Pre-qualify enquiries, provide knowledge, relieve people — with clear limits.
An AI assistant takes on clearly delimited, recurring steps in sales and customer contact: initial qualification, structured information intake, answers from verified materials. The decision stays with a human.
A process with rules, not a salesperson without supervision.
An assistant is useful when its task area is narrow, its knowledge base defined and its handover to a person regulated. That is exactly how it is built here: narrow the use case, define the sources, check the answers, document the limits. Whatever lies outside the assignment, the assistant passes on — it does not improvise.
What an assistant is good for
- Initial qualification of enquiries: does the enquiry match offer, region and minimum quantity?
- Structured information intake: asking the right questions in the right order
- Answers to frequent questions — exclusively from verified, approved materials
- Internal knowledge assistant for team and sales: prices, terms, product data, processes
- Preparation of replies that a person checks and sends
- Handover of qualified enquiries to the responsible person with all captured details
- Support for lead routing: which enquiry goes to whom?
- Documentation of recurring processes so they don't live only in one person's head
What applies without exception
- The human stays in control: decisions, commitments and prices are approved by a person.
- The knowledge sources are defined before launch — the assistant answers only from them.
- Answers are checked in the test phase and by sampling during operation.
- Personal data is limited to what is necessary; only what the process needs is captured.
- Providers, models and data flows are reviewed and documented before launch.
- AI output is not legal, tax or professional advice — and is not presented as such.
- Where labelling of AI content is required or sensible, it is implemented visibly.
How an assistant is built
- Narrow the use case: one task, one channel, one measurable goal.
- Define and approve the knowledge sources: materials, prices, rules, exceptions.
- Build a prototype with test questions — including the cases the assistant should not answer.
- Review and approval by you: answers, tone, limits, data scope.
- Define the handover to humans: when, to whom, with which information?
- Operation and maintenance: update sources, check samples, handle edge cases.
Scope and price
Scope and price are set after a needs analysis — depending on use case, knowledge sources, channels and the connection to your systems. An assistant can be built as a module within an Asset Sprint or a Sales Foundation, or as a project of its own.
Which enquiries cost you the most time today?
In an initial conversation we clarify which step is suited to an assistant, which sources it needs and where the handover to humans lies.
No obligation · clear use case first · no automation without review