open questions, no menus
People just say what they are calling about.
- No long menus
- Follow-up questions stay within the same contact
Unexus Connect · the Virtual Agent
a straight answer, no menu
Customers do not want a menu and they do not want to wait. They want to be helped. The Virtual Agent understands the question, responds straight away and handles the contact. On the phone, in web chat and on messaging channels such as WhatsApp, in plain language, with a smooth handover to your colleague when that is needed.
what it is
The Virtual Agent sits at the front of your customer contact. It listens to the whole story, works out what it is about and answers where it can. Where it cannot, it transfers to the right colleague, along with what has been asked so far.
It works on telephony, web chat and messaging channels such as WhatsApp, and switches language automatically. You describe topics in plain language with a few examples; elaborate scripts are not needed.
what you get out of it
why you want the Virtual Agent
helped sooner
Plenty of questions can be settled on the spot.
- Answers from your own knowledge base
- It arranges callbacks itself
handed over with context
When it cannot handle something, it hands over warmly.
- On web chat and messaging your colleague sees the conversation
- Transfers to the right team or to a partner organisation
phone, chat and WhatsApp
Telephony, web chat and messaging channels such as WhatsApp. Not email.
- Switches language automatically
- The same behaviour, suited to the channel
outside office hours too
Reachable at the moments your team is not.
- Whatever can be settled straight away, is
- The rest becomes a call-back request or a message for your team
context from your systems
With a link in place it pulls context from your CRM or back office.
- With identification it knows who it is talking to
- You decide which systems it is allowed to use
questions you probably have by now
How does the Virtual Agent decide what to handle itself?
It recognises the intent behind the question and checks whether it can answer it on its own. Where context, authority or human judgement is needed, it transfers deliberately. Where that line sits is yours to set.
What happens when someone changes the subject?
It follows the conversation on meaning and context, so the customer does not have to start again. Steering it somewhere else does not mean walking back through fixed steps.
How much preparation does it take before it can go live?
You describe topics in plain language, with a few examples alongside. Elaborate scripts are not needed. How big that job is depends on your organisation and the kind of questions you get.
How does it connect to our existing systems?
Where the link is in place it pulls context from your knowledge base, your CRM or your back office, so it knows who it is talking to and can act accordingly. You decide which systems it is allowed to use.
How do we tell whether it is working once it is live?
You get insight into every interaction: which questions come in, what was handled, and where things get stuck. That is what you use to extend it step by step.
how you put it to work
-
step 1
describe the topics
In plain language, with a short description and a few examples per topic.
-
step 2
set the boundary
What it handles itself and when it passes things to a person. That boundary is yours to set.
-
step 3
connect what is needed
Knowledge base, CRM or back office, so it has context and can act accordingly.
-
step 4
extend it on what you see
Once it is live you see which questions come in and where things get stuck. That is what you extend it on, step by step.
Multiple languages work from the start, so you do not have to begin again for each one.
the technical side
for your IT colleague
what the caller hears first
That they are speaking to an AI assistant, and that a colleague is always within reach. That is not our choice to make: the EU AI Act requires it.
which channels it works on
Telephony, web chat and messaging channels such as WhatsApp, with automatic language switching. Email is not one of them. The behaviour adapts to the channel.
how it recognises the question
On meaning and context, not on fixed keywords. You set up topics with a short description and examples, so you are not building a decision tree.
what it takes from your systems
Answers come from your own knowledge base. Where a link has been set up and the caller identifies themselves, it can also use context from your CRM or back office. You decide which systems it is allowed to use.
how the handover works
A warm transfer to the right colleague, the right team or a partner organisation. What was asked travels with it; on web chat and messaging your colleague sees the conversation itself.
where it runs and what is kept
Azure Speech for speech and Azure OpenAI for understanding and answering, both Microsoft services in the Azure region West Europe. Nothing is trained on your data. The conversation with the assistant goes into the customer contact history, for the retention period you set.
what it does not decide
The assistant makes no decisions about your customer: it grants nothing and refuses nothing. Anything that calls for judgement or authority goes to a colleague.
looking further
this connects to
shall we look at which questions it could take on for your team?
Half an hour about the questions that come in most often. That shows what it can handle itself and what you want to keep with your team.
No obligations. Would you rather look around on your own first? The hard facts are on security & compliance.
