Loomfield AI
What Loomfield is for: recurring reports, reviews, analyses, and follow-ups
Loomfield is for work you do more than once in a similar shape: a monthly report, a variance question, a quarterly comparison, a client follow-up. These look like different jobs, but they lean on the same files, checks, business context, and judgment. Loomfield is built to keep that shared method, so each new job starts set up instead of from scratch.
Status: Loomfield is in private beta and the learning loop is in development. The examples below illustrate the intended behavior using synthetic scenarios, not customer results.
What kinds of work is Loomfield for?
Recurring reports, analyses, reviews, meeting preparation, investigations, and follow-ups. Reports make the idea easiest to see, but the same approach applies wherever you repeat the same kind of thinking: find the right sources, run the same checks, make familiar judgment calls, and present the result in a format people expect.
How does a monthly report make the next job easier?
Take July's performance report. While you build it with AI, you point to the files that matter, correct a misread number, and settle on a layout. Normally all of that stays inside one chat.
Loomfield is designed to keep those sources, checks, corrections, and formatting connected to the finished report. When the August review begins, it starts with the trusted sources and structure ready, while changes and approvals remain visible to you.
Can the method carry over to a different task?
That is the point. A variance question, a quarterly comparison, or a client follow-up is not the same task as the monthly report. It can still inherit the analysis method that worked, such as which source is authoritative or which check catches a common error, without being forced into the shape of the last job.
Is there a working example today?
One is built in. Loomfield comes with a Weekly Review agent already scheduled for Monday mornings. You can open the dashboard's Getting started checklist or the Schedule page and click Run now to see what it does. When you are ready, the Agents page lets you build your own. The getting-started guide shows where each of these is in the app.
How much does Loomfield do without asking?
Only as much as you allow. Work can stop at any of four levels, and different work can stop at different ones:
- Prompted: AI helps while you lead and decide.
- Proactive: Loomfield notices due work or missing inputs and brings them to you.
- Supervised: it gathers sources, runs checks, and prepares a draft, then waits for your approval.
- Automated: stable routines run on schedule within approved boundaries, and exceptions come back to you.
Who is Loomfield for?
People who already pay for an AI subscription, such as Claude Pro or Max or a ChatGPT plan with Codex, and who do recurring knowledge work. You do not need to code, use an API key, or design a workflow first. If you can name one report, review, analysis, or follow-up you keep rebuilding, that is a good place to start. Request access from the homepage.