Salesforce subsidiary Informatica's CLAIRE multi-agent AI system, embedded across the Intelligent Data Management Cloud (IDMC), executes enterprise data workflows at a 90% task success rate and reduces month-long manual efforts to days by coordinating 50–60 specialized model calls per request through an orchestration agent with deterministic tool routing and validation checkpoints.
"Single-agent AI systems fail in enterprise workflows because they cannot simultaneously handle context-heavy reasoning, diverse toolchains, and multi-step execution. Each step — search, profiling, statistical analysis, rule generation, and code execution — demands different inputs, tools, and reasoning patterns that a single agent cannot reliably manage in one pass," Neha Awasthi, Senior Manager of Software Engineering at Salesforce, told the Salesforce Engineering Energizers Q&A series. Early single-agent designs that exposed multiple tools to one agent "selected incorrect tools, exceeded context limits, and produced inconsistent outputs across steps," driving the team to specialize.
Enterprise data workflows historically required business users and engineers to navigate multiple disconnected tools across discovery, governance, data quality, integration, and master data management. A business user identifying an issue had to pass it through translation, rule creation, validation, and pipeline integration across systems — each handoff introducing latency and context loss. CLAIRE collapses these stages into a continuous workflow orchestrated by AI agents that can interpret intent, operate across heterogeneous systems, and execute workflows spanning multiple domains.
CLAIRE predates "agents" as an industry term, born from the inherent complexity of enterprise data workflows that demanded autonomous, coordinated execution rather than copilot-style single-step assistance. The team initially built single-agent designs that exposed many tools to one agent, then decomposed the system into multiple specialized agents — each operating within constrained context and an optimized toolset — coordinated by an orchestration agent responsible for intent detection, plan generation, and routing.
The architecture has three layers: an orchestration agent (control plane) for intent detection, plan generation, and routing; specialized agents such as a data quality agent that bundles profiling, rule recommendation, rule generation, and cleansing with dedicated tools and model configurations; and a planning layer where users can review and modify high-level plans before execution. Deterministic tool routing ensures agents invoke the right tools based on intent, while validation checkpoints and strict data contracts between agents prevent invalid outputs from propagating downstream through the 50–60 model calls in a typical workflow — critical because a single malformed metadata output early on can otherwise cascade into total workflow failure several steps later.
CLAIRE operates across the IDMC for cataloging, governance, data quality, integration, and master data with two interaction models: CLAIRE GPT as a generalist interface for cross-domain requests, and specialized copilots embedded inside individual products for in-context execution. Workflows that previously required up to three months of manual effort now complete in days, with the multi-agent system executing at a 90% task success rate. The architecture also supports background and headless execution — agents can run data quality assessments, rule validation, and pipeline health monitoring autonomously on scheduled or event-driven triggers without a user present in the conversation.
The team continues to invest in adaptive planners that recompute dependencies when users modify execution plans in real time without disrupting in-flight workflows, and in deeper guardrails that contain anomalies early in execution. Going forward, the focus is on expanding autonomous data quality operations at scale and extending coverage to more heterogeneous enterprise data systems and domains.