Fini
Self-improving AI for customer experiences
299 followers
Self-improving AI for customer experiences
299 followers
Fini is the self-improving AI agent for customer support. It resolves 90% of tickets at 99% accuracy across voice, chat, and email, built for fintech, banking, and other regulated industries where a hallucination isn't an option.
Most support AI needs constant re-tuning. Fini detects its own knowledge gaps, learns from real resolutions, and gets sharper on its own. Live in 14 days. Fully autonomous in 30. Works on your existing helpdesk, no migration.
This is the 4th launch from Fini. View more
Agent Handshake by Fini
Launched this week
Your customers now send personal AI agents to contact support, and an agent will reword a request until the answer changes. Agent Handshake gives it the same answer every time. Fini decides from intent, live account data and your rules, and checks every reply.
Payment Required
Launch Team / Built With
I'd want the answer to change when the account facts change. Keeping that separate from someone just rewording the same request seems like the important distinction here.
@oleksii_sekundant You're absolutely right!
@oleksii_sekundant Exactly how it works. If the order is refunded or the plan changes, the next answer reflects that, because Fini reads the live account data each time. What it ignores is the wording, so asking again in a new way gets the same answer.
The approval line is the part I'm most curious about — anything that moves money can go to a person to approve, and you choose which actions need approval. How do you draw that line for the gray-area actions that don't move money directly, like changing a plan or issuing account credit? Is that a fixed list, or does repeated rewording from the same agent push it toward escalation?
@jcbuilds Good question. It's not a fixed list we set for you. Each team picks, action by action, what an agent can ask for on its own and what needs the customer's approval or a person. Checking an order status can go straight through, while a refund or an account credit can be set to wait for someone on your team. Plan changes sit wherever you put them. Rewording doesn't move that line: the same request gets the same decision however it's phrased, and a request that needs a human goes to your team.
Hi Product Hunt, I'm Deepak, co-founder of Fini. We built Agent Handshake because of something we started seeing in support queues.
Customers have started sending personal AI agents to contact support for them. Assistants like Muse and Dots can already do it. An agent is patient: it will reword the same refund request again and again until a different answer comes back. If that works, the customer who wrote in once got a worse deal than the one who sent an agent.
Agent Handshake is the part of Fini that handles these agents. The same request gets the same answer, whoever is asking and however many times they ask.
1. Same request, same decision. Fini decides from the customer's intent, the live account data and the rules you wrote, not from the wording.
2. Every reply is checked before it is sent. Guardrails check banned terms, confidential details, allowed links and your plain-language rules. A reply that fails is rewritten or handed to your team.
3. It knows when to stop. Fini escalates to your team, or sends the agent back to its owner, when a request needs a human. Anything that moves money can go to a person to approve.
Likely personal agents are tagged in the Inbox, and you choose which actions need approval.
It is built to serve agents, not block them. Agent Handshake is open to everyone: any team running an AI agent for support, ours or not. On Fini it is switched on per account.
Here's a test you can run on your own support: write one refund request twenty different ways, send them all, and check whether the decision moved. I'd love to hear what you find, and which agent you expect to see in your queue first.
How it works
Personal agents negotiate. Your policy shouldn't 😎
@zuzanna_piskorska BINGO!
customers sending their own agents to support and rewording until the answer flips is a real thing now, hadn't thougth about it from the support side
can Fini tell a personal agent apart from a human who just asks the same thing 5 times??? or does it treat both the same and stick to the policy???
@thamibenjelloun great question. Akash this side - one of the makers.
Fini actually does both - it identifies the personal agent and stick to the policy with same response every time. The identification helps support team when escalation happens and the context with which a human team can answer confidently.