The Platform User Evolution
From Analytics Consumers to Agent Architects/Engineers
Courtesy: napkin.ai
The intent of this post is to equip enterprise leaders, architects, and implementers with the clarity and depth needed to understand the potential and pitfalls, digest key system-level details, and confidently embark on this multi-year transformation journey—overcoming skepticism and caution along the way.
As we go through the early journey of talking to more enterprise leaders, we realize this isn't an easy journey and conversation around the topic is as important as actual execution to bring about "real change".
In this post, we look at shifts in roles and beachhead technology platforms that form the current state of enterprise architecture and skill inventory largely created by adoption of the previous transformation journey of Data and ML. Understanding inventory of current state of beachhead platform, technology and roles, understanding of organisational issues will help us come up with the enterprise agent strategy factoring many of those.
You may also want to read on the trends that Sanjeev Mohan and I wrote about 2025 data and AI trends.
We introduced Applied AI as a new category with newer platform and user personas. Platform personas included Agent Management System and Multi-agent systems as newer imaginations to watch out for.
On the data track, we positioned the Intelligent Data Platform as the enterprise’s strategic beachhead for unlocking all data and AI insights.
While the technical architecture of Intelligent Data Platform continues to mature, This post focuses on former, which is Agent Management System and Multi-agent systems.
As I engage with more and more conversations around Enterprise AI agents. It is becoming clear that we are experiencing a dramatic shift in user-persona and therefore an equally profound need that will transform who uses these platforms and how they approach their work.
This persona shift represents one of the most significant yet underappreciated changes in enterprise AI adoption.
The Traditional Data & ML Platform Persona (2020-2024)
The incumbent user archetype across platforms:
Data Engineers building ETL pipelines and maintaining data infrastructure
Data Scientists training models and running experiments for human consumption
Analytics Engineers creating dashboards and reports
Business Analysts consuming insights to inform human decision-making
Current User Workflow
Raw Data → Processed Data → ML Models → Human Insights → Business Decisions
Current Success metrics
Model prediction accuracy
Dashboard engagement rates
Time-to-insight for human users
Query performance optimization
The early signs of need for Agent-Centric Persona (2025+)
We're now seeing the rise of the AI Engineer, Agent System Architect as enterprises adopt agents and insert agents as part of the overall "Agent Strategy"
Agent System Architect and AI Engineer — is a fundamentally different user with different goals, workflows, and success metrics:
Building autonomous/semi-autonomous systems(HITL) rather than human-consumed analytics
Orchestrating multi-step workflows where agents execute business processes end-to-end
Managing agent-to-agent interactions rather than human-to-dashboard experiences
Optimizing for process completion rather than insight generation
New Workflow
Create Autonomous/Semi-autonomous Agents or Agentic Workflows (with HITL) → Test Agents -> Deploy Agents -> Deliver Business Impact -> Monitor Agents -> Iterate for issues and next value
This rise of AI Agents and AI engineer/architect persona shift fundamentally changes how platform success is measured
New Agent-Centric Success Metrics:
End-to-end process completion rates
Human intervention requirements
Agent reliability and uptime
Multi-agent coordination efficiency
Business outcome achievement without human loops
Agent First Platform Architecture
While a lot of these agent projects today start by picking a Foundation Model, an agent framework/an agent builder, an eval framework. This is unlikely how the enterprise agent story can pan out over a 1-3 year period.
Success with Enterprise Agentization need an Agent First Platform (Agent Management System/Multi-agent systems) at a strategic level which will reinforce this new AI user persona and accelerate creation and adoption of Enteprise Agents at scale.
A brief on Agent-First Evolution:
Data Connectors to Unified Data & Apps Connectors: As agents query and use tools directly, we are seeing collapse of ETL and Reverse-ETL technologies. Data connectors were often uni-directional, they brought data into data and AI platforms, however with agents we need bidirectional flow and further diversity in upstream applications. Agents need to be enabled to take actions.
Unified Agents Catalog & Governance layer for agents and multi-agent systems: Catalog and governance for agents subsume data and ml catalogs. Data catalogs should not just include data onboarded from other systems, but also importantly enable creation of synthetic data for training, data apps, that enable trajectory data capture for common use-cases, data generated by AI Agents themselves and making them available and formatted for the next iteration of the training run.
Data Exploration: With process and trajectory data becomes far more important, we need newer imagination on how do we visualize and explore trajectory flows.
Value Shift: Value shifts from "What insights can we extract?" to "What autonomous processes can we enable?"
Orchestration: Orchestration is not just about orchestrating SQL engine or a jupyter notebook or an ML model but also about Agent orchestration and Sandboxes that can load and execute a sub-agent or an agentic tool on the fly.
Monitoring and Observability: Investment in agents and multi-agent observability capabilities that compress agent OODA loop (build-deploy-iterate cycle)
Deployment: Emphasis on enterprise-grade agent deployment with end-to-end agents and multi-agent system reliability.
Positioning: From Data cloud to Agent cloud.
Collaboration: Collaboration must now address A2A, H2A, A2H all combination of need
Strategic Implications for Enterprises
Organizations must recognize that the highest-value use cases are shifting from "better human decisions" to "fewer required human decisions." This means:
Investment Priorities: Budget allocation should favor agent orchestration capabilities over traditional BI tooling
Talent Strategy: Hire for agent system architecture and engineering skills, not just core data engineering/analysis and model capabilities.
Platform Selection: Evaluate vendors based on agent management system, multi-agent system capabilities, not just data and AI platforms.
Success Measurement: Define KPIs around process automation, not insight generation
The Convergence point between current platforms and agent first platform architectrue.
This new persona and platform thinking is critical for AI agents driven business transformation over the next 1-3 years and needs to be a factor as enterprises plan to strategically move in that direction. The clear intent is to reinforce this new role personas with platformization of capabilities that enable and accelerate the adoption of agentization story. Its important to think and put the same on your strategic planner and realign your platform and deployment strategies accordingly.
Agent first platform architecture will achieve significantly compressed data-to-action lifecycle and sustainable competitive advantages through autonomous/semi-autonomous process execution with Human-in-the-loop(HITL) and provide enteprises to adopt agents at scale.