Enquiries, quotes and triage
Emails, forms and chat that need reading, qualifying and answering. An agent drafts, a person approves, the CRM is updated without anyone retyping.
One workflow, automated end to end.
Not a chatbot bolted onto your website. We pick the one process that eats the most hours, wire it to your real systems on the latest Claude models, and keep a person in the loop where a wrong answer would cost you.
Where it pays off
The work nobody should be doing by hand.
Extraction and checking
Invoices, CVs, contracts, reports. Pull the fields out, check them against your rules, flag the exceptions for a human.
Answers from your own material
Retrieval over your docs, tickets and product data so staff and customers get grounded answers, with sources, not guesses.
Selling and quoting
Conversational agents that recommend, configure and quote inside the tools you already use. We built one for an enterprise CPQ stack and one for Shopify stores.
Glue between systems
The spreadsheet someone updates every Monday from three other systems. Replaced by an integration that runs itself and tells you when it can’t.
When we say no
If the workflow has no clear owner, no data, or the cost of a wrong answer is unbounded, we will tell you on the scoping call rather than build something you can’t trust.
How a project runs
Audit first, then build, then hand over.
Audit
FirstWe sit with the people doing the work, map the workflow step by step, and find where the data actually lives.
- A written map of the process and its exceptions
- What to automate, what to leave human, and why
- A fixed quote for the build
Short, paid or free depending on scopeDecided on the scoping call
Build
Weekly releasesAgents, retrieval and integrations built on the latest Claude models, in your accounts, with evaluation sets so you can see accuracy, not just demos.
- Human approval steps wherever a wrong answer costs money
- Logging and monitoring from the first release
- Measured against the hours it removes
Fixed priceA working release every week
Hand over
IncludedEverything runs in your cloud under your API keys. You get the code, the prompts, the eval sets, written docs and a recorded walkthrough.
- No retainer, no lock-in
- Your team can change prompts and rules without us
- Further automations as new fixed-price phases
Fully yoursNothing to migrate later
Proof
We run this in production ourselves.
Aurevia is our own AI sales agent for Shopify stores: it reads the catalogue, talks to shoppers, recommends products and hands off to a human when it should. It is live on the Shopify App Store and has shipped a release every week for over a year.
For Expedite Commerce we integrated a conversational quoting agent into a live enterprise CPQ stack without breaking what was already there.
Read the Aurevia case study ↗
Questions
Before you reach out.
Which models do you use?
The latest Claude models by default, running under your own API keys so usage and cost are yours to see. We will use another model where it is clearly the better tool.
What about our data?
Everything runs in your cloud accounts. Nothing is trained on your data, and retrieval indexes live with you. We put this in writing before starting.
How do we know it is accurate?
Every build ships with an evaluation set drawn from your real cases, so you see a measured accuracy number before go-live and after every change.
Do you offer retainers or ongoing support?
No. Fixed price, fully handed over. When you want the next workflow automated, it is a new fixed-price phase, quoted the same way.
Let’s talk
Which workflow eats the most hours?
Tell us in a sentence and we will say on the call whether it is worth automating, and what it would cost.
Book a free scoping call
Or send a short brief and hear back within one business day.