It's listening.
So your reps can.
Codd Call Assistant is real-time agent assist for live calls. It hears the conversation as it happens and puts the next right thing to say on your rep's screen, grounded in this customer's actual history, your approved policies and your certified business metrics.
Scroll to follow one illustrative call, start to finish. Names, numbers and policies are fictional.
- Caller matchedDana Whitfield · Northwind Logistics
- AccountGold plan · customer since 2019
- Open ticket#48213 · billing
- Last callSummary from 12 days ago
- Account healthActive usage ↓ 18% (certified metric)
Customer“I was charged twice for my renewal. Again.”
▍I can see both charges from March 3. I'll reverse the duplicate right now. You'll see it back on your card in 3 to 5 business days.
The answer is ready when the customer stops talking.
Call Assistant transcribes both sides of the call as it happens. When the customer finishes a thought, it pulls what matters (their charges, the open ticket, the refund policy) and drafts one to three short sentences the rep can say out loud.
No searching three tabs while the customer waits. Our design target is under two seconds from the end of the customer's sentence to the first words of the suggestion.
Customer“Honestly, I'm starting to look at other options.”
▍Her team's active usage is down 18% since the plan change in January. Ask what changed for her team before offering anything.
Not a help article. Your numbers.
Most agent-assist tools retrieve a knowledge-base article. Call Assistant also reads this customer's real numbers, from the same governed semantic layer Codd AI uses for analytics, so a rep and a dashboard are working from the same definition.
So when it says “usage is down,” usage means exactly what your business has certified it means, not whatever a model guessed from a column name.
Read how a contextual semantic layer works in The CDO's Guide
Customer“Okay, the card ends in ”
▍Payment step detected. Transcription is paused: card details are not captured, stored or sent to the model while payment is taken.
Some things it should never hear.
- Transcription pauses while card details are read out, so they don't reach a transcript or a model.
- Personal details are redacted before anything is sent to the model, and transcripts follow your retention rules.
- Designed to follow your call-disclosure and recording-consent configuration, which varies by jurisdiction.
Rep“Let me see what I can do on price for you.”
▍Credits over one month need manager approval. You can offer the loyalty rate now: 12% off for 12 months, no approval needed.
It won't make promises you can't keep.
Call Assistant only suggests refunds, discounts and deadlines that exist in your approved policies and offers, and every suggestion shows where it came from.
Reps rate each suggestion with a thumbs up or down. That feedback, and whether they actually used it, keeps tuning what Call Assistant says and when it speaks up.
The call ends.
The memory doesn't.
Before the rep picks up the next call, Call Assistant writes a summary back to your CRM. It becomes part of the history the next rep sees the moment Dana calls again.
▍Duplicate renewal charge from March 3 reversed (3 to 5 business days). Customer mentioned looking at alternatives after a drop in team usage since the January plan change. Offered loyalty rate (12% off for 12 months); accepted. Follow-up: usage check-in scheduled for 30 days.
Everything you just watched, feature by feature.
Each feature is tagged with the moment it showed up. Tap a timestamp to jump back to it.
Context before hello
Matches the caller and loads their account, open tickets and last call summary while the phone is still ringing.
Live two-sided transcript
Transcribes the customer and the rep separately as they speak, tuned to your product names and jargon.
Speaks up at the right moment
Suggests when the customer finishes a thought or a key moment comes up, not on every word, so reps aren’t distracted.
Next best thing to say
One to three short sentences the rep can say out loud, plus the next action to take, streamed onto the screen.
Moment detection
Flags questions, objections, cancellation risk and compliance moments as they happen in the conversation.
Grounded in certified metrics
Uses governed metrics from the Codd semantic layer, so “usage” or “revenue” means what your business certified.
Your playbooks, with sources
Draws on approved policies, scripts and objection guides, and shows the source behind every suggestion.
Payment pause and redaction
Pauses transcription while card details are taken and redacts personal details before anything reaches the model.
Policy guardrails
Only suggests refunds, credits and offers that exist in your approved policies, and points to the approved alternative.
Rep feedback loop
Reps rate suggestions with a thumbs up or down. Their feedback tunes what Call Assistant says and when.
Automatic call summary
Writes a summary, outcome and follow-ups back to your CRM, ready for the next rep who takes the call.
Side panel next to the call
Suggestions, live transcript and key customer facts in one panel beside the call, in the tools your reps already use.
Streams live call audio from RingCentral (RingCX Call Streaming or RingEX Supervise), with each side of the call kept separate.
Customer history and open tickets from your CRM, approved policies, scripts and objection guides, and certified metrics from Codd's governed semantic layer.
Transcribes the call, decides when to speak up, and drafts suggestions grounded in your data, knowledge and policies.
Suggestions appear in a side panel next to the call. Summaries are written back to the customer record.
The next call is
already ringing.
Call Assistant is one of the new Codd AI Apps. We're opening early access to a small group of sales, support and contact-center teams on RingCentral. Bring your CRM and the policies you already have; Codd supplies the semantic layer.