As the Director of Delivery at APAX, my job is to make sure we’re all rowing in the same direction. Lately, the strongest current is Artificial Intelligence. The big question wasn't if we should use AI, but how to do it in a way that’s practical, effective, and feels like us.
We needed a human-centric guide, not a technical manual. So, we started with the systems that already define who we are.
Our Principles: Grounded in Our Values
At APAX, we run on the Entrepreneurial Operating System (EOS), and its foundation are our four core values: Be Good, Be Excellent, Be a Friend, and Be You. Instead of inventing a new AI philosophy from scratch, we used these values as our guide. This led us to a few key, common-sense principles for how we use AI.
- You Own Your Work: We see AI as a powerful assistant. But just like with any assistant, the final accountability rests with the person using the tool. We check the output, we understand it, and we are fully responsible for the final product.
- Keep a Human in the Loop: AI is here to augment our team, not replace it. The final judgment, the creative spark, and the strategic oversight will always come from a person. This ensures our work always has that essential human touch.
- Be Transparent: We're open about our use of AI, both with each other and with our clients. Using these tools is a strength, and it should never be a surprise when and how AI helped us get to a better result.
Our Playbook: From Play to Purpose
With our principles in place, we needed a process for bringing new tools into our daily work. For this, we were inspired by the "Total Motivation" framework, which shows that people are driven by a sense of play, purpose, and potential. We built our adoption process to reflect that. It starts with curiosity, not spreadsheets.
- Discover & Play: The process begins with exploration. We encourage everyone to be curious, experiment with new tools, and see what’s possible without the pressure of an immediate result.
- Align and Adopt: When a tool shows real promise, we align as a team. We decide what we’re using and, just as importantly, how we’re using it to ensure consistency.
- Train & Integrate: We teach each other the ropes and weave the new tool into our established, everyday workflows so it becomes a natural part of what we do.
- Upgrade & Evangelize: We’re never done. We constantly look for ways to get better results, and we share what we learn—the wins and the lessons—with our team and the wider world.
Our Actions: Focused on Impact
As we seek to get intentionally better every 90 days through Quarterly Rocks, several team members used the principles and playbook we created to apply AI to specific pieces of our software development process.
- Documentation: often a “nice-to-have” in the moment, but a lifesaver in the future, Rolando found that building more complete documentation is a great use for AI. From solidifying coding patterns and listing services used, to summarizing complex business logic and providing high-level project overviews, documentation reduces risk and saves money by making it easier to onboard new team members to a project (even if that’s “future you” coming back to a project).
- PR Review Assistance: a key role of our senior software engineers is reviewing chunks of work (Pull Requests) before it moves into production. With many engineers per each senior, this can create a bottleneck. Removing this step would eliminate a key quality check, so we’ve got to go through it. Ben found that specialized tools like CodeRabbit, CodeAnt, and even Claude Code help by providing a first-pass scan on code so that senior engineers can focus on the truly value-adding, human work of making judgement calls and prioritizing issues.
- Design-to-Dev Handoff: our design team works in Figma, our engineering team lives in VS Code. In theory, Figma makes this process easy, but there are always edge cases and incomplete information. Throw in a CSS framework like Tailwind and the questions start piling up. This is usually resolved through back-and-forth communication between designers and engineers, but Andrew found that integrating Figma’s MCP server with Claude Code helps smooth the process making it easier to maintain the fidelity of designs while translating them to code.
By grounding our strategy in our values and building a process that starts with curiosity, we’re not just using AI; we’re shaping it to fit our culture. It’s an approach that keeps us agile, accountable, and authentically APAX.
