Data in 2026: Interchangeable Models, Clouds, and Specialization

The New Stack looks ahead to data in 2026 and predicts a world of specialization + interoperability: interchangeable models, more portable cloud choices, and AI systems built from coordinated components (including agents) rather than one monolithic stack. Franz Inc. CEO Jans Aasman was interviewed for the piece, sharing how knowledge graphs and multi-model strategies can make agentic systems more accurate, auditable, and sustainable. If you’re tracking where the modern data/AI stack is heading, it’s worth reading the full article.

Dr. Jans Aasman, CEO, was quote throughout the article:

“All the inputs and outputs of the agents, every decision, goes into the orchestrating knowledge graph.”

Why it matters: agent systems will need an authoritative “memory + audit layer” for governance, traceability, and accountability.

“You have three or five LLMs read a document… and then there’s a resolver system… so we ultimately get data that’s 99.9% correct. If you use one, it might be only 60% correct.”

Why it matters: accuracy will increasingly come from model collaboration + reconciliation, not just picking a single “best” model.

“It tells me what’s possible and I say what I really want, and it says ‘let’s try this’… ‘this is cool, please store this in a visualization,’ and it keeps going.”

Why it matters: the interaction pattern is shifting toward AI-as-coach/collaborator that iterates with users and produces durable artifacts (like visualizations/structures).

“We can’t sustain the current investments in AI… This is unsustainable. So, the only way to go forward is with small models.”

Why it matters: economics will push architectures toward smaller, specialized models—which increases the need for interoperability and orchestration across many components.

Visit The New Stack to read the full article.