4 trends that will define data management and AI in 2026
2025 was about building agents. 2026 is about trusting them. As agents move from pilots to production, context, interoperability, automation and platform consolidation will set the agenda.
If you lead a team or a budget, these are the moves to make now - before the vendor talks start driving your roadmap.
1) Context becomes non-negotiable: semantic models go front and center
Connecting agents to data is table stakes. Connecting them to meaning is the difference between useful and unreliable.
Semantic models - the shared definitions, business logic and relationships that give data its meaning - stepped out of the background in 2025 and will be a strategic priority in 2026. As David Menninger put it, accessing data without a semantic model is like driving without a roadmap.
Vendors such as AtScale, DBT Labs, Google's Looker and ThoughtSpot have offered semantic layers for years, but agents raise the bar. "Companies realize their AI doesn't fail for lack of intelligence, but for lack of business context," said Baris Gultekin of Snowflake. Agents also need a shared vocabulary across departments or their reasoning breaks down.
There's work to do. Menninger notes most models still fall short on metrics - too many definitions trapped in SQL that can't express real business logic. The Open Semantic Interchange initiative is pushing for progress, but buyers should press vendors to expand beyond simple SQL expressions.
What to do- Fund an enterprise semantic layer with clear ownership and a backlog tied to top use cases.
- Standardize business metrics outside of SQL-only definitions; require vendor support for richer logic.
- Make semantic governance part of production-readiness for every agent project.
2) Interoperability moves from hype to decisions: A2A vs. MCP expansion
Multi-agent systems need a way to talk, coordinate and hand off work. That's where agent-to-agent protocols come in.
Google Cloud's Agent2Agent (A2A) launched in April and has backing from AWS, Microsoft, Oracle, Databricks, Snowflake and others. It merged with IBM's Agent Communication Protocol in September 2025, narrowing fragmentation. Still, Donald Farmer cautions that A2A is only essential when you're actually running agent swarms - adoption will follow real enterprise demand, not vendor enthusiasm.
Others expect MCP to absorb A2A-style capabilities. "This market rewards the protocol that becomes the default, and MCP is rapidly becoming that default," said Chris Aberger of Alation. Expect convergence - not coexistence - as enterprises push for one standard.
What to do- If you're not running multi-agent workflows at scale yet, avoid locking into a single protocol.
- Prioritize vendors that support both MCP and A2A-style patterns or commit to compatibility.
- Design for loose coupling between agents so protocol changes don't break production.
Resources:
3) Agent-fueled automation accelerates - with guardrails
With context and connectivity in place, automation will extend from data plumbing to business workflows. Think data quality monitoring, pipeline generation, catalog enrichment - and then customer support, finance tasks like invoice processing, and supply chain optimization.
Examples are already live. Monte Carlo's agent monitors data quality. Informatica's Claire Agents help with data exploration and pipeline building. ThoughtSpot is rolling out agents that automate dashboard development, semantic modeling and embedded BI. "We've barely scratched the surface," said Menninger.
What to do- Pick three high-friction workflows with measurable ROI and automate end-to-end with human approval gates.
- Define runbooks, escalation paths and decision guardrails before scaling to more use cases.
- Track cost, cycle time, error rates and compliance impact, not just "time saved."
4) Cost pressure and complexity drive consolidation
Agentic AI is expensive and complex compared to traditional analytics. That's pushing buyers toward unified platforms and pre-integrated stacks - and prompting more M&A across data integration, transformation and streaming.
Deals across 2025 showed the direction of travel, and more are likely in 2026. Farmer expects data catalogs, observability and ETL to be active targets, with private equity consolidating fragmented middleware. At the same time, advanced teams will keep a few specialized tools where they see a clear edge, as Satyen Sangani notes.
The buyer's job: simplify without boxing yourself in. Balance platform consolidation with selective best-of-breed where it truly matters.
What to do- Decide your default stance: platform-first with targeted specialists, or best-of-breed with strict TCO gates.
- Add consolidation and change-of-control clauses to contracts; demand exit paths and data portability.
- Standardize on common governance, lineage and security controls across all tools to reduce integration drag.
Your 90-day plan
- Fund a semantic layer program tied to 3 priority agent use cases; publish a metrics catalog.
- Run one multi-agent pilot with explicit handoffs and observability; keep protocol choice flexible.
- Automate two production workflows with human-in-the-loop controls and clear KPIs.
- Rationalize your vendor list: cut overlaps, set platform standards, and document migration paths.
The shift is clear: from experimenting with agents to trusting them in production. Invest in context, pick an interoperability path that won't trap you, automate where the numbers prove out, and simplify your stack before it complicates your plans.
Upskill your team
Want practical training for managers and operators rolling out agentic AI? Explore role-based programs here: Courses by Job.
Your membership also unlocks:
