In our last post, we shared the chaos of trying to collaborate with two completely different writing workflows: I relied on Google Docs with ChatGPT, while my co-founder, Vagmi, preferred NeoVim with AI plugins. Syncing Markdown files from GitHub, dealing with AI model lock-in, and struggling with tools like Notion that couldn’t adapt to our needs made collaboration a nightmare. After multiple attempts at fixing our workflow, we realized that the problem wasn’t just about writing tools. The issue was deeper: AI was bolted onto existing tools as an afterthought rather than being embedded into the workflow itself.

AI Workflows Are Broken

Most AI-powered tools today suffer from three fundamental flaws: 🚨 AI as an extra step – You have your source of information in one place and AI in another, leading to inefficiencies. 🚨 Limited model flexibility – Many tools lock you into a single AI model, restricting adaptability and control. 🚨 No real-time interactivity – AI assistance typically comes in separate, asynchronous interactions instead of seamlessly integrating into the workflow.

We didn’t just want AI to assist us—we wanted it to augment our thought process in real time. That’s when it became clear: what we needed wasn’t another AI-powered writing tool. We needed an AI workflow engine that could power such tools at their core.

From AI Writing to AI-Powered Workflows

The real insight wasn’t just about improving our writing experience; it was about rethinking how AI fits into creative and operational workflows. The problem we faced with writing was just one example of a much broader issue that exists in multiple domains:

Sales and Marketing – AI should assist sales and marketing teams in real-time by generating personalized outreach messages, optimizing campaign content, and providing instant insights on customer data within their CRM systems.

Research & Analysis – Analysts should be able to query AI inline while writing reports, rather than switching between documents and chat interfaces.

Customer Support – AI should work within a team’s existing CRM and support tools, rather than requiring agents to manually extract and input information. By solving our writing workflow issue, we realized we were building something much more powerful—an AI-first workflow engine that could be adapted for multiple use cases.

What We’re Building

We’re designing an AI workflow engine that powers writing tools and beyond: ✅ Define workflows – Structured, reusable AI prompts and logic embedded directly into your workflow. ✅ Use any AI model – No lock-in to OpenAI, Gemini, or Claude. Choose the best model for the task. ✅ Iterate in real time – AI suggestions appear as you write and work, rather than as an afterthought.

To validate our ideas quickly, we’re first building this as an Obsidian Plugin—since Obsidian provides a solid Markdown-based foundation while allowing deep extensibility. However, this is just the beginning. The same AI workflow engine will eventually power a range of use cases across industries.

This isn’t just about writing. It’s about reimagining how AI augments human workflows at every level.