Our process
How it works, start to finish.
What happens after the first call, what an agent can and can’t touch, what it connects to, and what you see every month. Most first agents are live in three to four weeks.
Building an agent
Six steps. You can stop at any of them.
01 —
Intro call
We walk you through how the studio works, show you what we've built, and explain the pricing. Your chance to ask everything before committing to anything.
We'll also get a sense of which processes in your business are good candidates for automation — invoice chasing, data entry, customer follow‑ups, reporting. No pressure, no pitch — a conversation about what's possible.
02 —
Discovery
Time to map the process. You bring something you want to automate and we break it down together — what happens first, what decisions get made, what systems are involved, what counts as done.
We get into the weeds: triggers, exceptions, the cases that need human judgment. We'll usually sketch two or three approaches and pick the pilot — something high‑frequency your team does the same way every time.
03 —
Proposal
We design your agent and show you exactly what it will do: a written proposal and a demo video of the agent handling real‑world scenarios with sample data.
This is the test drive. If something should work differently, we adjust the design until you're confident it fits your process.
04 —
Build & test
Once you approve, we build the production version and connect it to your systems, then put it through extensive testing — scenarios, edge cases, exception handling.
There's no charge for any of this. We build and test at our cost.
05 —
Launch
Your agent goes live and starts handling real work — automatically, consistently, around the clock.
You put a balance on account for the processing it consumes, and from then on we're paid per job it completes — nothing else. If it isn't earning its keep, switch it off and owe nothing.
06 —
Expand
We monitor continuously — if it fails, we fix it, and support is included in the fee.
As your agent works you'll spot the next process. Most clients are back within 60 days: “can we do this too?” Almost always, yes. Invoice chasing becomes data entry becomes follow‑ups becomes reporting — each compounds on the last.
Timeline
Three to four weeks, first call to live.
Day 0
Intro call
Thirty minutes. We find the one process worth doing first.
Week 1
Discovery
We map the process properly — triggers, decisions, exceptions.
Week 2
Proposal & demo
You watch a demo of your agent working on sample data before we build.
Weeks 3–4
Build, test, live
We build and test at our cost, then it goes live in your environment.
Websites run two to three weeks from brief to launch. Films are days — usually inside a week for a brand film and cut‑downs.
Integrations
It works with what you already use.
Finance
Xero, Sage, QuickBooks, FreeAgent, Stripe, GoCardless
Email & calendar
Microsoft 365 and Outlook, Google Workspace and Gmail
CRM & sales
HubSpot, Salesforce, Pipedrive, Zoho, Monday
Support
Zendesk, Freshdesk, Intercom, shared inboxes
Commerce & ops
Shopify, WooCommerce, Airtable, Notion, Slack, Teams, WhatsApp Business
Running something older or bespoke? If it has an API, an export, or an inbox, we can usually work with it — and we’ll tell you honestly on the first call if we can’t.
Access & safety
What an agent can and can’t touch.
It runs on your accounts
Access is granted by you, scoped to the specific inbox, folder or records the job needs — and revocable by you at any moment, without asking us.
Least access that works
An invoice agent sees invoices. It does not get your whole finance system because that was easier to set up.
It stops rather than guesses
Anything outside what it was built for goes to a human. Agents escalate; they don’t improvise.
Hard spending limits
A ceiling per conversation, a daily cap, and a balance it cannot exceed. Runaway costs are structurally impossible.
Your data isn’t training anything
We don’t train models on your business, and we don’t reuse your data for anyone else. We keep only what the task needs to run.
Everything is logged
Every action an agent takes is recorded and visible to you — not a black box you’re asked to trust.
Reporting
What you see every month.
A live balance
What’s left on account and what the agent has consumed, updated as it runs.
A running count of outcomes
Every invoice recovered, ticket resolved or meeting booked — itemised, timestamped, and exportable.
A monthly statement
What the agent did, what it cost to run, and what we’re charging for. The count is the invoice.
Checkable against your own systems
Every outcome we bill for exists in your CRM, inbox or ledger. If our number and yours disagree, yours wins.
Websites & films
Shorter, and fixed price.
01
Brief
One call. What it’s for, who it’s talking to, what you already have.
02
Quote
Fixed price and a date, in writing, before anything starts.
03
Build
You see it in progress, not at the end. Two rounds of revisions included.
04
Launch
It goes live and we invoice. Everything — files, accounts, footage — is yours.
The awkward questions
What if you get hit by a bus?
Your agents run on your own accounts and your own infrastructure, not on a Ruxley platform you’d lose access to. Every build ships with its documentation, and if we ever stopped, the agents keep running and another developer can pick them up. You are never holding a system only we can operate.
What if the agent does something wrong?
Agents are built to stop rather than guess, and they escalate anything outside their brief. If one misbehaves it’s ours to fix, at our cost, and support is included — we don’t bill you to repair our own build.
Do I need anyone technical?
No. We handle the integration, the testing and the monitoring. You need someone who understands the process being automated — that’s all.
What if we want to stop?
Switch it off. No notice, no exit fee, and whatever’s left on your balance comes back to you.