Innovation and AI Analytics for Call Centers Conference

In April 2022 SentiSquare organized a conference in WorkLounge dedicated to innovation and analysis using artificial intelligence for call centers.

The event brought together many industry experts from contact centers, call centers, and customer service departments from banks, insurance agencies, energy providers, retail companies, and more.

The speakers were Tomáš Brychcín from SentiSquare, Michal Pleyer from Atos, and Tomáš Roubíček from IXPERTA.

For SentiSquare, we presented the topic “Contextual call center analytics”. Tomáš Brychcín, CEO of SentiSquare, explained how we perform contextual analytics for call centers. Let’s take a brief look at what the attendees learned from SentiSquare’s presentation at the conference:

The main ingredient of SentiSquare technology is called Distribution Semantics.

The main idea is that “Meaning comes from context”. As a result, one is able to discern the meaning of an unknown word.

Distributive Semantics acknowledges the ideas of English linguist J. R. Firth, which he made as early as the 1950s when uttering the now-famous quote:

“A word is characterized by the society in which it occurs.” - (J. R. Firth, 1957).

Just as the human brain works, so do the SentiSquare algorithms. They learn meanings from context. And how does Distributional Semantics help us in analyzing calls?

We face many problems in call analysis:

  1. Information density is much thinner
  2. Spontaneous talk (lots of ballast)
  3. Frequent word splitting (parasitic words)
  4. “Non-speech” sounds are often  transcribed (parasitic words)
  5. The need to perceive a much wider context than for normal text

Call analysis can’t be made without context. Thanks to distributional semantics, SentiSquare technology can tackle these problems while also providing the following benefits:

  1. Algorithms learn meanings directly from text (language independence)
  2. Algorithms are adapted to client data (higher success rate)
  3. We implement projects in various languages including Czech, German, English, Hungarian, Polish, Japanese, Chinese, etc.

With these capabilities, SentiSquare technology can analyze calls by knowing the true meaning of the call from its context.

Having explained the principle behind SentiSquare technology and how its approach overcomes the challenges of call analysis, we also answered the following question:

Why analyze calls in the call center?

According to our experience, we’ve put together the top reasons that drive our existing clients to call center analytics:

  • Identifying trends and themes in calls
  • Insight into customer sentiment and the way customers express themselves
  • Overview of dissatisfied callers
  • Agent performance review
  • Checking for adherence to call scripts
  • Searching for cross-selling opportunities
  • Reviewing successful and unsuccessful calls (subsequent call-script improvements)
  • Not too lax/aggressive operator?
  • Predicting customer churn

And finally, perhaps most importantly:

What all can be extracted from calls with Distribution Semantics?

Before starting the “Contextual Analytics for the Call Center” project, we ask our clients the following question: “Do you want to focus your analytics on your customers or your operators? Or both?”

And what is the difference between these two approaches?

Focusing on the customer will give you the answers to these questions:
  • Why are they calling? What is the reason for the contact?
  • How do they express themself?
  • The sentiment (satisfied/dissatisfied?). Focus on finding the problem. Was the problem resolved during the call?
  • What products do they mention?
  • What competitors do they mention?
  • Customer journey.
  • Reason for rejecting the operator’s offer.
  • Do they not want to leave?
Focusing on the operator will give you a detailed evaluation of the agents and their performance:
  • Did they follow the call script?
  • Did the customer understand what they were saying?
  • What reason did they have for declining the operator’s offer?
  • Did they attempt to cross-sell?
  • Did they waste time?
  • Do they negatively impact customers?
  • How is the operator performing relative to others?
  • Are they improving?
  • What types of calls do they have problems with?

The software automatically detects everything. Your people no longer have to conduct interrogations. Instead of having a percentage of calls, you have an overview of everything that’s going on at the call center.

Finally, we’d like to thank our partners IXPERTA and Atos for co-organizing this conference. We hope everyone enjoyed the event, and we’ll see you at the next one!

Call Center Conference

Read more