Deepground
Prototype
AI Taskforce Platform
AI Taskforce Platform
I'm prototyping Deepground, an AI taskforce for brick-and-mortar small businesses. Owners get a team of specialist agents that answer their calls, book their jobs, and help them show up better online.
I'm building it end to end, from the marketing site and free grader to the dashboard, agent platform, and integrations.
I'm building it end to end, from the marketing site and free grader to the dashboard, agent platform, and integrations.
Problem
Small business owners are on a job, with a customer, or closed when the phone rings. Callers who reach voicemail usually try the next business, and the owner's website often can't answer a question or take a booking either.
Existing options are an answering service that only takes a message, or software that takes a developer to set up. I wanted owners to get a working team of agents without writing a prompt or touching an API key.
Existing options are an answering service that only takes a message, or software that takes a developer to set up. I wanted owners to get a working team of agents without writing a prompt or touching an API key.
Free Grader
Every owner starts by searching for their business. The grader scans their Google Business Profile and website, scores their online health out of 100, and estimates the monthly revenue they're missing. No signup required.
The estimate is a transparent model with stated assumptions, shown as a range instead of a single number, so owners can see exactly where the money is leaking.
The estimate is a transparent model with stated assumptions, shown as a range instead of a single number, so owners can see exactly where the money is leaking.
A report built from a sample business
The Taskforce
Each report ends with a recommended taskforce: specialist agents for reception, scheduling, follow-up, invoicing, reviews, and analytics, matched to the gaps the grader found. Signing up turns those recommendations into real agents in the owner's workspace.
Agents can be grouped into teams, where a lead routes each request to the right specialist. They run on Claude, GPT, and Gemini models, and connect to Gmail and Google Calendar to take real action. The orchestration is built on Mastra.
Agents can be grouped into teams, where a lead routes each request to the right specialist. They run on Claude, GPT, and Gemini models, and connect to Gmail and Google Calendar to take real action. The orchestration is built on Mastra.
Meet the team that would work for you
Reception Agent
The first live agent is a voice receptionist built on Retell. It answers inbound calls, collects the caller's details, checks real availability in Jobber, and books the job while the caller is still on the line.
Every call is transcribed and logged in the dashboard with its outcome, and owners can invite their team to the workspace.
Every call is transcribed and logged in the dashboard with its outcome, and owners can invite their team to the workspace.
Reception Agent booking a job (product preview)
Site Builder
The newest piece rebuilds an owner's website from their Google profile and current site. It keeps their services, story, and details, then writes a page for every service, set up for local search.
Each site goes live on its own address with a contact form that emails requests to the owner, an inline editor for changing any word or image on the page, and an Ask AI button that answers visitors' questions from the business's own details.
Each site goes live on its own address with a contact form that emails requests to the owner, an inline editor for changing any word or image on the page, and an Ask AI button that answers visitors' questions from the business's own details.