach for modern data management.
The companies that understand this will be the ones positioned to successfully transform their businesses with AI. Those that don't risk building impressive demos that stall when faced with the complexity of real-world data environments.
Preparing for a data movement now can help businesses ensure they're ready to build scalable, effective AI applications.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
