CoreCx AI Hub
Speech analytics on every call: transcribed, measured for sentiment, silence and overtalk, summarised, searched semantically and graded against your QM forms. Run it on-premise or in your own cloud, where models can work fully offline.
Speech analytics, conversation intelligence and autonomous quality management, on-premise or in your own cloud.
Understand every call
Speaker-separated transcripts with word timestamps, language detection, tension scoring and recovered-call flags.
Automate quality management
Auto-QM scores every call against your own QM forms, instead of a small sample.
Search by meaning
Semantic search over calls and segments, theme discovery, AI summaries, intent, chapters, silence and overtalk.
From call to insight, automatically
Every recorded call flows through the same five steps, with no sampling and no manual tagging.
Calls arrive from iFlux Recorder, from every telephony vendor, sealed under your keys.
Speaker-separated transcripts with word timestamps, in many languages and accents.
Sentiment, tension, silence, overtalk, keywords, phrases and the reasons customers call.
Every call graded against your QM forms, with calibration and disputes.
Coaching drafts, worklists and reports by email, SFTP or webhook.
What speech analytics measures on every call
Sentiment and tension
How the customer feels, and whether it rises or falls during the call.
Talk, silence and overtalk
Talk-to-listen ratio, long silences and interruptions, per agent and team.
Keyword and phrase spotting
Find every call where a word or phrase was said, such as “cancel my card”.
Call drivers and themes
The real reasons customers call, discovered and trended automatically.
Intent and summaries
A short summary, intent and chapters for every conversation.
Quality scores
Every call scored against your QM forms instead of a sample.
Language detection
The language of every call, including calls that mix languages.
Recovered calls
Calls that started badly and ended well, so good work is seen.
Auto-QM, explained
Autonomous quality management scores every call against your own QM forms, the way your best evaluator would, and leaves people in charge of the final score.
How Auto-QM works
Sections, questions and weights, with critical items marked. Forms are versioned with effective dates, and each queue can have its own.
A phrase rule, a call metric, an AI judgement or a manual answer. You decide per question.
After transcription, 100 percent of calls are scored automatically. A missed critical item fails the call, and N/A questions are left out of the score.
Supervisors confirm or override each score. No automatic score becomes final without a person, and blind calibration keeps evaluators aligned.
Agents see their results and can dispute them. Coaching drafts, worklists and reports show where to act next.
Four ways a question can be answered
Did the agent give the recording disclosure?
Words and phrases, wildcards, NEAR, BEFORE and AFTER, said by the agent or the customer.
Was dead air kept under ten seconds?
Silence, overtalk, hold, talk-to-listen ratio and other call measurements.
Did the agent show empathy?
A language model reads the transcript and answers questions that need understanding.
Was the complaint handled to policy?
Left for the evaluator, alongside the automatic answers on the same form.
Traditional QA compared with Auto-QM
| Traditional QA | CoreCx Auto-QM | |
|---|---|---|
| Calls reviewed | A small sample, picked by hand | Every call, automatically |
| Time to feedback | Days or weeks | The same day |
| Consistency | Depends on the evaluator | Same form and rules on every call, plus blind calibration |
| Compliance misses | Found by chance | Critical items checked on every call and auto-fail |
| Evaluator time | Spent listening | Spent confirming and coaching |
What CoreCx AI Hub does for you
Understand every call
Speaker-separated transcripts with word timestamps, language detection, tension scoring and recovered-call flags.
- Speaker separation
- Language detection
- Tension score
Automate quality management
Auto-QM scores every call against your own QM forms, instead of a small sample. Critical misses fail the call automatically, a supervisor confirms every final score, and coaching, disputes and calibration all happen in the same place.
- Auto-QM on your own forms
- Critical items auto-fail
- Supervisor confirms every score
Search by meaning
Semantic search over calls and segments, theme discovery, AI summaries, intent, chapters, silence and overtalk.
- Semantic search
- Theme discovery
- AI summaries
Speech analytics on every call
Sentiment and tension, talk time, silence, overtalk and interruptions, keyword and phrase spotting, and the reasons customers call, measured on 100 percent of calls and trended by agent, team and queue.
- Sentiment and tension
- Talk, silence and overtalk
- Keyword and phrase spotting
- Call drivers and trends
Built for every language
A multilingual portal, speech and language models for many languages and accents, and reports on schedule by email and SFTP.
- Multilingual models
- Mixed-language calls
- Scheduled reports
Works with what you run
Built in, not bolted on
- Multi-tenant with row-level isolation
- Granular roles and permissions
- PII redaction by default, unmasking audited
- Offline, signed installer with SBOM
Where CoreCx AI Hub makes the difference
Score every call, not a two percent sample
QA teams listen to a sample and miss the calls that matter.
Autonomous quality management grades every call against your QM forms, with calibration, disputes, coaching drafts and worklists.
Cloud analytics send your audio to the vendor to be scored. CoreCx scores 100 percent, on-premise or in your own cloud.
Conversation intelligence without the audio leaving the building
Speech AI means shipping customer audio to a foreign cloud.
Speech, language, tone and embedding models run in the environment you choose, including fully offline on-premise.
Competitors’ AI is cloud-only. CoreCx runs on-premise or in your own cloud.
Understand every language your customers speak
English-first analytics mis-transcribe other languages, accents and calls that mix languages.
Speech and language models for many languages and accents, a multilingual portal and glossaries for your industry.
Many vendors treat languages beyond English as an add-on. CoreCx treats them as the default.
Put CoreCx AI Hub to work on your own infrastructure
A pilot against your real telephony, on your hardware, in days.
Frequently asked questions about CoreCx AI Hub
What does CoreCx AI Hub do?
Speech analytics on every call: transcribed, measured for sentiment, silence and overtalk, summarised, searched semantically and graded against your QM forms. Run it on-premise or in your own cloud, where models can work fully offline.
Can CoreCx AI Hub run on-premise or in your own cloud?
Both. Most customers install it on-premise from a signed offline bundle, with no internet connection needed. You can also bring your own cloud and run it in your own cloud subscription. Wherever it runs, it never sends your data back to CoreCx.
What does CoreCx AI Hub integrate with?
iFlux Recorder, On-premise or your own cloud speech and language models, Single sign-on, Email, SFTP and webhook delivery.
How is CoreCx AI Hub secured?
Multi-tenant with row-level isolation. Granular roles and permissions. PII redaction by default, unmasking audited. Offline, signed installer with SBOM.