US-United States
AU-Australia
CA-Canada
CH-China
EEN-Eastern Europe and Nordics
FR-France
DE-Germany
IT-Italy
JP-Japan
KR-Korea
LATAM-Latin America
META-Middle East
NL-Netherlands
ES-Spain
UK-United Kingdom
From Data Mesh to AI Mesh: Taking the Next Step
Published: October 8, 2025
Every major shift in technology brings changes the organizational structure, to processes and responsibilities. Take a look at the shift to cloud: before centralized IT departments were managing large on-premises infrastructures, handling hardware, networking, and software deployment. Developers had limited control over infrastructure. After the migration the structure has shifted to decentralized, cross-functional teams (e.g., DevOps, FinOps, CloudOps) and Infrastructure as Code (IaC) enabled automated provisioning and IT operations evolved into cloud governance, security, and cost optimization roles. As enterprises recognized the importance of data, different technology and organizational approaches emerged, one of the recent and successful approaches being Data mesh.
As Wikipedia defines it:
“Data mesh is a sociotechnical approach to building a decentralized data architecture by leveraging a domain-oriented, self-serve design (in a software development perspective), and borrows Eric Evans’ theory of domain-driven design and Manuel Pais’ and Matthew Skelton’s theory of team topologies. Data mesh mainly concerns itself with the data itself, taking the data lake and the pipelines as a secondary concern. The main proposition is scaling analytical data by domain-oriented decentralization. With data mesh, the responsibility for analytical data is shifted from the central data team to the domain teams, supported by a data platform team that provides a domain-agnostic data platform. This enables a decrease in data disorder or the existence of isolated data silos, due to the presence of a centralized system that ensures the consistent sharing of fundamental principles across various nodes within the data mesh and allows for the sharing of data across different areas.”
That’s a mouthful — but here’s the essence: data mesh is about moving from centralized control to decentralized ownership. Instead of one big central team managing all the data, responsibility shifts to the domain teams but with a federated governance, and tech stack that the business teams can leverage — the people closest to the business problems. Data becomes a product, accountability increases, silos break down, and overall quality improves.
Data mesh marked a paradigm shift from a centralized structure to decentralized ownership, empowering teams to treat data as a product and fostering better outcomes. My view is that AI should follow the same route as the technology matures, and early indicators and lessons learned show that this is the right approach.
When something new and disruptive like AI and multi-agent systems (MAS) emerge, the first instinct is to often centralize them — to create a single, fast-moving team that can innovate quickly, this is great for getting started as that approach creates speed, but it’s not enough on its own. To be truly effective I the long run, you also need to embed experts from that team inside the business units and make the technology accessible to the people with domain expertise.
The goal isn’t just training new models. It’s about equipping AI agents with the knowledge and tools to be productive in their domain — and then connecting those agents through a larger multi-agent system that enables cross-domain solutions. Finding the right balance between centralized efficiency and decentralized ownership will be crucial for success.
Just as organizations once had data ambassadors, today we need AI (and Data) ambassadors embedded directly in the business units to drive change from within. These ambassadors serve as catalysts for transformation — not just adopting AI, but rethinking business processes with an AI- and data-first mindset. They work side by side with the AI experts from the central team and the business teams to foster understanding, encourage adoption, and ensure the responsible use of artificial intelligence within each domain.
At the same time, organizations need system thinkers who can connect the different domains. This is what makes AI truly powerful at the enterprise level — unlocking outcomes that were previously out of reach. Beyond efficiency and automation (which are expected and no longer differentiators), this approach enables companies to design entirely new experiences and products for their customers. It’s how organizations can stay ahead of the competition, disrupt themselves before others do, and build a foundation for long-term innovation.
How could this materialize in different verticals? Let’s look at a few examples.
Banking
Imagine a Personalized Financial Wellness Agent. Instead of one monolithic system, it’s powered by multiple specialized agents: one analyzes spending patterns, another evaluates credit history and risk, a third pulls in market data and investment options, and another manages customer interactions and compliance. Together, they create a unified financial assistant that can deliver personalized budgeting, credit improvement plans, and tailored investment suggestions — aways on, delivering insights in real time, every day of the year.
Venue and events
This is already a major vertical for Extreme Networks in the U.S., and it’s expanding rapidly worldwide. Imagine a Smart Event Experience Agent also powered by multiple agents working together. One manages ticketing and seating, another optimizes crowd flow and safety by guiding attendees to key services, while other agents handle food & beverage recommendations, entertainment scheduling, and real-time offers. Combined, they create a personalized itinerary, that helps guests avoid queues, discover new experiences, and enjoy exclusive offers based on their preferences and location.
As enterprises embrace - AI-driven - transformation, the principles of data mesh — decentralization, domain ownership, and platform thinking and governance — remain relevant as ever. But to unlock the full potential of AI agents, we need to take the next step: what I call an AI Mesh. The AI Mesh approach brings together domain experts and AI specialists in a distributed model that embeds intelligence directly into business processes. By striking the right balance between centralized expertise and decentralized ownership, organizations can build cross-domain systems that fuel innovation, efficiency, and create entirely new customer experiences.
Enterprises that embrace AI Mesh won’t just optimize for today — they’ll invent the customer experiences of tomorrow.