Aqvitus answers your customers before anyone on your team has to, on chat, email, SMS, voice and RCS — and shows you every action it wants to take before it takes it.
A product of EvolvLabs, LLC · built on Exolvra engineering
Where is my order?
Out for delivery — arrives Thursday.
Can someone look at invoice INV-2291?
The AI finished the first one on its own. The second is Priya’s, and she did not have to ask Mara to start again.
Support software asks you to choose between answering customers quickly and knowing what was said.
Deflection tools answer fast and leave you guessing. Ticketing tools keep the record and leave the customer waiting. The AI ones in between will happily issue a refund because a sentence convinced them to.
Aqvitus is built the other way round. The AI drafts the reply and proposes the action; the platform then decides, in deterministic code, whether that action is allowed — and holds it for a person when a rule says a person should see it. Nothing happens because the AI said so.
How an action gets approved“I’m sorry the lamp arrived damaged. I can refund it to your card today.”
The reply is written and the refund is ready. Neither is sent: a rule you wrote says a refund this size waits for a person.
Six positions the product has to keep, not slogans. Each one is something you can go and check in the console.
The AI takes the first turn on every conversation. When it should not finish the job, the conversation steps down to a person with the history already summarised.
Every action the AI proposes is checked against rules you wrote, in ordinary code, before anything happens. You can read what it wanted to do and why it was allowed.
A resolution the platform confirmed and one it only assumed are never added together — not in the inbox, not in a report.
Easy enough to set up on your own in an afternoon. Roles, approvals, audit and workspaces for a support team that has to answer for it.
A router picks a model per task, so no part of the product is welded to one provider. Bring your own key, or run Aqvitus self-hosted with local models.
The proof is your own conversations. Start a trial, put the chat on your site and talk to your own AI before you decide anything.
The first two beliefs are only worth writing down if they hold when the AI wants to spend your money. Here is the same customer as the lane under her changes.
Where is my order?
Out for delivery — arrives Thursday.
The AI knew the answer and nothing it had to do was risky, so it answered and closed the turn. Nobody on the team saw this one.
“I’m sorry the lamp arrived damaged. I can refund it to your card today.”
The lamp arrived cracked. The reply is written and the refund is ready, and neither has happened: a rule this workspace published holds a refund over $25 for a person.
Thread summaryDamaged lamp on order A‑1042. A $48.00 refund is drafted and waiting. Two earlier messages, both answered.
Priya picks it up already knowing what was said and what the AI wanted to do. Mara is not asked to start again.
A conversation only travels in this direction. Aqvitus can step it down toward a person on its own; it never quietly gives the AI more room than you granted it.
Short, because there is nothing here we would have to dress up.
Aqvitus is a product of EvolvLabs, LLC, built on Exolvra engineering. The platform, the chat widget, the carrier integrations and the console are ours; the AI models behind them are interchangeable by design, and you can supply your own.
There is one address for everything — sales, support, self-hosting, a hard question about how the rules engine decides — and no contact form standing in front of it.
Contact usAqvitus is new. When customers want to say something about it, they will say it themselves. Until then the screens on this site are the product, and the trial is the proof.
No per-resolution fees. And our reports never count a customer going quiet as a success.
See pricingThe three lanes, the rules that decide, and every module in the console.
See the platformWorkspace isolation, per-workspace keys, redaction before any prompt, and an audit trail you can read.
How we handle dataChat, email, SMS, voice and RCS in one inbox, with one customer staying one conversation.
See the channels
Start a trial, put the chat on your site, and watch the first thing the AI wants to do before it does it.