Conversations
Every call, text, chat, and web session in one place — how to filter it, what the statuses mean, and the habit that makes agents better.
Conversations is every interaction your agents have had, across all four channels, in one list. It's the screen you'll spend the most time in, because it's where you find out what your agents are actually saying.
The list
Each row shows the date, channel, contact, direction, agent, result, and status. Search across contact name, phone number, and message content.
Result is the outcome your agent recorded — Appointment Booked, Follow Up, Voicemail, Wrong Number, Not Qualified, Do Not Call. These are the ones you renamed to match your business in Define what counts as a win.
Status is what happened mechanically — Active, Replied, Opted out, Timed out, Closed, Completed, No answer, Voicemail, Busy, Failed, In progress.
Result and status answer different questions. Status tells you whether the conversation worked. Result tells you whether it achieved anything. A call can be Completed and still have gone nowhere.
Reading a conversation
Open any row for the full detail: the transcript, the recording where there is one, the contact, which agent handled it, and any post-conversation analysis answers.
This is the loop that makes agents good:
- 1
Filter to a result you're unhappy with
Not Qualified, or a status of Timed out. Successes teach you less than failures.
- 2
Read five of them end to end
Read, don't skim. The problem is usually one specific moment.
- 3
Find the pattern
Five transcripts is enough to see whether it's the same thing every time.
- 4
Fix the single worst thing
One change. Then come back in a week and read five more.
Read the ones that went well, too
Occasionally read a conversation that ended in your best result. It tells you what to protect when you're editing — people quite often "improve" an agent by removing the line that was doing the work.
Filtering
The filters are worth learning because they turn a firehose into an answer:
| Question | Filter |
|---|---|
| Is my new agent working? | Agent + last 7 days |
| Why is my booking rate down? | Result = Not Qualified, this week |
| Did that campaign land? | Date range around the campaign |
| Are texts getting replies? | Channel = SMS, status = Replied |
| Who's asking to be removed? | Result = Do Not Call |
That last one is worth a weekly look regardless. A rising count means your list or your targeting is off, and it's the earliest warning you'll get.
Acting on a conversation
Three actions matter beyond reading:
Reply sends a text yourself, in the thread. Read the warning below first — this one has a consequence.
Block stops any further texts to that number, adding them to your blocked list. Use it when someone asks to stop in words that aren't the standard opt-out keywords — "please don't message me again" won't be caught automatically.
Create a test from this conversation turns a real interaction into a repeatable test case, so a conversation that went wrong once can be checked against every future change. See Testing your pathway.
Replying by hand takes the conversation over — permanently
The first manual reply you send in a text thread switches that conversation's automated sequences and follow-ups off for good. There's no way to hand it back to the agent, and no undo. The reply box tells you before the first one, and a banner marks the thread afterwards with who took it and when.
This is deliberate — an agent cheerfully texting over the top of a colleague is worse — but it catches people out. If you're only looking, don't send. If you want the automation to keep running, act somewhere other than the reply box.
Turn a bad conversation into a test before you fix it
Then you know your fix actually worked, and it stays fixed. See Testing your pathway.
The four channels
- Call — voice, inbound or outbound. Transcript plus recording where recording is on.
- SMS — the full text thread, including messages sent by a campaign sequence and by follow-ups.
- Chatbot — a text chat session from the widget on your website.
- Web Agent — a voice conversation held in a visitor's browser.
They're deliberately in one list. A lead who filled in a form, got a text, and then called you is one relationship, and seeing it as one sequence is usually the point.
What to look at weekly
- Five transcripts from your worst-performing result.
- Anything with a Do Not Call result — and whether those numbers made it onto the list.
- Timed out conversations — usually a texting timeout set too short, or an agent asking questions people don't answer.
- Your best conversation of the week. Then work out what it did differently.