An illustrative example: a customer calls to reschedule a meeting, but the conversation receives a low score against a sales scenario. The salesperson did not ask about budget or make an offer because they were solving a different problem. The red number looks convincing, but first establish what it measured.
AI means artificial intelligence. It helps identify moments worth reviewing. Value comes when the manager checks the score explanation and turns it into a specific working agreement.
What the update describes
The official September 2026 roundup describes Speech Analytics 2.0: BitrixGPT creates and supplements script rules, explains scores in more detail and provides managers with summaries. This is an update from the September roundup, not a launch announcement for today.
The tool guide says updates roll out gradually. Before implementation, check portal availability, plan conditions, permissions and call processing settings. Do not promise a feature based only on a news illustration.
Check the original conversation first
Use an anonymised training example with a clear customer need. Compare the recording, transcript and score explanation. If an important phrase was recognised incorrectly, the subsequent conclusion needs checking.
For a team in Kazakhstan, separately test the languages used and switching between them. Do not generalise one clean test to every recording, device and connection condition.
Clarify the person's goal: a new purchase, support or changing an existing agreement. Not every conversation needs the same question sequence.
Check the appropriate script
Here, a script describes the expected conversation and assessment rules. Read the criteria: which observable actions do they test? “Communicated well” is too vague for a useful review.
The guide allows a conversation to be assessed again, including with another suitable script. Do not change criteria just to improve the number: first explain why the new option fits the task better.
If rules are supplemented automatically, assign someone to review their meaning. Keep the conditions under which a result was discussed; comparing calls across different periods may otherwise mislead.
End with one improvement
Discuss a specific moment and next useful action with the salesperson: clarify the customer's question, name the agreed next contact method or summarise the outcome more clearly. A summary helps select a call for attention, but does not replace its content.
Do not equate a high score with a sale or a happy customer without checking the outcome separately. Begin with a small set of varied scenarios and compare system explanations with the manager's review.
QQDev implements Bitrix24. On the implementation page, discuss a speech analytics pilot: which calls to analyse, which criteria help and who reviews disputed scores. Start with an anonymised example where the team and system understood the call's purpose differently.