A portfolio manager asking a simple question—“Which positions drove active risk this quarter and why?”—may require data from multiple systems to perform attribution, analyze factor exposures and benchmarks, review market events, reconcile calculations, and manually piece together insights before they can be shared with clients or presented to a risk committee. Although the intelligence already exists within the organization, the challenge is connecting it together.
In enterprise AI, competitive advantage increasingly hinges on how well firms connect the intelligence they already have. The differentiator has moved beyond who builds the most powerful models to who can connect data, systems, and workflows in a secure and scalable architecture, turning fragmented information into intelligence that drives better decisions.
Every major wave of enterprise technology has solved a different challenge. Databases organized information. Enterprise applications digitized business processes. Cloud computing transformed infrastructure into an elastic resource. APIs connected previously isolated systems. Artificial intelligence introduced the potential for reasoning.
Today, the next challenge is to enable intelligence to move securely, consistently, transparently, and in real time across the enterprise so that it reaches the workflows where decisions are made.
As a result, the conversation is moving beyond large language models toward interoperability, standardized interfaces, governed data access, and architectures that allow AI to participate safely within enterprise workflows rather than operate alongside them. This is the emergence of connected intelligence.
Given organizations have already invested billions of dollars building trusted data platforms, analytical models, governance frameworks, and business applications, their investments remain enormously valuable. The next generation of AI systems should amplify the earlier investments rather than replace them.
For example, imagine the portfolio manager asking another question through a trusted AI interface: "Which positions contributed most to our active risk this quarter, and how does that compare with last quarter?"
Instead of the PM searching across multiple systems, AI assembles the response nearly instantly from validated analytics, governed data, and existing risk models so that calculations are transparent, numbers are traced back to their source, and conclusions are supported by the same trusted analytics the organization already relies upon.
That’s the fundamental shift occurring in enterprise AI; the architecture itself is becoming the defining source of competitive advantage. Organizations are prioritizing architectures that reduce the distance between intelligence and decision making. They will have architectures that allow trusted intelligence to move securely, seamlessly, and transparently across the enterprise.
The first major breakthrough in enterprise AI was retrieval augmented generation (RAG). It solved a fundamental problem: large language models had no direct access to current, proprietary, or contextual information. RAG has enabled AI to retrieve trusted information from enterprise documents, databases, and research repositories. That enables AI to ground its responses in verifiable data, improve accuracy, and reduce hallucinations.
It was a major step forward. However, retrieval alone doesn't enable action.
Consider an analyst preparing for an earnings call. RAG can quickly retrieve and summarize revisions in expected or actual earnings, shifts in analyst sentiment, and changes in peer valuation multiples. However, the analyst still needs to update financial models, validate assumptions, prepare presentation materials, notify stakeholders, document the analysis, and develop investment recommendations across multiple enterprise systems. RAG delivers the intelligence, but it doesn't execute the workflow.
Workflow execution is the gap that Model Context Protocol (MCP) is starting to close. MCP enables AI to:
Identify available tools, systems, APIs, governed analytics, permissions, and enterprise workflows
Understand how they work
Interact with them safely and consistently
Together, RAG and MCP transform AI from a conversational interface into an driver of action in enterprise workflows. Instead of simply answering questions, AI can retrieve trusted information, invoke governed analytics, coordinate actions across enterprise systems, and help move work from insight to decision to generate measurable business value.
As AI moves closer to influencing real investment decisions, trust becomes the defining requirement. Every insight, recommendation, and workflow must withstand scrutiny from portfolio managers, research analysts, risk managers, compliance officers, auditors, and regulators.
A model can be impressive, a response can be fast, and a recommendation can even be directionally correct. Nevertheless, none of those capabilities matter if the result cannot be traced, validated, and reproduced across governed systems. Intelligence that cannot be defended cannot be trusted.
In the earliest AI interfaces, success meant receiving a useful answer. In capital markets, success requires understanding why the answer is trustworthy. Every result should be able to answer questions such as:
Where did the data originate?
