Osmoti - AI Marketing Department for Small Businesses

An AI marketing department that small business owners run by text. Owners talk to Osmo, the agent, and it runs ads, social, lead follow-up, and reviews behind their approval.

Role: Co-Founder & CTO. Team: 2 co-founders: Om Patel (CEO) and Rahul Mehta (CTO). Status: Live. Year: 2026.

Overview

Osmoti gives small businesses an AI agent, Osmo, that owners talk to in a web chat or in iMessage. Osmo drafts and publishes ad campaigns, posts to social, texts back new leads, and asks customers for reviews. Every action waits for the owner's approval or runs inside limits they set, and revenue is traced back to the ad or post that produced it.

Osmoti is live with paying customers. I co-founded it with Om Patel (CEO) and, as CTO, built the platform end to end: the agent, the integrations, the approval system, and the owner app.

How it works:

  • The owner texts Osmo or opens the web app. Osmo knows the business, its brand, and its connected accounts.
  • Osmo drafts ad campaigns and social posts. Nothing goes live without the owner's yes.
  • New leads get a text back, either automatically or after the owner approves the reply.
  • Review requests go out to customers, and Osmo helps respond to reviews.
  • Booked revenue is tied back to the source ad or post.
  • Spend limits, an audit log, and undo keep the agent inside the lines the owner draws.

Real results: An Atlanta med spa turned $360 in ad spend into $6,500 of booked work in its first 14 days, with zero hours from the owner.

Website

Key features

  • Owner-facing agent, Osmo, in a web chat and in iMessage
  • Ad campaigns on Meta and Google, drafted by the agent and published only after owner approval
  • Social posting across the major platforms
  • Text-message lead follow-up in automatic or approve-first mode
  • Review requests and responses
  • Revenue attribution from ad or post to booked work
  • Guardrails: spend limits, approvals, an audit log, and undo

Technical details

  • Frontend: React and TypeScript
  • Backend: Node.js, Prisma, and Postgres
  • AI: Large language models for reasoning and drafting, plus generative image and video
  • Integrations: ad platforms, social platforms, text messaging, and payments

Challenges

  • Letting an AI agent act on real ad accounts and customer messages without losing the owner's trust
  • Making approvals fast enough to happen over text, so owners actually use them
  • Tracing revenue back to the specific ad or post that produced it
  • Connecting many third-party platforms behind one agent

Tags: AI Agents, TypeScript, React, Node.js, Postgres