The first phase of enterprise AI was about generation. Models proved they could summarize documents, answer questions, and produce drafts that sound plausible. That race is effectively over, and access to a capable model is now table stakes. As AI becomes the interface for financial decision-making, the question that matters has changed. It’s no longer who has access to the best models. It is who controls how those models are applied to real-world analysis.
Skills are emerging as the mechanism that defines that control. A Skill encodes the data, the methodology, and the standards that govern how a piece of analytical work is performed, and makes that encoding available to an AI agent on demand. In doing so, Skills move enterprise AI from generating text to producing decision-grade output. For financial services, where the consequences of getting it wrong are regulatory as well as reputational, that shift is decisive.
Where generic AI fails
Most professionals who have experimented with AI have had the same experience. The technology is clearly powerful, but the output needs so much checking, reformatting, and supplementing with real data that the time saved starts to feel marginal. You end up doing the work twice: once to get the AI's draft and again to make it usable.
In regulated, high-stakes workflows, this is disqualifying. A general-purpose model knows a great deal about the world in general and nothing about how your institution does things. Its output cannot be traced to authoritative sources, it might not follow a defined methodology, and it might not even be consistent from one run to the next. Each of those failures alone would rule it out as the basis for a credit decision, a KYC determination, or anything a firm would defend to a supervisor. Together, they explain why so many enterprise AI initiatives remain impressive demos rather than production systems.
Skills are the control in AI
A Skill closes that gap by encoding institutional knowledge as an executable process. It is a structured specification that captures the steps, the standards, the data sources, and the quality bar for a specific piece of work. The model already knows how to read, reason, and write. The Skill tells it how to do those things your way, every time.
The simplest analogy is a chef and a recipe card. The model is a skilled chef who already knows technique, ingredients, and flavor profiles. The Skill is the recipe card: same chef, new dish, no retraining and no rebuilding the kitchen.
What makes Skills the control layer, rather than merely a productivity feature, is that they govern the three things financial institutions most need to govern.
Control over data. Every output is grounded in specified sources. When a Moody's Skill produces a Public Information Book, it draws from Moody's proprietary ratings, financials, research, and risk intelligence rather than a general-purpose web crawl. The data behind every claim is traceable and auditable.
Control over process. The analytical steps, the structure of the deliverable, and the quality standards are encoded, not improvised. Every run follows the same process, which means the output is consistent regardless of who triggers it or when. Where a prompt says “analyze this”, a Skill specifies the seven steps of the analysis, the order in which to apply them, the conditions under which to deviate, and the format in which to present findings with every assumption flagged and every source cited.
Control over composition. Skills are modular and designed to work together. An Earnings Brief can feed a Peer Analysis, which can feed a broader report, building complex analytical workflows from reliable components rather than monolithic, fragile prompts.
Why this matters most in financial services
Every industry benefits from consistency, but in financial services it is the price of admission. Outputs must be consistent, explainable, and defensible. An analysis that cannot show its sources will simply not survive an audit. A process that produces different answers on different days cannot support a rating, a lending decision, or a sanctions determination. Repeatability is not an efficiency metric here, but a genuine regulatory requirement and the foundation of client trust.
This is why the bar for AI in financial services has never been “can it generate accurate text” but “can it produce work that meets the standard required to act on.” Skills are how that bar gets met. They make AI-assisted analysis traceable to authoritative data, repeatable across runs and users, and aligned to institutional methodology. That combination is what separates decision-grade intelligence from confident-sounding text.
Moody's point of view: expertise as the execution layer
This is where Moody's holds a structural advantage. Our Skills are not built on generic data or abstract workflows. They encode decades of credit expertise, risk methodology, and connected intelligence, and they run against the same authoritative data layer that powers our analytics, delivered through Moody's Model Context Protocol (MCP) servers. MCP is an open standard that lets an AI model connect directly to external data and tools, rather than relying on what it learned in training. This is the execution layer that sits on top of Moody’s Connected Intelligence — the knowledge graph, MCP Servers, and authoritative data architecture described in our earlier work.
The architecture is what makes that advantage durable. A Skill is the encoded methodology, the steps, the standards, the format. It runs against Moody’s authoritative data through our MCP servers, the governed data layer underneath. That separation is the point. The SKILL.md format is open, and anyone can write one; what they cannot replicate is a Skill that executes against Moody’s ratings, research, and risk intelligence through the same data layer that powers our analytics. The methodology and the data are inseparable, and both belong to Moody’s. Giving the format to an open standard does not give away the advantage, because the advantage lives at the data layer, not in the file.
The first wave covers the workflows where that expertise is most concentrated: an Earnings Call Summary across multiple companies; a Peer Analysis with investor-grade comparison tables; a Public Information Book that builds a full dossier on a single entity; a Rating Pitch covering sector context, rating history, and peer positioning; and a Sector Analysis combining Moody’s proprietary research with live market intelligence.
Each of these represents hours of expert work, compressed into a single natural-language request, and each produces output that is structured, sourced, and formatted to institutional standards. The analyst still reviews and still applies judgment. What changes is where their time goes. The hours of gathering, structuring, and formatting collapse, and what remains is the high-value analytical thinking where human expertise matters most.
Where the market is going
The direction of travel is already visible in the market. The SKILL.md format has quickly become an open standard, with dozens of platforms, including those built by Microsoft, Google, Amazon, and OpenAI, now reading the same file format and a large and growing library of Skills published across the ecosystem. A Skill built for one platform works on any compatible platform, which means the institutional knowledge it captures is a durable, portable asset rather than a prompt that only works in one tool.
For Moody’s, that portability is the strategy, not a side effect. A client working in Claude, in Microsoft Copilot, or in their own internal pipeline does not get a different Moody’s — they get the same Skill, running against the same data, because the methodology travels with the MCP layer beneath it. We go to the environment the client has already chosen rather than asking them to come to ours. Build the expertise once; deliver it everywhere our customers work.
As this model takes hold, Skills will do more than improve productivity. They will define how analytical work is executed across the industry, and the firms that lead will be those whose expertise is embedded directly into that execution layer. Models will keep changing hands at the top of the leaderboard. The methodology encoded in a well-built Skill library, connected to authoritative data, belongs to the firm that built it and compounds with every production run.
That is the position Moody's has already built. Decision-grade intelligence, powered by Skills, running wherever our customers work. In a market where everyone has access to the same models, the durable advantage belongs to whoever controls how the work gets done.
Interested in learning more about Moody's GenAI solutions?