Which version of the dataset was used?
Which analytics engine produced the result?
Which assumptions and methodologies were applied?
Can the same result be reproduced under the same conditions?
We have long recognized that need. For example, with the FactSet Workstation for datasets such as FactSet Fundamentals, users can trace calculated values back to their underlying financial statements, formulas, and source filings, allowing analysts to understand exactly how a metric was derived and sourced.
Similarly, when AI surfaces insights from unstructured content, FactSet's document intelligence capabilities can link responses directly to the location in the filing, transcript, or research document where the citation originated so users can immediately verify the source and context.
Consider the portfolio manager reviewing performance attribution. Knowing that a position contributed positively or negatively to returns is the starting point. The PM requires confidence the attribution was calculated with the correct pricing source, benchmark, risk model, portfolio holdings, and methodology.
Now consider a compliance officer reviewing portfolio exposures. A summary alone is insufficient. The officer must be able to trace every figure back to governed data sources, documented methodologies, entitlement controls, and the firm's compliance policies.
In both situations, uncertainty is a real operational risk.
That is why governance, lineage, transparency, and reproducibility are architectural requirements. AI systems operating in financial markets must explain how they reached every conclusion and consistently reproduce the same result under the same conditions.
Given financial institutions have spent decades building governed analytical environments for consistency, auditability, and trust across teams, enterprise AI must integrate and operate within those environments, not outside of them. When AI is embedded into governed financial systems rather than loosely connected to them, it becomes part of the institutional decision-making fabric.
Connected intelligence becomes operational by creating a common framework for interactions, where MCP architectures shift AI from passive retrieval to active participation and execution. They enable intelligence to move through systems with context intact, respecting permissions, preserving lineage, and maintaining traceability from end to end.
In practical terms, an AI system is no longer just answering “What changed in portfolio risk?” It can also execute the downstream workflow as it accesses the latest risk analytics from governed models, compares changes against prior periods with consistent methodologies, surfaces drivers of change across exposures and factors, generates an audit-ready summary with full lineage, and routes outputs into reporting or investment-review workflows.
In other words, AI is now embedded within the system to participate in it and drive action.
Enterprise finance platforms have been moving toward connected intelligence embedded directly into trusted workflows.
Initiatives such as our FactSet Portfolio Analytics MCP and collaborations with cloud ecosystems including Google Cloud reflect a broader industry transition from static data delivery toward interoperable, governed intelligence layers. The intent is to extend the existing analytical foundations.
Portfolio analytics, risk models, performance attribution, and market data become callable, governed components that AI can safely interact with. As a result, computation moves closer to the point of decision and preserves the rigor on which financial institutions depend.
Data remains governed and traceable while analytics remain consistent and institutionally validated. AI becomes the interface layer that orchestrates them. Decisions happen faster within the same control framework. The result is amplification of existing systems.
As AI architecture matures, a new hierarchy of competitive advantage is emerging. It’s defined more structurally, such as who can move intelligence most effectively across the enterprise without breaking trust, governance, or workflow integrity.
Organizations that solve the challenge of intelligence fragmentation by connecting models, data, analytics, and workflows into a coherent intelligence network will fundamentally change how decisions are made.
The future of enterprise AI will also be defined by how well the models become enterprise aware. Intelligence that can move securely, transparently, and reliably across the systems where real decisions happen is where competitive advantage begins.
The shift is already underway.
Organizations that gain the greatest advantage from AI will build the architecture that enables intelligence to move securely and seamlessly across data, agents, workflows, and people. That is the promise of connected AI and the vision realized through FactSet Intelligence.
Learn more:
This blog post is for informational purposes only. The information contained in this blog post is not legal, tax, or investment advice. FactSet does not endorse or recommend any investments and assumes no liability for any consequence relating directly or indirectly to any action or inaction taken based on the information contained in this article.