Dream Awake LabsGet in touch

AI agents & business automation

AI agents for everyday business work.

Reading inquiries, updating the CRM, and chasing follow-ups take time. An agent can help with that work, but the team still needs to see its progress and review actions such as contacting a customer or requesting payment.

For owners and operations leads whose team handles inquiries, bookings, quotes, and follow-ups by hand, and who want to see what an agent is doing before trusting it.

Design an agent workflow
Agent City in the Rectangle scenario while "Send deposit invoice" waits for approval: an isometric city of colored district buildings, a toast reading "Waiting for approval: Send deposit invoice", the Human Approval Hall showing one item awaiting decision, and a work queue listing a DJ rehearsal inquiry as Awaiting Approval.
The deposit invoice cannot go out until a person approves it. Synthetic data.

Worked example

Inquiry handling at The Rectangle

Every inquiry The Rectangle’s website receives used to be read, sorted, entered into the CRM, and answered with a deposit request by hand. We built the system behind the site: the inquiry API, AI screening, routing, CRM sync, notifications, the owner admin, the deposit flow, and the tools browser agents call.

An inquiry arrives from a form or from a browser agent, is validated and saved before anything else, and is screened for spam, intent, and fit by a model that fails open, so an error never drops an inquiry. Studio bookings and business projects go to separate pipelines, and the owner gets an alert.

Only a person’s approval sends the deposit invoice. Agent City shows the same flow as a city: the deposit invoice waits in the approval hall until someone decides, then the agent resumes.

Pipeline diagram in eleven numbered steps: inquiry from a site form or browser agent; validation with schema check, rate limit, and honeypot; saved to Postgres first; AI screening that fails open; routing to a studio or business pipeline; HubSpot sync; owner alert; a highlighted step where a person approves; Shopify deposit invoice; payment webhook; and a confirmation email with a calendar file.
Every inquiry is saved before any external call, and nothing is invoiced until a person approves.

Read the case study: The Rectangle

How it works

  • Agent Forge

    defines each action once with an effect class and a consequential flag, so an agent knows which calls need a person.

  • Atlas

    is our own operations agent system and the pattern we build from: a policy check on every tool call, human approval in Slack before anything goes outside, per-agent budgets, and a tamper-evident audit log.

  • Agent City

    shows what each agent is doing, what it waits on, and which steps are held for approval.

What we can build

  • Custom agents

    Agents built for one job in your business, with clear inputs, tools, and limits.

  • Business workflows

    Multi-step flows from intake to follow-up across CRM, email, payments, and calendar, with retries and a record of every step.

  • Policy and approvals

    Every action is classified by its effect, and consequential ones wait for a person in the tools your team already uses.

  • AI operating systems

    Several agents with shared memory, budgets, and an audit trail, run as one system for your company.

  • Operational visualization

    A view of what each agent is doing, what it waits on, and what it may do, readable without training.

  • Organizational knowledge

    Meeting notes and chat threads turned into decisions, tasks, and summaries your team can search.

Projects

Projects in this area.

  • Agent City in the Rectangle scenario while "Send deposit invoice" waits for approval: an isometric city of colored district buildings, a toast reading "Waiting for approval: Send deposit invoice", the Human Approval Hall showing one item awaiting decision, and a work queue listing a DJ rehearsal inquiry as Awaiting Approval.

    Systems & Research

    Agent City

    A live 3D city that shows a business owner what their AI agents are doing and holds every consequential step for a person to approve.

  • Five visitors face a wall-sized projection of Agent City inside The Rectangle, with districts labeled Plan, Build, Test, Create, Learn, and Deploy, while a white robot with a screen reading "Ask me anything" points at the map.

    Systems & Research

    The Rectangle

    A working creative space in Santa Cruz for DJ rehearsals, media production, and XR experiments, supported by an agent-ready booking website.

  • Systems & Research

    Agent Forge

    An open-source TypeScript library that lets a web app expose its real operations to AI agents as typed tools, each with a declared effect and a flag for consequential actions.

All projects

Where to start

Workflow pilot

Who it is for
An owner or operations lead whose team handles inquiries, bookings, quotes, or follow-ups by hand.
The problem
Routine work depends on one person reading, sorting, entering, and replying, and nobody can see where each item is.
What you get
An agent handling one workflow from intake to follow-up, with approval before actions such as customer messages or payment requests. We compare the results with your current process.
What you bring
  • The workflow as it runs now, in your own words
  • Access to the tools it touches: the CRM, the inbox, the calendar, the payment system
  • One person to review and approve actions during the pilot
  • A sample of recent cases to test against
What you learn
How much of the workflow an agent can carry, where approval belongs, how long each case takes, and what it costs to run.
Evidence

Next step

Design an agent workflow

Get in touch

Tell us what you have in mind.

Describe it in a few sentences. You get a draft brief to review before anything is sent.

Design an agent workflow Email hello@dreamawakelabs.com

We use cookieless analytics and first-party cookies that remember how you found us. Nothing is sold or shared with ad networks.