Agents are ready but your architecture probably isn't

2026-04-29 · Source: Databricks · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure, Robotics & Autonomous Systems · Depth: Intermediate, medium

Summary

Databricks co-founder Arsalan Tavakoli-Shiraji observes that many enterprise AI initiatives generate activity without delivering value, primarily because organizations often start with technology rather than desired outcomes. Most companies are still in experimental or task-automation phases, leading to "AI sprawl." A common architectural mistake is underestimating the complexity beyond model selection, particularly regarding data connectivity across disparate systems, robust governance for agentic actions, and deep semantic understanding of organizational context. Traditional dashboards and batch pipelines are inadequate for agentic systems due to latency and slow response times. Databricks' Lakebase addresses this by providing a transactional database optimized for the low-latency, high-scale demands of agentic applications, complementing existing analytical layers. Effective governance is crucial for agents that take actions, as they lack human situational awareness, necessitating a "prerequisite" approach to permissions and oversight.

Key takeaway

For VPs of Engineering or Data grappling with stalled AI initiatives, recognize that simply adopting models without re-evaluating your underlying data architecture and governance strategy will lead to "AI sprawl" rather than value. Focus on defining clear business outcomes first, then ensure your transactional data infrastructure (like Lakebase) and agent governance frameworks are robust enough to support agentic actions at scale, rather than retrofitting AI onto outdated systems. Your system is your strategy.

Key insights

AI value stems from outcome-driven design, robust data infrastructure, and proactive governance, not just model selection.

Principles

• Start with desired outcomes, not technology.
• Governance is a prerequisite for agentic systems.
• Rethink processes, don't just add AI on top.

Method

To achieve successful AI agent deployment, define specific success outcomes upfront and establish small, isolated pilot teams to iterate quickly, free from legacy constraints, before scaling learnings across the organization.

In practice

• Prioritize data connectivity and semantic understanding.
• Implement transactional databases for agentic workloads.
• Treat agent permissions as a critical governance layer.

Topics

• Agentic Systems
• AI Governance
• Data Architecture
• Lakebase
• Enterprise AI Strategy

Best for: VP of Engineering/Data, CTO, Director of AI/ML, AI Architect

Related on AIssential

• The Agentic Enterprise Just Got Real — AI's primary challenge is context management, not intelligence, requiring robust governance for agentic systems.
• Architecting Data And AI In The Era Of Enterprise Intelligence: Meet Shylaja Nathan, Principal Analyst — Agentic AI demands modern data architectures that provide context, responsiveness, and semantic consistency at scale.
• Your AI Strategy Isn’t Failing Because of the Model — It’s the Data — Data governance and quality, not AI models, are the primary bottlenecks preventing enterprise AI from scaling to production.
• A Strong Suite in Agentic Strategy: Data Apps — AI agent success in enterprises depends on architecting governed, queryable data applications, not just advanced models or semantic layers.
• What it takes to scale agentic AI in the enterprise — Scaling agentic AI demands robust governance, data infrastructure, and system integration, not just advanced technology.
• The Software Collapse & The New Stack — The agentic era shifts software from human-informed systems of record to autonomous systems of action, inverting traditional SaaS value.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Databricks.