Videos
Call-Based Query Resolution on the Kognitos Platform
An end-to-end demo of call-based query resolution: transcription, classification, SOP lookup, and structured note generation — all in one Kognitos automation.
What's in this video
This demo walks through a complete customer-support automation that resolves phone-call queries without legacy RPA, intelligent document processing (IDP), or speech-to-text point solutions stitched together. Everything runs in a single Kognitos process.
The seven steps Kognitos automates
- Reference setup: The initial SOP mapping document and email template are loaded into Kognitos so the process has context.
- Recording upload: A customer call recording is provided as input — the trigger for the whole flow.
- Audio-to-text transcription: Kognitos uses a large language model to convert the recording to a written transcript.
- Query classification: The transcript is classified into one of several intents — voice troubleshooting, APN settings, slow connection, or, in this run, an order cancellation before delivery.
- SOP lookup: Using the classification, Kognitos pulls the correct SOP from the mapping file and verifies it by opening the file URL.
- Private and public notes: Private notes are generated in a structured "problem / finding / solution" format. Public, customer-safe notes are derived from the transcript plus the private notes.
- Agent email: Kognitos emails the agent the classification, the matched SOP, and both note sets so the human resolver can pick up where the automation left off.
Why teams care
Most call-center workflows require an RPA platform for system actions, a speech-to-text vendor for transcription, an IDP tool for SOP lookup, and custom code to stitch them together. Kognitos collapses all of it into one English-defined process that any operations lead can read, audit, and modify.
Questions answered in this video
Does Kognitos transcribe the call itself or rely on an external service?
Kognitos performs the transcription as part of the automation by leveraging large language models — no separate speech-to-text vendor needs to be stitched into the workflow.
What categories does Kognitos classify customer calls into?
The demo shows classification across categories including voice troubleshooting, APN settings, slow connection, and order cancellation. The category list is defined by the business and pulled from the SOP mapping file.
What gets emailed to the support agent at the end of the process?
The agent receives the issue classification, the extracted SOP, the private notes (formatted as problem / finding / solution), and the public notes generated from both the transcription and the private notes.
How is this different from a traditional RPA call-resolution bot?
Traditional setups need RPA for system actions, an IDP product for document lookup, a speech-to-text tool for transcription, and custom glue code. The video shows Kognitos handling transcription, classification, document retrieval, summarisation, and email in one English-defined process.