‍Last month at the AI Training Company, we cancelled a $300/month SaaS contract because we built our own. 

‍

Like most off the shelf software, our old CRM was a generalist, performing lots of different functions at a mostly satisfactory level. It wasn’t a bad service, but for $300/month in this builder's climate, you have to do better than satisfactory. 

‍

So last month, we pulled the plug. Not because we migrated to a competitor, but because we built our own, fully customized CRM and task manager using Claude Cowork, in about two weeks. 

‍

Automating individual business functions with ‘Skills’

‍

Looking over our utilisation, we saw that we were using maybe 20% of the CRM features we were paying for. 

‍

This is what kicked us into gear. Before we replaced the tool entirely, we wanted to see if it was possible to automate some individual business functions. 

‍

Enter Claude Cowork and its ‘Skills’. You can think of these as custom recipes of logic and automation ready to be dropped into any project. 

‍

We started by automating the high friction tasks that historically live across three different browser tabs. These included:

‍

• The client quotation bot: We built a skill that takes raw meeting notes and instantly generated a structured, professional quote. 
• Daily conductor: This skill handled our daily preparation, organisation and delegation. It checks the pipelines and tells us exactly who needs to do what to keep things on track. 
• Post-training packs: After a session, we were spending at least an hour compiling tailored resources for packs. This skill builds and formats these packs in seconds. 
• The universal style guide: We probably use this one the most. It fits into every project ensuring that each line of text, or bit of code matches our brand voice and technical standards.

‍

‍

Connecting the dots and building ‘AITC Hub’

After a few weeks of  playing around with these skills, we realised they had quickly become key to every team member’s daily work. In fact, the skills were doing about 80% of what our expensive CRM was supposed to do. 

‍

It was clear at this point that with Claude’s capabilities and our team’s AI expertise, that we had what we needed to make something totally unique that worked for us. 

So, we took these skills and used Claude to weave them into a unified task manager and CRM. Because we built it from the ground up, there’s zero fat and zero waste. The tool has:

‍

• No cluttered menus scattered with features we never use
• No rigid data structures that don’t fit our training model
• Total ownership of our data and processes

‍

The result is a dashboard that has exactly what we need at this given moment in time. If that changes, we’ll be quick to pivot with updated skills.

‍

‍

Is it ‘enterprise ready’?

‍

Do we think another, larger company would migrate from any of the major software providers  for our customer build tool? Probably not. At this stage, our CRM doesn’t have the requisite million integrations or 50-page data security whitepaper. 

‍

And with the business moving fast, there’s a lingering concern that we could be creating a headache down the track if our custom CRM starts flaking out.

‍

But for us, right now, it’s perfect. It’s fast and easy and most importantly it’s saving us $300 a month.

New standards for SaaS

‍

We are entering an era where good enough Saas isn’t cutting anymore. Particular for small business and agency customers. Why would a team of 5 or 10 pay thousands for a platform built to service enterprise clients, when they can spend a few weeks designing a solution tailored exactly to their needs. 

‍

Of course, if you do choose to take the builder’s route, there are plenty of roadblocks you’ll have to navigate. For example, our team works in the Google suite, and syncing Claude’s code with Google Drive remains a manual hurdle, with asset generation across two platforms getting a bit clunky. 

‍

We’ve also found Claude has a tendency to squash PDF content together in ways that require a human eye and there remains a lot of back and forthing, trying to nail the more technical elements of the build.

‍

So, are you still paying for features you don't use, or are starting to trim the fat by building your own?

‍