Applications & systems

AI chatbots & workflow automation

Assistants and automations that draft, route and summarise — with people approving anything sensitive.

Illustrative demonstrationAn enquiry, handled with a person in charge

01 Inquiry

“Hi — do you install kitchen backsplashes, and could someone visit next week? Budget is flexible.”
Sam · website chat · sample enquiry

02 Intake

Topic: Kitchen backsplash installation
Request: Site visit next week
Budget: Not stated — marked for a person to ask
Contact: Sam (sample)

Details are organised from the message; anything uncertain is flagged rather than guessed.

03 Routing

Routed to Estimates — rule: installation request with a site visit.

Estimates
Support
Billing

04 Draft

AI draft — not sent

Hi Sam, thanks for getting in touch. Yes, we install kitchen backsplashes. We can arrange a visit next week to measure and talk through materials — would Tuesday or Thursday morning suit you?

05 Human approval

A person reviews the draft before anything leaves the business. Sensitive decisions — prices, refunds and commitments — always stay with people.

Returned to draft with your note: “Mention the tile samples.”

Approved by a person · draft locked

06 Complete

Simulated: the approved reply would be sent and the enquiry logged with its review.

Usage is logged for cost visibility. In this demonstration nothing was sent and no model was called.

A synthetic enquiry and a pre-written demonstration draft. No AI model is called and nothing is sent.

We build AI-assisted workflows such as website assistants, enquiry triage and document drafting. Sensitive decisions — refunds, contracts, pricing and releases — always stay with people.

How it is priced

Quoted separately as custom development after a scope and feasibility review.

Custom projects are billed 50% as a deposit, 25% after the approved design milestone and 25% before production launch.

What is included

Website or internal assistants grounded in your approved content
Enquiry triage and routing
Draft generation with human review
Usage logging and cost visibility

Good to know

AI output can be wrong; workflows are designed with human review where it matters
Model provider usage fees are passed through or billed by the provider