Questions about TelAI
How the agent sounds and behaves on a call, what you can control, and what you get afterward. If your question isn't here, email us.
The conversation
What happens when a caller interrupts?
The agent lowers its volume and trails off within a couple of words, then lets the caller speak. It also keeps track of how much of its sentence the caller actually heard, so it won't assume they got information it never finished saying.
Short listening cues such as "mm-hm", "yeah" and "go ahead" don't count as interruptions, so the agent keeps talking through them.
What if the caller is on speakerphone?
TelAI rejects speakerphone echo, so the agent's own voice coming back through the caller's phone is never mistaken for the caller talking.
What if the agent mishears a word?
Phone audio garbles words all the time. TelAI reads unclear words against the context of the conversation and the knowledge base, so it can tell that "Otto" was really "auto" instead of asking the caller to say it again.
Does it read long scripts?
No. Answers are kept short and conversational, the way a person would talk on the phone. You can still give it detailed instructions for what to cover.
Knowledge, instructions and voices
How does the agent know about our business?
You give each call or campaign a knowledge base of product docs, policies or FAQs. It can run to tens of thousands of words, and replies stay fast with a large knowledge base from the start of the call to the end. You also give it a scenario, a persona and instructions for what the call is for.
Can the agent go off script?
The knowledge base and your rules are what keep it on script. The agent answers from the material you load for the call, and you set strict rules for what it may offer, what it has to confirm first and what it must never say. If a caller asks something your material doesn't cover, the rules decide what happens next, such as taking a message.
Can different lines or campaigns behave differently?
Yes. The scenario, persona, instructions, knowledge base and voice are all set per call or campaign.
Can we use our own voice?
Yes. Along with the premium voice library, TelAI supports custom and cloned voices, and you choose the voice for each call. If the chosen voice is ever unavailable, the call uses your default voice instead.
A cloned voice requires written permission from the person whose voice it is.
Is there a pause before the first reply?
No. The voice is loaded while the phone is still ringing, so the agent is ready to speak as soon as the call connects.
Can the call sound like it is coming from an office?
Yes. You can add optional background ambience, such as office sound, to make the call environment sound natural.
Calling
Does TelAI answer calls, place them, or both?
Both. It can answer inbound lines and place outbound calls.
Can it get through a phone menu?
Yes. It detects automated menus such as "For English, press 1" and presses the keypad options. It also recognizes hold announcements, waits through them, and starts the conversation when a live person picks up.
What happens if nobody answers?
TelAI can retry a call automatically when there is no answer, and it detects when a call reaches voicemail.
What number does it call from?
You choose the caller ID from your verified numbers.
How does the call end?
When the conversation is over, the agent wraps up and hangs up gracefully. A supervisor can also end any call with one click.
Can the agent transfer a call to a person?
Yes. TelAI can transfer a call to someone on your team when your rules call for it, such as an emergency, a caller who asks for a person, or a question only a specialist can answer.
Can TelAI connect to our own software?
Yes. Our engineers build custom integrations with the systems you already use, so the agent can look things up and take actions during the call, such as checking an order, booking an appointment or creating a work order, and send each call's details into your software afterward. See custom integrations.
Is there an API?
Yes. A simple HTTP API lets you place calls and pull call status, history and analytics into your own systems. For lookups and actions during a call, see custom integrations.
Supervision and review
Can someone on our team listen to a call while it is happening?
Yes. Anyone with access can listen to any active call from the browser, see real-time audio levels, and read a live transcript as both sides talk. See live control.
Can a person step in?
Yes, in two ways. You can send the agent new instructions mid-call to change its direction, or type a line and the agent says it right away in its own voice. You can also hang up with one click.
Are calls recorded?
Yes. Calls are recorded in two channels, and you can play recordings in the browser, download them or delete them. The full call history can be searched by number, by scenario or by anything said on the call.
What does the post-call analysis include?
For each call, TelAI writes a one-line summary and scores the customer's emotion, mood, stability, frustration and annoyance. It records whether the call reached its goal, and it flags profanity and pulls out key quotes and behavior tags.
The analytics dashboard rolls these up into a success rate, mood distribution and frustration and tag trends, next to call volume, answer and failure rates, talk time, status breakdown, top numbers and voice usage.
Reliability
What if a response is slow?
Slow responses are retried automatically. If there is still a gap, the agent covers it with a natural line like "Sorry, you cut out" so the caller doesn't sit in silence.
What if the connection drops?
Connections recover on their own and reconnect in the middle of a call.
Do software updates interrupt calls?
No. Updates wait for live calls to finish, so an update never drops a call in progress.
Compliance
Does the agent tell callers it is an AI?
Yes. The agent says it is an AI at the start of each call, and callers are told the call is recorded. You choose the wording in your scenario, but the disclosure itself is required.
Do we need consent to place outbound calls?
Yes. You must have prior express consent from the person you are calling before TelAI places an outbound call to them, and prior express written consent for any marketing call. In February 2024 the FCC ruled that AI-generated voices count as artificial voices under the Telephone Consumer Protection Act, so these calls need the same consent as prerecorded calls.
What keeps TelAI from calling the wrong people or at the wrong time?
Outbound calls are placed only between 8 a.m. and 9 p.m. in the recipient's local time. Numbers are checked against the National Do Not Call Registry and your opt-out list before dialing, and anyone who asks not to be called again goes on that list automatically. Calls carry STIR/SHAKEN caller ID attestation from your verified numbers.
Are our calls used to train TelAI?
No. TelAI was trained on hundreds of thousands of calls over eight years of development, but your calls and transcripts are never used to train it. Call data is encrypted in transit and at rest and stored in the United States, and a business associate agreement is available for healthcare customers.
What uses are not allowed?
Calls to people who haven't consented, political robocalls, impersonating a real person or another company, cloning a voice without permission, and debt collection. The compliance page has the details.
Still have a question?
Email us and one of the engineers who built TelAI will answer. You can also read how it works or the full feature list.
Request a demo