Agents
Conversation design by situation
How each type of conversation is designed for the situation the customer is in, from a new enquiry to a cancelled flight to a claim, and how those designs are tested.
On this page
A customer with a new enquiry, a passenger whose flight has just been cancelled and a policyholder reporting a burst pipe at midnight all need different conversations. The difference is not only in the content. It is in the order of things, how much to ask before answering, when to offer options and when to stop and hand over. Telonic designs each type of conversation the way your industry requires it to be held, because how a conversation feels is most of what a customer remembers.
Why the situation shapes the conversation
An enquiry needs qualification woven into a helpful exchange. A cancelled flight needs acknowledgement before options. A claim needs accuracy and patience. These are different conversations, not different content in the same one.
An agent that answers every situation with the same pattern gets the facts right and the moment wrong. It lists rebooking options to a passenger who has not yet been told why they are stranded, or asks a buyer for their budget before answering their question. The facts come from your systems. The design decides how they are delivered.
How each situation is designed
Each design starts from what the customer needs in that moment, then sets the order of the conversation around it.
| Situation | What the customer needs | How the conversation is designed |
|---|---|---|
| A new property enquiry | A useful answer first, from someone who will not treat them as a form | Answers the question from your inventory, then asks about budget, timing and whether they are buying to live in or to invest, as part of the exchange. Offers a viewing when it fits |
| A delayed or cancelled flight | To know what has happened, and that someone understands | Acknowledges the situation first, states what has changed from the live schedule, then sets out the options the fare permits, then acts on the choice |
| Reporting a claim | Accuracy, patience and a clear next step | Checks the person is safe. Asks the questions the claim type requires, in order, one at a time. Confirms the details back. Ends with a reference and what happens next |
| A payment reminder | Clarity, without damage to a long relationship | States the amount and the date from the schedule, firmly and courteously. Offers a payment link. Routes any request to change the plan to a person |
| A coverage question | An answer they can rely on | Answers from the approved policy wording. If the customer disputes it, records the question and hands it to a person |
| A complaint | To be heard, and to reach someone with authority | Acknowledges the complaint without arguing it, records what the customer says, and hands the conversation to your team with a brief |
| A customer in distress | A person | Recognises the signs, stops the process in hand, and routes the customer to a person with the full conversation |
Designed per intent, per industry
An intent is what the customer is trying to do: book a viewing, change a date, report a claim. Conversation patterns are designed for each intent in each industry by people who know that industry, not assembled from a general template.
A design sets the sequence of the conversation, what must be established before an answer is given, when to offer alternatives, how to confirm what has been agreed, and when to stop and hand over. It draws on the industry model, so the questions are the ones the process actually needs. A motor claim and a home claim are designed separately, because the questions and their order differ.
How the channel changes the shape
The design of a conversation holds across channels. Its shape adapts to each one.
| Channel | How the design adapts |
|---|---|
| Voice | Short spoken sentences, one question at a time, and numbers and references read back to confirm them |
| Tappable buttons and lists where a choice is clearer than a typed answer, configured during implementation, and documents or links sent in the chat | |
| A complete reply in the same thread, covering every question the customer raised | |
| Web and app | The same as messaging, with the customer often already identified if they are signed in |
When a customer is upset or distressed
The agent notices how the customer is feeling as the conversation goes on, from what they say and how they say it. When a customer becomes frustrated, the agent changes its approach: it slows down, acknowledges the problem and stops pressing on with the process. If frustration continues or rises, it hands the conversation to a person.
Some signs mean the conversation should go to a person straight away: distress, a bereavement, a serious injury, a customer who may be vulnerable. The signals that trigger this are configured with you during implementation. The customer is routed with the full conversation, so nobody asks them to repeat something painful. See Sentiment and frustration and Handover to your team.
The agent offers a person to any customer who asks for one. It does not argue with an upset customer, or try to keep a customer who wants to speak to someone.
How designs are tested
Every conversation design is tested before it reaches your customers. Telonic runs it against test sets drawn from conversation patterns in your industry, scored against what a correct outcome means there. Simulated conversations built from your own scenarios are configured during implementation. See Testing before every release.
Your team takes part. Before go-live, the people who own each process read and listen to sample conversations, and tell us where the design is wrong for your customers. After go-live, every conversation is reviewed against your quality standard, so a design that works in testing but not in practice shows up in the scores rather than in complaints. See Quality review on every conversation.
What we need from your team
Your team knows things about your customers that no industry model holds. During implementation, we ask for the scripts and guidance your team already uses, the questions your process requires in each situation, and the moments where you would always want a person. The people closest to the conversations, such as senior agents and team leaders, are the most useful reviewers.
In practice
Wadi Assurance, an insurer, uses a Telonic agent for home claims out of hours.
- At 11:50pm, Noura calls to report water coming through her ceiling from the flat above. The agent asks first whether everyone in the home is safe.
- It confirms her identity, finds her home policy, and asks the questions the insurer's escape-of-water process requires, one at a time: when it started, where the water is coming from, which rooms are affected.
- It reads the details back to her, and explains her excess from her policy.
- It logs the claim, gives her a reference, and sends a WhatsApp message listing the photographs and reports the claim needs.
- Near the end of the call, Noura says she cannot cope with this after her father's death last week. The agent stops, acknowledges what she has said, and tells her a colleague will call her. It routes the claim to the senior handler on call with a brief, as Wadi Assurance specified for bereavement.
What your team controls
- The sequence and questions for each type of conversation, reviewed before go-live.
- The signs of distress or vulnerability that send a customer straight to a person.
- The scripts, disclosures and approved wording the agent uses in each situation.
- Changes to any design, tested before they go live.
Related
- Tone, voice and brandAgents
- Handover to your teamAgents
- Industry modelsAgents
- Sentiment and frustrationQuality and analytics
- Testing before every releaseGovernance and control
Product names and logos are trademarks of their owners. Their mention shows systems Telonic connects to and does not imply partnership or endorsement.