Alice, AI Task and Content Studio
Task tracking and content generation in one studio, writing in a real person's voice.
The problem
Generic AI writing is easy to spot and easy to ignore. The hard part is sounding like a specific, credible person while writing at volume. The briefs, deadlines, and approvals around the writing usually sit in a separate tool, so the work gets split across two places.
What I built
A studio that holds the tasks and the content together. A brief goes through a pipeline of five agents: a Router decides what kind of piece it is, a Retriever pulls reference posts and the author's own drafts, a Generator writes variants, a Critic scores them, and a Polisher tidies the ones that pass. A copilot works on tasks, drafts, and the library through tools, and every write shows a confirmation card first. The same tools are exposed through an MCP server, over stdio and HTTP, so other agents can drive the studio.
How it remembers
Voice comes from three layers: a style learned from each author's own drafts, a shared house style, and an optional brand modifier. Trends are stored as facts with an expiry date. Duplicate topics are merged by meaning, and anything a person has edited is protected from being overwritten by a later run.
My approach
Each stage of the pipeline is small enough to test alone. The Critic scores every variant on voice match, hook, clarity, and platform fit, and if polishing lowers a score the original wins.
The result
On-brand writing at the pace of a team, in a voice readers recognise as the author's, with the tasks behind it tracked in the same place.
Stack
Connect
Building something, or thinking about it? Book a 1:1 and let's dig in.
Gbolagade Ishola