Eric E.

The thing that made me so excited about Clay in the early days was that they saw the importance of integrations at Scale, and the thing that really sold me on it was Kareem’s conviction and his emphasis on creativity and not business. The main things to focus on when choosing between startups, as cliche as this may sound, is can you get behind the vision. I fundamentally believe that your work product is only as good as your conviction in what you’re working on. If you don’t believe in what you’re building, you’re leaving room for your own growth on the table

The way we decided what gets built next has changed over time. However, it started with people on a demo using Clay and telling us what was missing. That turned into customer requests as we grew. It continues to remain customer requests depending on our customers data and AI needs. We do a weekly review of every request and decide what would be the most impactful thing to build. Does anything need to be bumped up in priority from the current roadmap etc. The one that I think was used way more than I ever though, which is cheating a bit, is “Clay Send Data to other Table”. In the world of workflows, it’s a bit outdated now, but I remember one day looking at a literally 15 table chain and thinking “wow, this a level of complexity I could never have anticipated”. For the least used, I’d say Google Slides, because I think it’s the coolest thing I built, and I want to see people use it more. But, more similar providers are coming soon as well. Custom generated slide decks are just super fun and creative to me 🙂

The biggest thing that changed was our need for controls. We built Clay in the early days very much like the business we wanted to build. It was a business of full transparency, anyone can see anyones workflows and help them iterate on them. Transparency is key when you want to build full alignment. With that being said, as we continued to grow, Enterprise’s needed controls on who could use what in Clay, who could edit certain things in Clay, etc. As we’ve grown, it’s important to still aim for that transparent mindset, but also allow you to lock that down if you need to.

I’m not sure I would actually build Clay all that differently in full transparency, as I think that each step we’ve taken as defined who we are now, and I think each of those successes, mistakes, and periods of Clay are part of that development. Additionally, I think adoption of MCPs depends on the types of companies in your customer base. Larger enterprises may struggle to adopt MCPs as quickly as small to medium size teams that are focused on the cutting edge. So, I think it’s important to maintain both ways of using Clay. We want to adapt to your use-cases. In terms of how to use AI and Clay together, the biggest thing that I can say is that as Clay becomes your source of truth for your operational data, and we orchestrate your workflows, Claude can then use all of the data we have to make better decisions for the pieces of your business we don’t know about

In my mind, the most underutilized feature is also my favorite. I’d love to see more folks use our Google Slides (and similar integrations coming soon 👀) to generate dynamic slide decks, and allow you to be fully creative at scale.

The biggest thing I’ve learned as my literally job has become sharing context with all of Clay is that nobody wants to read docs anymore. So, the biggest thing that I can suggest is:

Focus on building the primitives that AI can build upon. You can get those primitives from existing libraries, existing public system prompts, looking at the open-source codebases of brands you admire. If you focus on having essentially an OS of primatives that an AI can build on top of, you’ll have less duplicated code, and less management of code later on

My favorite Clay feature that I built has to be our Google Slides integration. Before Claude Design, being able to generate custom slide decks off of a meeting brief, in seconds was incredibly powerful and just spoke to my brain as the creative engineer. The feature that I built that taught me the most lessons was Shared Tables in Clay. We viewed shared tables as the biggest growth driver at Clay because people would share their workflows and their brains to power the GTM community. What I learned was that, while the sharing is cool, a GTME’s workflows are their brand and are their business, so without some form of monetization, sharing may be tricky. What it influenced was me remembering that we’re helping other people build businesses on top of our system, and each feature we build needs to have that in mind

I’m going to loop this one into this response: Giving AI business context has become far more important than writing code 🙂

My planning process is gaining as much business context as humanly possible. Then, translating that into a spec that includes context on what’s already been built. Then I ask AI to present a plan for how it would build that spec, and I fact check each part of the plan and I ask more questions about why it made certain decisions. Then, I ask it to update the plan with my feedback and only after my final review does it start building. Because I work mainly on billing and enterprise, my delegation process is all about context building. AI doesn’t have great context on scale or proper billing processes, so the planning process becomes far more important