Support that answers itself
Braid puts an AI agent on WhatsApp, Instagram, email and live chat. It answers from knowledge you curate, and hands the conversation to a person the moment it should.
The loop
One thread, however the customer arrives
Four channels, one inbox, one agent. The customer never sees the seam and neither does your team.
- 01
A message arrives
WhatsApp, an Instagram DM, your support mailbox or the chat widget on your site. All four land in the same place.
- 02
The agent answers from your knowledge
It reads what you have given it — your site, your documents, your articles — and answers with a citation rather than a guess.
- 03
A person steps in when it matters
Ask for a human, sound frustrated, or ask something the sources do not cover, and the thread is handed over with a brief of what happened.
- 04
You see what happened
What resolved without a person, how fast the first reply went out, and every question your agent could not answer.
Knowledge
It answers from your material, not from the internet
Point Braid at your website and it crawls it. Upload your documents and it reads them. Write an article and it uses that. Every answer traces back to a source you put there, and the questions it could not answer become a list you can act on.
Handoff
The part most AI support gets wrong
An agent that will not let go is worse than no agent. Braid hands over on your rules — when a customer asks for a person, when they are unhappy, when the answer is not in your sources — and the person picking it up gets a brief instead of a transcript to scroll.
Tools
Give it something to actually do
Connect a tool server over MCP and the agent can look up an order, check a delivery or start a refund inside the conversation — with the actions you choose to gate held for a person to approve.
Put it on one channel and watch
Connect a channel, add your knowledge, and see what it resolves before you trust it with more.