← All work
Case study | Claude skills · MCP · custom GPTs · Gemini gems

AI tools built for your business

A Claude skill, a Gemini gem, a custom GPT. Your process, your systems and your files built into a tool anyone on your team can use by asking for it.

The problem

There are jobs your business does the same way every time. Drafting a particular kind of letter, putting together a client summary, checking a file against a standard, researching a property or a counterparty before a meeting. The procedure exists. It lives in the head of whoever has done it longest, and it takes them an hour each time.

Handing that to a general purpose AI doesn't fix it, because the model doesn't know your process, your systems or your files. Every request starts by explaining all of it again, and most people explain about half, so what comes back is generic and someone spends twenty minutes fixing it.

What gets built

A named tool that lives inside whichever AI your team already uses. The procedure is written down once, your systems and files are built into it, and using it is a matter of asking the AI. Your brand voice enforced on everything that gets drafted. A document put together the way you put documents together. A research routine that checks the sources you would check. The design standards you want followed, actually followed.

What holds it together

It meets people where they already work. Nobody opens a second application or changes what they do. The tool is inside the conversation they were having anyway.

The procedure is written once and used by everyone. The quality stops depending on who is asking and how well they phrase it.

It's easy to pick up. There's no new software to learn and not much to explain. People ask for what they need in their own words.

Where this fits

Anything your team does repeatedly with AI where the output has to match how your business does it. If you can describe the job to a new hire, it can be built into a tool. I build with AI every day, and most of this work is teaching a business to do the same with their own material behind it.

An example: one I built and shipped

I had a problem when I started developing tools with AI. Work that gets redone because what was decided last time isn't written anywhere the tool can reach. So I built a tool to save my notes right from my AI. In the middle of whatever I'm doing, I say "save this," and it's filed. Later, in any AI I use, I ask for it back and it's there.

That's the same mechanism your tool would use. Say what you want in plain language, and the procedure, the files and the systems behind it do the rest. You can check out the Claude skill at scribld.io.

Schedule a consult