Second Chair is built for three. It knows who is talking, warns the professional privately the moment they cross a regulatory line, and says nothing at all for the other ninety percent of the conversation.
Waiting for the first turn.
Nothing to flag yet.
An adviser is sitting with a couple, talking through a retirement plan. He says “this one is guaranteed to return eight percent.” He didn’t mean anything by it. It slipped out.
That sentence breaks a rule, and nobody in the room notices. Six weeks later a compliance officer finds it in a transcript, and the firm has a problem that costs real money.
Second Chair tells him the second he says it, in his earpiece, where only he can hear it — along with the words to put it right. He corrects himself in the same breath. The problem never exists.
Drop a conventional voice agent into a meeting with three people and it fails in three specific ways at once. It cannot tell you who said what, so the transcript is useless as a record. It answers questions that were never put to it, because it assumes every utterance is addressed to it. And it can only tell you about a problem after the call, when the sentence that caused it is already on the recording.
Second Chair inverts all three. Speakers are separated as they talk. The agent answers only a direct address. And the warning that matters arrives while the sentence is still being spoken — early enough to change what comes next.
The whole design follows from a single decision: what the agent says to the professional and what it says to the room are different things, carried separately.
Compliance warnings, coaching, and the phrasing to recover. It fires on a deterministic rule tier that measures 0 ms, so it lands while the professional is still mid-sentence. The client never knows it happened.
Spoken answers, drawn from the client record. It opens its mouth only when someone addresses it by name — twice across a twenty-turn call — and treats being mentioned in the third person as conversation, not a summons.
The earpiece is the version people remember, and it is the one running today. Two more are on the way, for the places most regulated conversations actually happen.
A laptop and one earbud
For meetings that happen face to face. A laptop or tablet on the table picks up the room, and the professional wears a single wireless earbud. The warning is panned to that ear alone, so nobody else hears a thing. This is the setup the demo runs today.
Nothing to install, once it ships
Second Chair alongside Zoom, Teams or Meet, taking the call audio directly and putting warnings on screen instead of in an earpiece. No hardware, nothing for the client to notice. Most regulated advice happens here now, so this is what we are building next.
Connects to your telephony
Audio from the phone system you already run, with warnings in the screen the agent is already watching. This is where call volume is highest — claims lines, advice desks, contact centres — and where the integration work is heaviest.
The pattern repeats anywhere a professional advises two people at once and the record of what was said carries legal weight.
Suitability, fee disclosure and performance claims are all decided by exactly what was said out loud. When a second household member is present, the person who kills the deal is usually the one who stopped talking twenty minutes ago.
Intake conversations routinely include an adult child who answers on the patient's behalf. Attribution matters clinically and legally, and the transcript is full of protected health information from the first minute.
The adjuster is capturing dates, amounts and locations while a distressed claimant talks, often with a family member interjecting. Missed details mean a callback; a recorded misstatement means a complaint.
An intake call has to gather enough to assess a matter without crossing into advice before an engagement exists. That line is crossed in a sentence, and nobody notices until the transcript is read back.
The demo runs the full reference call in your browser — no signup, no key. Put headphones on: the earpiece channel is panned left, the room channel sits centre.