Skip to main content
Intent understanding

Intent understandingIdentify customer needs and route them to the right next step

“It still hasn't arrived” could mean a tracking check, a missing parcel or a refund request. AI reads business intent from the conversation and order context, and routes each inquiry down the right path.

Less back-and-forth for customersInquiries land on the right pathUnclear requests get clarified firstSee top issue types at a glance
Intent understandingLive Chat · Order #S10482

The sofa still hasn't arrived. It's been two weeks.

Order #S10482 linked · 3 parcels · 2 delivered

AI detected · IntentMissing parcel (not a delay or a refund): the order shipped in 3 parcels and the customer has received 2
Missing parcel 92%Shipping delay 6%Refund 2%

You've received two of the three parcels. The third (the sofa base) is still in transit and is expected to arrive by May 27. I'll follow up on that one for you now.

How intent understanding finds different needs behind the same words

“It has not arrived” could mean a delayed shipment, a missing parcel or a refund request. Order and conversation details help identify the need; when unclear, AI asks first.

4 customers, one message: “still hasn't arrived”Read against order status
2 deliveredMissing parcel inquiry3 parcels in the order · Routed to multi-parcel shipping
5 days overdueShipping delayShipped 12 days ago, 7-day delivery promised · Follow up on the delay and offer compensation
Marked as deliveredDelivery exceptionCustomer says it never arrived · Reship after carrier check
Not yet shippedRefund requestCustomer asks to cancel · Routed to order cancellation workflow

The same “still hasn't arrived”
read against order status to find the real request

  • Read the order, not just the words“Hasn't arrived” can be four different issues depending on order status, so AI checks the order before deciding.
  • Classify common enquiries automaticallySales, shipping, after-sales and complaint enquiries can be classified and sent to the right workflow.
  • Use your team’s service categoriesKeep your existing service categories and adjust the criteria to fit how your team works.
When intent is unclear, ask firstLive Chat

Can I return this?

Intent unclear: return, exchange or refund · Clarify first

Happy to help. Would you like to return it for a refund, or exchange it for a different size or color?

One size up, please.

Detected · Exchange requestRouted to exchange workflowSize L in stock · Within 30-day exchange window

Fewer misroutes, fewer follow-ups
inquiries go straight to the right workflow

  • Ask when unsureWhen intent is unclear, AI asks first instead of guessing a path.
  • Route to the right workflowExchanges go to the exchange workflow and complaints to the human queue, each under its own rules.
  • See the most common enquiry typesSee enquiry volumes and misrouted cases by category to identify what to improve first.

Less explaining for customers
fewer detours for every issue

Common customer service challenges

  1. Every “still hasn't arrived” gets treated as a tracking request, so missing parcels and refund requests go down the wrong workflow
  2. One wrong transfer and the customer has to explain it all again
  3. Returns, exchanges and refunds get lumped together, with no clear path

What changes with Solvea

  1. Order status reveals the real request, which is routed to the matching workflow
  2. Right path the first time, so customers don't have to repeat themselves
  3. Similar requests are kept apart, each following its own rules
93%Ticket reply rate

unibuying has AI take every ticket first and routes complex issues to the right path by rule.

unibuyingConsumer electronics & accessories · Smart securityView customer stories →

About Intent understandingFAQ

How accurate is intent recognition?

Accuracy depends on how intents are defined and on the connected order and knowledge data. Common intents are validated on real conversations before launch, and when unsure, AI asks instead of guessing.

Can intents be customized?

Yes. Define customer intents and workflows in your team’s own terms, and adjust how common enquiries are classified.

What if AI gets it wrong?

Once the customer or agent corrects the intent, the conversation moves to the right workflow. Reviewed mistakes can then improve the detection and routing rules.

How does Intent understanding relate to the Knowledge base?

Intent decides which path to take; the Knowledge base decides what to say along it. The two work together.

Which products include Intent understanding?

Intent understanding is available in AI Customer Service Agents, Live Chat, Email, APP and Voice AI Agent, and often works alongside Chatflow, Human–AI collaboration and AI Tags.

See Intent understanding in your workflow

Bring a recent customer question. We’ll show how this works in your process.

Book a feature demo