Service · AI automation
AI automation for real estate.
The routine work, handled inside limits you set, with a record of everything it did.
- Typical timeline
- 3–5 weeks to first workflow
- You receive
- 13 deliverables
- Built for
- Agents, landlords, managers
- Safety
- Limits set before it runs
What is included.
13 deliverables
Every engagement includes all of it. Nothing here is an upsell.
Controls, built first
- Spending caps
- Approval steps per channel
- After-hours rules
- Stop switch
- Full activity logWhat it sent, and when
What it handles
- Lead response and follow-up
- Client and tenant messaging
- Document and data extraction
- Email and SMS workflows
- Calendar and viewing booking
Handover
- Accounts in your name
- Written runbook
- Training session
An agent that can act on your behalf without a spending limit is a liability, not a saving. We build the guardrails before the automation.
How it goes.
4 steps
Watch the actual week
The shared inbox and the group chat, not a workshop about them.
List of what repeats
Draw the lines
What it may spend, what it must ask about, what stays yours.
Written limits
One workflow live
The most repetitive task only, with you approving everything.
Working automation
Widen slowly
Approvals come off one at a time, where the log has earned it.
Activity review
Whether this is for you.
Working with us
An off-the-shelf tool
Built around how you already work
You change to fit the software
Limits enforced in the system
Limits described in a settings page
Every action logged and reviewable
Partial history, if any
Your data is never used for training
Check the terms carefully
We turn things off if they misfire
A support ticket queue
Where we have done this.
Questions.
If yours is not here, ask it on a call. We would rather answer it than have you guess.
Safety and control
What stops it doing something stupid?
Limits you set before it goes live: what it may spend, what needs your approval, when it may act at all, and a switch that stops everything. It stops at those lines rather than guessing, and every action is written to a log you can read. In practice most people start with approval required on everything and relax it once the log has earned their trust.
Will it message our clients without us knowing?
Only if you decide it should, and you decide that separately for each channel. You can require your approval on everything, on some things, or nothing at all. Whatever you choose, everything it sends appears in the activity log with a timestamp, so "what went out last night" is a question with an answer.
Do you use our data to train AI models?
No. Your data is processed to run your automation and for nothing else. It is not used to train models, ours or anyone else's.
What happens when it gets something wrong?
It will, occasionally — anything that drafts language sometimes drafts the wrong language. That is why the approval steps exist for anything consequential, why spending has a ceiling, and why there is a log. When something misfires we turn that workflow off, fix it, and turn it back on. We would rather it did less unattended than more.
Scope and fit
Is this just ChatGPT with extra steps?
No. A chat window waits to be asked; this acts on events in your business — an enquiry arriving, an application submitted, a document landing — and it runs inside limits you set. The useful part is not the language model, it is the plumbing around it and the controls on top of it.
We already use Doughy. Do we need this?
Probably not for the landlord operations Doughy already covers — rent, maintenance, leads and books are in the product. This service is for the work that is specific to your business and sits outside it. If what you describe is already a Doughy feature, we will tell you so rather than quote for it.
Can you automate cold outreach at volume?
No. We will not build a spam machine. Following up with people who have actually enquired is a different thing, and that we will happily build.
Tell us what is eating your week.
The other three