# How I Use AI Daily

> How I live and breathe AI as both practitioner and creator, from building Plumb to using AI in every part of my workflow.

Source: https://curiouslychase.com/posts/how-i-use-ai-daily/

Zapier's application asked:

>How are you currently using AI in your personal and professional life?

I love this question, because my mission is building the future of AI-assisted work, both as practitioner and creator.

# I Live and Breathe AI

For the last 2 years as founding engineer and technical co-founder at [Murmur Labs](https://curiouslychase.com/posts/murmur-labs/), I built [Plumb](https://curiouslychase.com/posts/plumb/), an AI workflow automation platform where builders could use AI in deterministic workflows that they could sell to their subscribers. 

We used AI for generating output schemas and prompts. We built a proof of concept where anyone could generate workflows through natural language ("[Magic Mode](https://curiouslychase.com/posts/magic-mode-the-workflow-builder-you-deserved/)"), then refine them visually: bridging speed and precision. Before that, Supermanage (another product I built) used a complex graph of data transformations and AI nodes to create a 1:1 brief based on Slack data for managers to be prepared for their 1:1s.

One of my favorite uses at [Plumb](https://curiouslychase.com/posts/plumb/) was leveraging our mdc files in [Cursor](https://curiouslychase.com/tools/cursor/) for the shape and structure of a new node. We could use [Cursor](https://curiouslychase.com/tools/cursor/) to generate new nodes from a spec (90% implemented in the product), then do the last 10% to ensure the integration did what we wanted. This allowed us to create new nodes and integration configurations really quickly.

Another favorite: my product manager and I would meet for an hour every week and walk through the features we wanted to build or bugs we wanted to fix. We'd record our call using Granola or Spark email meetings, take the transcripts, and run them through a workflow we built. It would determine which features and bugs to extract, run each through an AI prompt with a template for our spec, then create entries in our Notion product roadmap database. It allowed us to discuss things in an unstructured way, then have AI create structure from our conversation. I'd use those specs as the starting point for features, combining them with our mdc files in [Cursor](https://curiouslychase.com/tools/cursor/) to generate 90% of the code we needed, sometimes in one request.

Daily, I use Claude Agents Parallel Worktrees for development, run [voice-to-sense-making-to-content pipelines (audio monologue → transcript → Notion → newsletter draft)](https://curiouslychase.com/thinking/the-high-throughput-zero-writing-documentation-system/), and use [Claude](https://curiouslychase.com/tools/claude/), Factory Droid, and [Raycast](https://curiouslychase.com/tools/raycast/) for various tasks.

I also write a lot about how I use AI in my daily work in my newsletter, raising the tide for others in our industry and beyond. I helped a friend who's a top-selling Amazon author use Claude Code projects to track his character arcs and story development.

I've taught workshops on using AI in daily work and written about [using AI as a collaborator](https://curiouslychase.com/posts/how-an-ai-collaborator-can-amplify-your-abilities/). I use AI every day to riff on ideas, ask for feedback on things I've written or done, and generate prompts for creating consistent but creative images.

I co-hosted [AI Builders Club](https://curiouslychase.com/posts/ai-builders-club-welcome-to-the-club/), a podcast dissecting practical AI automation workflows (inbox management, meeting prep, proposal generation) with actual tools and honest "would-this-work?" assessments.

Lately I've been using AI ([Claude](https://curiouslychase.com/tools/claude/) or Factory Droid) to generate interactive components based on my writing. It's been amazing to see how TypeScript/ShadCN/Tailwind is so easy to generate with a single prompt if you have a good spec for the core design and what you want to achieve.

# Core Belief
Deterministic, schema-driven workflows that can be triggered by a user in a chat (we also built a proof of concept for workflow invocation through chat), scheduled, or by integration events beat black-box agents for mission-critical work.

AI is amplifying my velocity to 10 one-day sprints instead of one two-week sprint.