Claude, ChatGPT, and Gemini function like operating systems. Here's how to start using them like one.
This piece is adapted from an Operators Guild Focus Session on AI as infrastructure, led by Aneesh Sachdeva, Founder of Glia Intelligence. Glia works directly with companies like PayPal, Opendoor, and Sphinx to help teams become AI-native and rebuild how they get work done. Focus Sessions are small-group, member-only conversations where operators compare notes on decisions in flight, pressure-test tradeoffs, and surface the operational realities that rarely show up in polished frameworks or vendor decks.
If you want access to sessions like this, including the recordings, live demos, and the community behind them, you can apply to join OG.
Most operators are drowning in new AI tools. A new product lands in the inbox every week. Vendors are upselling AI features on contracts you already own. Startups are offering free design partnerships in exchange for your process. The noise is real, and finding the signal is genuinely hard.
The teams that actually change how they operate tend to have one thing in common. They stop looking for new tools and start building on what they already have.
What they have being Claude, ChatGPT, or Gemini.
These platforms are not apps. They function more like operating systems, the same way Windows or macOS sits underneath everything else you do on a computer. Once you start treating them that way, the question shifts from “what tool should I buy for this?” to “how do I build this capability on the platform I already have?” That shift changes everything about how fast you can move.
Stop adding tools, start adding connectors and skills
Most teams underuse these platforms because they treat them like a smart search bar. Ask a question, get an answer, move on. That approach gets you maybe 10% of what is available.
The real leverage comes from two things: connectors and skills.
Connectors are integrations. They give Claude access to the systems your team already touches every day: email, Slack, your CRM, your data warehouse, your documentation. Without connectors, you have a very capable model with no visibility into your actual work. With them, you have something that can act on your behalf across your entire environment.
Skills teach Claude how your team works. Not just what to do, but how you do it: your templates, your preferences, your process, your standards. Think of a skill like a thorough onboarding doc, the kind you would hand a new hire so they could get up to speed without shadowing you for two weeks. Write it once, and Claude carries it forward every time it does that type of work.
The combination is what makes transformation possible. Connectors give Claude access. Skills give it context. Together, they turn a general-purpose platform into something that actually knows your business.
Three levels of adoption
There are three levels to how teams typically build out their infrastructure, and the progression matters because each one unlocks meaningfully more capability.
The first level is personal productivity. Connect your calendar, email, and communication tools. At this stage, with no custom skills built yet, Claude can already function as a capable executive assistant: triaging your inbox, flagging action items, resolving scheduling conflicts, making sure your meetings are set up correctly. Most teams are surprised by how much this first level changes their day-to-day before they have written a single skill.
The second level is cross-functional operations. Add your documentation tools, project management systems, and CRM. This is where Claude starts functioning more like a chief of staff. One example: a team that connected Jira and Confluence used Claude to run a thorough review of all engineering pods before quarterly planning, analyzing burndown rates, surfacing bottleneck ticket types, and comparing actuals against the roadmap. No custom skills required. Just the connectors, and Claude had the context to do the work.
The third level is financial and data infrastructure. Connect your data warehouse, your ERP, your financial systems. This is the most sensitive data, and also where teams have seen the biggest transformations. When an FP&A team connected their Snowflake instance to Claude and built a skill that taught it the business context behind their data, their decision speed increased tenfold in two days. They went from asking two or three questions per day to thirty. They avoided backfilling a departing analyst role. And the work their analyst org does today has shifted entirely: the job is no longer to do the analysis, it is to help Claude become a better analyst for everyone in the company.
The shift from doer to director
Becoming AI-native is not primarily a technical skill. The most important capability is learning to be a good manager and reviewer of AI’s work. When you were doing the work manually, review was embedded in doing. You caught errors in the process of producing output. When Claude is doing the work, review becomes its own explicit step, and the quality of your direction determines the quality of what you get back.
The fundamental shift for teams that make this transition: everyone becomes a manager. That means giving clear instructions, checking outputs against expectations, and iterating on the process when something is not working. The operators who move fastest are the ones who already think like that.
A practical framework for build versus buy
With AI reshaping what is possible on general-purpose platforms, procurement decisions need a new starting point.
Two questions worth asking before buying anything new:
- Does it integrate with Claude? Every tool you add outside your primary platform fragments the network effects you are building. If a new tool does not connect back to Claude, you are adding a silo.
- Are you still owning your data and process? One of the underappreciated costs of adding vendors is that your process becomes theirs. Design partnerships exist because your operating knowledge has value. Keeping that knowledge inside your platform protects it.
Some categories are worth buying: specialized agents for customer support, tools for labor you were already outsourcing, platforms where the ROI on external infrastructure clearly outweighs the cost of fragmentation. The question is whether you have exhausted what your existing platform can do before you go looking.
Most teams have not.
Go deeper with OG
This session was just one example of the work happening inside OG every day.
If you are an OG member, you can watch the full recording and access the session presentation, and connect directly with Aneesh in Circle. He is also hosting in-person follow-on sessions in New York and San Francisco for members who want to go hands-on, see a live demo, and get tactical advice on how to apply this inside their own team.
If you are not yet a member and want access to sessions like this one, including recordings, working examples, and a community of senior operators comparing notes on what is actually working, you can apply to join OG today.