AI "Skills" Are Not What You Think

Anthropic released Agent Skills as an open standard in December. Salesforce’s Agentforce platform now serves over 18,500 enterprise customers, with Agentforce accounts in production growing 70% quarter over quarter. OpenAI has adopted a structurally similar skills architecture across ChatGPT and its Codex CLI.

Three very different companies arriving at the same architectural conviction, which should tell us something worth paying attention to.

A skill, in this context, is a structured set of instructions that teaches an AI system how you do your specific work.

Your conventions, your workflow logic, your domain vocabulary, your quality standards, packaged once and reusable indefinitely. Anthropic defines them as markdown-based directories that agents load dynamically when a task calls for it. Salesforce builds them into Agentforce actions and flows. OpenAI organizes agent capabilities as modular tools and function calls.

The format differs but the underlying idea is identical: codify the way work actually gets done so that an AI agent can perform it reliably without being re-taught every time.

To understand why this matters, consider what it replaces.

For decades, the most valuable institutional knowledge lived in people’s heads. The senior analyst who knew exactly how to structure a board memo for a particular CFO. The sales operations lead who could configure a forecast model accounting for every regional quirk because she had been burned by every regional quirk. The compliance officer whose contract reviews were better than anyone else’s, not because he had internalized twelve years of edge cases.

That knowledge was perishable. It walked out the door every time someone left, and replacing it took years, if it got replaced at all.

“Skills” change the economics of expertise.

When a financial analyst builds a skill that encodes how their firm evaluates risk (the specific weightings, the regulatory language, the formatting conventions that leadership expects) they are not writing a prompt. They are creating institutional memory that compounds. Every time an agent invokes that skill, it performs at the level of someone who has internalized years of context without requiring those years.

Take the spreadsheet as an example. When VisiCalc and Lotus 1-2-3 arrived, every accountant could use pre-built templates. That was useful. But the accountants who could construct their own models from scratch, custom formulas and macros and structured logic tailored to their actual business, had a fundamentally different relationship with the technology. They did not just use the tool. They taught it how to think about their problems. Over time, that capability went from impressive to expected.

We are at this same inflection point with AI skills.

Pre-built skills from Atlassian, Canva, Figma, Notion, Cloudflare, Stripe, and Zapier are already available in Anthropic’s partner eco-system. Salesforce’s out-of-the-box agents can prospect, qualify leads, and route cases on day one. That is valuable and it is table stakes.

But the person who builds a custom skill encoding how their team actually runs quarterly business reviews, or how their compliance department evaluates vendor contracts, or how their editorial voice should sound across a dozen content formats, that person has created something no marketplace can sell them.

The skill captures judgment, not just process, and judgment is the part that does not generalize.

Andrej Karpathy and others have argued for “context engineering” as a better frame than “prompt engineering,” describing it as the discipline of filling an AI’s context window with exactly the right information for each step. Ethan Mollick at Wharton has noted that this approach encodes business logic itself.

Skills are context engineering made durable. A prompt disappears after the conversation ends. A skill persists, improves, and compounds, which is exactly what institutional knowledge is supposed to do but rarely does.

I have been writing for three years that when intelligence becomes infrastructure, judgment becomes the differentiator. Skills are how that judgment gets operationalized.

Learning to build them, job-specific, task-specific, shaped by whatever domain expertise you have spent years accumulating, will be a core competency for knowledge work going forward. Not just for developers. For every analyst, strategist, operator, and leader who works with information for a living.

Every time intelligence has been turned into infrastructure (the printing press, the public library, the internet) the initial fear was displacement. What actually happened was expansion. More people creating, more knowledge circulating, more problems getting solved that could not have been addressed before. AI skills are the next layer of that same pattern. We are building the infrastructure for it to matter at a scale that was never previously possible.

The best part is that the barrier to building your first skill is absurdly low.

You do not need to write code. You do not need an engineering team. You need to know your work well enough to explain how it should be done, and then package that explanation in a format an AI agent can follow. That is it.

The people who start building now, even imperfectly, will have a library of compounding institutional intelligence by the time their competitors are still debating whether to try.

The window for early advantage is open right now.

Until next time,

Ram

PS - here’s a full guide from Anthropic on how to build “Skills” → download here: https://resources.anthropic.com/hubfs/The-Complete-Guide-to-Building-Skill-for-Claude.pdf

— 

Ram Srinivasan

MIT Alum | Author, The Conscious Machine | Global Future of Work and AI Adoption Leader published in Business Insider, Fortune, Harvard Business Review, MIT Executive Viewpoints and more.

—

A Message From Ram:

My mission is to illuminate the path toward humanity's exponential future. If you're a leader, innovator, or changemaker passionate about leveraging breakthrough technologies to create unprecedented positive impact, you're in the right place. If you know others who share this vision, please share these insights. Together, we can accelerate the trajectory of human progress.

Disclaimer:

Ram Srinivasan currently serves as an Innovation Strategist and Transformation Leader, authoring groundbreaking works including "The Conscious Machine" and the upcoming "The Exponential Human."

All views expressed on "Substrate" and across all digital channels and social media platforms are strictly personal opinions and do not represent the official positions of any organizations or entities I am affiliated with, past or present. The content shared is for informational and inspirational purposes only. These perspectives are my own and should not be construed as professional, legal, financial, technical, or strategic advice. Any decisions made based on this information are solely the responsibility of the reader.

While I strive to ensure accuracy and timeliness in all communications, the rapid pace of technological change means that some information may become outdated. I encourage readers to conduct their own due diligence and seek appropriate professional advice for their specific circumstances.