AI sales engineer for technical products
Help customers choose the right technical products.Your team approves the basket.
Sentara AI builds an assistant for your website. It asks customers the questions your best salesperson would ask, picks parts only from your catalogue, and follows your technical rules. If a request breaks a rule, it says no. Your team checks every basket before the customer sees it.
If your customers ask "which one do I need?" and the honest answer starts with "it depends", this is for you.
Live at flue-pipes.com
Sentara AI runs live on flue-pipes.com, where customers configure a complete flue system in one conversation and the shop's staff approve the basket before it reaches them.
One live shop so far, in UK English, with a small number of real customer conversations. A German or US version is not live yet.
See it live on flue-pipes.comOpens in a new tab. Accept the shop's cookie notice to see the assistant.
Regulations and manufacturer instructions encoded as rules, not left to the model to remember
Every configuration evaluated against real customer cases before launch
Runs on the shop's own pages and hands the basket to the shop's own checkout
For products where the right choice depends on the details.
This is not a general chatbot that has read your website. Your products and your technical rules are built into the system, and it says no when a request breaks a rule.
The customer is guided, one question at a time
Mostly by tapping a choice or typing a measurement, with pictures and a short helper on every question. They can also attach a photo or sketch, and the assistant asks them to confirm what it saw. It asks only the questions that change the answer, worded the same way every time.
Your rules decide, not the AI
Regulations, manufacturer instructions and your own policies are written down and checked by the system, not left to the AI to remember. If a request breaks a rule, the assistant says no and explains which one. The AI never types a product code itself. Parts come only from your active catalogue, in combinations that fit together.
Your team approves the basket
The assistant builds the full basket, every part and quantity, with the whole conversation attached. Your staff can change quantities, add items, ask the customer a question, or reject it with a reason. Only after approval does the customer get a link that loads the basket into your shop's cart. They pay at your checkout. The assistant never takes payment.
It hands off when it should
If a customer asks for a person or seems frustrated, your staff take over the chat. If it cannot build a basket, a person quotes it by hand and the customer still gets the normal confirmation.
A real test conversation on flue-pipes.com, shortened
Are you fitting into an existing brick chimney, or do you need a new flue system?
New flue system
Where will the new flue run?
Internal
Great, let's design the flue that runs inside the house and out through the roof. What flue size does your stove need?
5 inch (125mm)
Our team will review your specification and send your basket link to name@example.com by email. We'll also post it here in the chat if you're still on the website, so you won't miss it either way.
Basket, first parts shown
- Roof support
- Straight length 1000mm with locking band
- Single to twin wall adapter
Your basket is being reviewed and will be with you shortly.
Your team stays in control.
The assistant comes with an admin dashboard. It is where your staff approve baskets, step into a chat and see how the assistant is doing.
Live today for flue-pipes.com.
Basket approval
The customer's notes and the basket side by side. Staff change quantities, add items, ask the customer a question, or approve or reject with a reason.
Live chats
Active conversations at a glance, with a flag when a customer needs a person. Staff can take over, use saved replies, and hand the chat back to the AI.
History, analytics and alerts
Conversation history with search and export, analytics on baskets and approval times, and alerts by browser, phone and email.
Sample data. The customers, conversations and analytics figures in this video are made up for the demo.
Built for businesses that sell complex technical products.
Online shops, trade counters, distributors, and manufacturers with dealer networks.
Proven in flue and chimney, where it is live today. The same approach suits other catalogues where the right product depends on rules:
Flue, stoves and heating
Live today
HVAC and ventilation
Same approach
Plumbing and electrical supplies
Same approach
Building materials and roofing
Same approach
Industrial parts and machinery
Same approach
Lab and measurement equipment
Same approach
Who builds it is who you talk to.
Sentara AI is a German company. Whoever designs your system also writes down your rules with you, runs the evaluations and answers the phone. You will not be handed to an account manager.
Design, build, evaluation and operation stay in one place. Nothing is subcontracted.
A small first step, then a pilot, then production.
Typical starting prices are below, so you can judge fit before we talk. Your own price depends on your catalogue, and you get it in writing before anything starts.
Fit call
Free, 30 to 45 minutes
We look at your catalogue and how a quote happens today. If it is not a fit, we say so.
Feasibility and design
Where the rules are complex
A short paid piece of work to map the rules and the data before anyone commits to a build. Priced after the fit call.
Pilot
Typically from around 5,000 EUR
One product family or one use case, live with real customers, evaluated against real cases.
Production
Typically from around 10,000 EUR
Integration with your shop or quoting system, all product families, evaluation before launch.
Operations
From around 1,000 EUR a month
Monitoring, catalogue and rule updates, regular evaluations, and a person who knows your system.
Questions people ask before a fit call
Short answers. The long ones are what the call is for.
Live today on Magento. Other shops and quoting tools are integrated per project, and the feasibility step establishes what that takes.
Yes. The chat and the dashboard are styled to match your company when we set them up: colours, logo, name and wording. That is part of the build for each client, not a settings screen you configure yourself.
It reads a synced copy of your active products, refreshed every few minutes, so a discontinued product is not offered. Prices come from your shop at checkout. New product families need their rules added, which is part of the monthly operations work.
If a customer asks for a person, or seems frustrated, your staff take over the chat with the whole conversation attached. If it cannot build a basket, a person quotes it by hand and the customer still gets the normal confirmation. It is built to refuse to guess: a wrong part is a safety problem or a returns problem, never an acceptable answer.
It declines a request that breaks a regulation and explains which one, for example a flue offset that is too large for its height. It also refuses unsafe situations, such as a stove in an unpartitioned garage with a petrol vehicle, or any work through an asbestos roof, and tells the customer what has to happen first.
The rules come from regulations, manufacturer instructions and your own policies, agreed with you before launch and tested against real cases. Responsibilities are set out in the agreement, and the system hands off to a person rather than guess.
It is live on one shop, flue-pipes.com, in UK English, with a small number of real customer conversations so far. A German or US version is not live yet.
The database and back end are in the EU. The AI model (Anthropic) and the front-end hosting (Netlify) are in the United States, under Standard Contractual Clauses. The full list is on the subprocessors page. Sentara AI is the data processor for your customers' conversations and signs a data processing agreement with you.
A pilot on one product family is typically live within a few weeks of the rules being agreed. A full catalogue depends on how many product families and rules there are, which the feasibility step establishes.
No. You need someone who knows the products and the rules, usually your best salesperson or technical buyer, for a few hours during setup. We handle the technical side.
We read every message and reply within two working days. If it looks like a fit we book a 30 to 45 minute call. If it does not, we will say so and, where we can, point you somewhere useful.
Could this work for your products?
Tell us what you sell and where customers need help choosing. We reply within two working days.
Request a free fit call