AI Assistants
Assistants that
know your business.
Customer service, shopping and internal knowledge assistants grounded in your own information, connected to your systems and handed to a person when it matters.
The work
AI Assistants
A useful assistant is not a chat window on top of a general model. It answers from approved sources in the languages your customers use, knows what it may and may not do, says when it does not know, and passes the conversation to your team with context. We design, build, evaluate and operate assistants on those terms.
Customer service assistants
Answer questions about orders, delivery, returns and products from your policies and systems, with order lookups under strict access rules and handover to a human agent.
Shopping & product assistants
Help visitors find, compare and choose products using live catalog data, stock and prices, so recommendations reflect what you can actually sell.
Internal knowledge assistants
Give employees answers from manuals, procedures, contracts and tickets with source references, respecting the permissions your systems already enforce.
Sales & quoting assistants
Qualify inbound requests, gather requirements, draft quotes or proposals for review and record a structured summary in your CRM.
What you receive
A production assistant with a documented knowledge scope, an evaluation test set, access controls, a clear AI disclosure, human handover, conversation analytics, cost controls and an operating guide, with an optional monthly improvement cycle run by HUB.
How we deliver
Built, measured
and operated
Products, policies and customer questions change after launch. We treat an assistant as a product with owners, tests and controlled releases.
- 01 /
Scope
Users, questions, allowed actions.
- 02 /
Ground
Sources, permissions, integrations.
- 03 /
Build
Conversation design, tools, handover.
- 04 /
Evaluate
Test sets, misuse testing, human review.
- 05 /
Operate
Monitor, improve, update knowledge.
A practical starting point
Start with the
questions you get.
Send us a sample of real customer or employee questions, the sources that answer them and the systems involved. We will tell you what an assistant can reliably handle and what should stay with your team.
Will the assistant make things up?
Any model can produce a wrong answer. We reduce the risk by grounding answers in approved sources, limiting scope, testing against real questions, showing references and routing uncertain or sensitive cases to a person. We measure the error rate rather than assume it is zero.
Do customers need to know they are talking to AI?
In the EU, the AI Act’s transparency rules apply from 2 August 2026 and generally require that people are told when they interact with an AI system, unless that is obvious. We build a clear disclosure in by default. This is not legal advice.
What happens to customer data?
We agree the data flows before building: which systems the assistant reads, what is logged and for how long, which AI providers process data and on what terms, and who can access transcripts.
Where can the assistant run?
Usually on your website or web application and in internal tools your team already uses. Other messaging channels depend on their platform APIs and terms, and we assess them per project.
Which AI models do you use?
We choose per use case for answer quality, speed, cost, data processing terms and language support, and keep the design portable so the model can be replaced later.
Explore our capabilities
Back-office automation and agents ↗ Can AI automate this? ↗ Agentic Commerce Lab ↗ All engineering services ↗What does your software
need to do next?
A new build, a difficult codebase or a system that needs support. Let’s talk.