Cisco Contact Center Topic Analysis AI Transparency Technical Note

Publisher

Cisco Systems, Inc. (Cisco Trust Portal)

URL

https://trustportal.cisco.com/c/dam/r/ctp/docs/transparency/ai-transparency-topic-analysis-technical-note.pdf

Published

2026-04-09

Updated

unknown

Accessed

2026-09-05

HTTP status

200

License

Cisco proprietary documentation, all rights reserved; no-redistribution; short excerpts and locators only

Claims citing this source

• The Contact Center Topic Analysis technical note contains no restriction on using its output for employment decisions and the word employment does not appear in it. fact in context
• None of the 8 transparency notes read in this tranche publishes a list or count of supported languages. inference in context
• All 5 contact centre transparency notes are titled and scoped by feature rather than by model, and 1 model - GPT-4o mini - appears in 2 separately titled notes rather than in 1 consolidated model note. disputed in context
• The Topic Analysis note cites 2 further AI Transparency Technical Notes covering the transcription step that feeds it: an In-House Transcription note and a Third Party Transcription and Real-Time Translation note. fact in context
• Cisco states that for European Union customers and for India customers, Topic Analytics transcription and analysis are processed locally within those regions, and that other global customers' data is processed in the relevant United States European Union and Asia Pacific Japan and China regions. fact in context
• Cisco states that during Topic Analytics inference customer data is not stored or retained by the model and that only the generated call drivers and topics are produced. fact in context
• The Topic Analysis note states that Cisco frequently assesses the feature and publishes no cadence, threshold or metric, where the Auto CSAT and Agent Burnout Detection notes both publish retraining cadences. inference in context
• Cisco states that Topic Analytics input transcripts are stripped of personally identifiable information using Microsoft's Presidio before call driver extraction. fact in context
• Cisco states that Topic Analytics has a toxic hallucination filter which prevents the model generating new toxic harmful or unsafe language, and that toxic language already present in the source transcript may still surface because the filter does not alter the input. fact in context
• Topic Analytics chains 2 third-party models: GPT-4o mini hosted by the Azure OpenAI Service for call driver extraction, and Mistral-7B-Instruct-V0.2 hosted on Amazon Bedrock for labelling call topics. fact in context

Cite this source record

APA

WarmTransfer. (2026, April 9). Cisco Contact Center Topic Analysis AI Transparency Technical Note. WarmTransfer. https://warmtransfer.net/knowledge/sources/cisco-trustportal-tn-wxcc-topic-analysis

BibTeX

@misc{warmtransfer-cisco-trustportal-tn-wxcc-topic-analysis,
  title  = {Cisco Contact Center Topic Analysis AI Transparency Technical Note},
  author = {{WarmTransfer}},
  year   = {2026},
  url    = {https://warmtransfer.net/knowledge/sources/cisco-trustportal-tn-wxcc-topic-analysis},
  note   = {Cisco Systems, Inc. (Cisco Trust Portal), accessed 2026-09-05}
}