AI Customer Assistant
A support agent that knows your product and can take action.
Request DemoA large share of support volume is the same small set of questions: where is my order, how do I reset this, what does this charge cover. Each one is quick to answer and none of them needs a specialist, yet together they consume the hours that should go to the difficult cases.
AI Customer Assistant answers those questions from your own documentation, help centre and past tickets, so the replies match what your product actually does rather than a generic script. Where a question requires looking something up — an order status, a subscription state — it can perform that lookup and answer with the real value.
Its most important behaviour is knowing when to stop. When a question falls outside what it can answer, or when a customer is clearly frustrated, it hands over to a human with the conversation history attached rather than looping. Deflection that comes at the cost of a bad experience is not a saving.
The problem
Support teams spend most of their time answering the same repetitive questions. The volume is predictable and the answers are already written down somewhere, but each one still costs a person a few minutes. Response times drift, the genuinely complex cases wait behind the simple ones, and hiring to absorb the volume is expensive and slow.
The solution
AI Customer Assistant answers common questions from your knowledge base, escalates complex issues to a human, and can perform simple actions like status lookups. It works from your own content, cites what it drew on, and hands over cleanly when a question needs judgement rather than information.
How it works
- 1
Point the assistant at your existing content — help centre, product documentation, policy pages and, where useful, resolved support tickets.
- 2
It indexes that material and answers from it rather than from general knowledge, which keeps replies aligned with how your product actually behaves.
- 3
Connect the read-only lookups you want it to perform, such as order status or subscription state, so it can answer with real values instead of generic instructions.
- 4
Set your escalation rules: which topics always go to a human, and what signals — repeated rephrasing, explicit frustration, an explicit request — trigger a handover.
- 5
Deploy it where your customers already are, on your site, in your help centre widget or through your existing support tooling.
- 6
Review the questions it could not answer. That list is the most direct indication of what your documentation is missing.
Features
- Answers grounded in your own content
- Source citation on every answer
- Read-only action lookups
- Rule-based human escalation
- Multilingual conversations
- Conversation history passed on handover
- Unanswered-question reporting
- Configurable tone and response length
Benefits
- — Repetitive questions resolved without a queue
- — Support time redirected to complex cases
- — Consistent answers regardless of who is on shift
- — Coverage outside business hours
- — Documentation gaps made visible
- — Volume absorbed without proportional hiring
Where it fits
Typical situations this application is built for.
Order and delivery status
The most common question in retail support is answered directly, with the actual status retrieved from the order system, rather than with an instruction to log in and check.
First-line product support
Configuration and how-to questions are answered from the product documentation, with a link to the page the answer came from so the customer can read further if they want to.
Out-of-hours coverage
Questions arriving overnight receive an immediate answer where the documentation covers them, and are queued with full context for the morning where they do not.
Finding the gaps in your help centre
The questions the assistant could not answer are collected into a report. Rather than guessing what to document next, the support team writes the pages customers are actually asking for.
Integrations
Security
- — Answers restricted to your approved content
- — Read-only access to connected systems
- — Escalation rules enforced before any action
- — Conversation logging and review
- — Configurable retention of conversation data
- — Scoped credentials per connected system
Pricing
Starter
Contact us
- — Knowledge base answering
- — Website widget
- — Human escalation rules
- — Unanswered-question reporting
- — Email support
Growth
Contact us
- — Action lookups on connected systems
- — Ticketing tool integration
- — Multilingual conversations
- — Custom escalation logic
- — Priority support
FAQ
What stops it from inventing an answer?
It answers from your indexed content rather than from general knowledge, and cites what it drew on. When the content does not cover a question, it says so and escalates instead of producing a plausible-sounding guess.
When does a conversation reach a human?
On the rules you set. Typically that means topics you have marked as always-escalate, an explicit request for a person, repeated rephrasing of the same question, or signals of frustration. The conversation history goes with it, so the customer does not repeat themselves.
Can it change anything in our systems?
By default it performs read-only lookups. Any action that writes — cancelling an order, issuing a refund — is something to scope deliberately, with approval controls, rather than something enabled out of the box.
What if our documentation is incomplete?
Then coverage starts lower and improves as you fill the gaps. The report of unanswered questions is designed for exactly this: it tells you which pages are worth writing first.
Does it work in more than one language?
Yes. It can hold a conversation in the language of the customer while drawing on documentation written in another, which avoids maintaining a translated knowledge base for every market.
Will customers know they are talking to an AI?
That is your decision to configure, but we recommend disclosing it. It sets the right expectation and makes the handover to a person feel like an escalation rather than a discovery.
How is it kept up to date when our product changes?
It re-indexes your content as it changes, so an updated help page is reflected in its answers. Updating the documentation is what updates the assistant — there is no separate script to maintain.
Related applications
AI Report Generator
Automated reports, written and delivered on schedule.
Automatically generate written business reports from your data on a recurring schedule.
AI Data Analyst
Ask questions about your business data in plain language.
Connect your data sources and ask questions in natural language — get charts, trends and explanations back instantly.