Inbound lead agent
Reads every enquiry, researches the company, scores the fit against your best customers, prices the job from your rate card and drafts the reply for a person to approve.
Works with Gmail, Outlook, HubSpot, Slack
Custom AI agents that do the work, with a person in charge
An AI agent reads what comes in, decides the next step, uses your tools to do it and reports back. We build agents for specific jobs in your business, connect them to the systems you already use and keep them running.
Agents we build
Each agent does one job well. We connect them when a job crosses teams.
Support agent
Answers customers from your help docs and order data on your site, WhatsApp or email, and hands the conversation to a person when it should.
Works with Shopify, Zendesk, Intercom, WhatsApp Business
Document agent
Reads invoices, contracts, delivery notes and forms, checks them against your rules and files them in the right system. Anything unclear goes to a person.
Works with Gmail, Google Drive, SharePoint, Xero
Research agent
Finds companies that look like your best customers, researches each one and writes a first line that proves it, ready for your outbound.
Works with Apollo, Clay, Instantly, Smartlead
Testing agent
Reads your analytics and payment data, suggests split tests, ships the variants and keeps the winners. Outtest, our own product, runs on agents like this.
Works with PostHog, GA4, Stripe
Reporting agent
Pulls the numbers from your tools every Monday at 8am, writes the summary and posts it in Slack before your first meeting.
Works with Stripe, Xero, HubSpot, GA4, Slack
An agent at work
Seven minutes from email to quote. The person on the sales team spent about 20 seconds on it.
- Email in from Sarah Hale at Northfield Logistics asking for a quote for 40 pallets a month to Leeds.
- Looked up Northfield Logistics. 42 staff, uses HubSpot, hiring two sales roles.
- Compared them with your best customers. Fit score 86 out of 100.
- Created a deal in HubSpot and attached the email.
- Priced the job from your rate card.
- Drafted a reply with the quote and three call slots from your calendar.
- Sent to #sales in Slack for approval.
- Tom approved. Reply sent. A follow-up goes out on Thursday if Sarah doesn't answer.
Example only. The names and numbers are made up.
How we keep agents safe to run
- A person approves anything that sends money, emails a customer or deletes data, until you decide to let it run on its own.
- Each agent gets the narrowest access it needs, read-only wherever possible.
- Every action goes into a log your team can read.
- Before launch we run the agent on your real past data and compare its work with what your team did.
- We build on the business tiers of AI providers such as Anthropic and OpenAI, whose terms don't allow training their models on your content.
How a build works
A free call, a blueprint that comes off the build price, weekly demos and a fixed quote.
- 01
Scope call
30 minutes, free
We pick the one job that costs your team the most hours and walk through how it runs today, step by step.
- 02
Blueprint
About a week, credited to the build
A written plan covering what we'll build, what it connects to, what it costs to run each month and a fixed price. It includes a clickable prototype.
- 03
Build
2 to 6 weeks
We build in your accounts and demo working software every week. Your team tests it on real work before it goes live.
- 04
Care
Monthly, optional
We monitor it, fix what breaks, move to better models as they ship and keep adding to it. Cancel any month and keep everything.
Questions about this service
Something else? Ask it on the scope call.
What is the difference between an AI agent and an automation?
An automation follows fixed steps you set in advance. An agent reads the situation and picks the next step itself, so it can handle messy inputs like free-text emails, PDFs and customer questions. Most of our systems use both. Fixed steps run as plain code, and the agent handles the parts that need judgement.
Which AI models do your agents use?
We pick the model for each job, usually from Anthropic or OpenAI, and send routine steps to smaller, cheaper models to keep running costs down.
Where do the agents run?
On hosting in your name, connected to your tools with your own API keys. You own the code, the prompts and the data, and the agent keeps running if we stop working together.
How long does it take to build an AI agent?
An agent for a single job usually goes live about two weeks after the blueprint. Agents that work across several systems take three to six weeks.
What if the agent gets something wrong?
We design for it. Anything that sends money, emails a customer or deletes data waits for a person to approve it until you're happy to let it run. The agent logs every action, and we test it on your real past data before launch.
Other services
Most clients start with one system, then add the next once the first is running.
AI sales systems
A CRM built around how you sell, cold email that researches every lead, call notes that file themselves and proposals in minutes.
Read moreAI marketing systems
X and LinkedIn posts in your voice, on-brand ad creative, split tests that run themselves and pages that rank in Google and ChatGPT.
Read moreAI for operations
Internal tools that replace the spreadsheet, documents that read and file themselves, onboarding from one form and an AI assistant for the whole team.
Read moreAI customer support
A support agent that answers from your docs and order data, inbox triage with drafted replies and one help centre for your agent and your team.
Read moreTell us which job you want an agent to take.
We reply within 24 hours with how we'd build it.