Back to vendors
F

Fini

Also known as: Fini AI, Sophie, Fini Inc.

Visit site
Entry price$0.89 per resolved ticket on Growth; $0.69 on Scale; $0.49 on EnterpriseFull pricing detail

Enterprise support agent for fintech and regulated sectors that resolves across voice, chat and email on one reasoning layer, takes real actions in billing and CRM systems, and maintains its own knowledge graph from resolved tickets.

Fini is an AI customer support agent from Fini Technologies Inc., backed by Y Combinator and Matrix Partners, built for fintech and other regulated sectors as well as ecommerce, software, health and fitness, gaming and insurance. One agent resolves tickets across chat, email and voice, with voice answered in real time, on shared reasoning, shared policies and a single audit trail, and it escalates to people when it should.

It processes refunds, reschedules bookings, updates accounts and checks billing and payments through actions into systems such as Stripe, and a worked example finds a duplicate charge in Stripe, refunds it and closes the ticket. Customers include Atlas, Unit, Wefunder, Column Tax, Bitdefender, Qogita, PostFinance and Swisscom, and Fini reports 3 million resolutions a month.

The platform is a five layer architecture. Guardrails check every query in real time, mask personal data and enforce company rules such as refund policy. An LLM Supervisor chooses between retrieving, reasoning, replying and acting. Skill modules search, write responses, follow internal processes such as refunds, cancellations and verifications, and call business systems.

A live feedback and evaluation layer checks every reply for accuracy, tone and safety before it is sent and learns from corrections, and a traceability layer records why each answer was given. Knowledge Atlas takes in the knowledge base and past tickets as a graph, finds gaps in resolved tickets and writes the missing articles itself.

Around the agent sit newer pieces. Agent Handshake answers the personal AI agents that customers now send to support, checking identity against live account data and sending refunds, credits and exceptions to an owner for approval. Scout is an AI coworker in Slack that reads tickets, reviews and social threads, answers questions with sources, drafts articles and bug reports for a named approver, and tests the support agent. Fini layers onto Zendesk, Intercom, Salesforce, HubSpot, Front, Freshdesk, Gorgias, LiveChat, Zoho Desk and ServiceNow, plus Slack, Discord, Shopify and Stripe, in more than 130 languages.

Fini holds SOC 2 Type II and ISO 27001 and states GDPR, with PCI DSS Level 1 and HIPAA shown as badges, and its trust portal names OpenAI as its model provider on Google Cloud. Pricing is per resolved ticket with no seat fees or usage minimums: Growth at $0.89 for up to 3,000 tickets a month, Scale at $0.69 for up to 10,000, and Enterprise from $0.49 with unlimited tickets, a 90 day free pilot, data residency options and admin and audit logs. Voice is $0.89 a call for the first 10,000 calls a month, and new enterprise customers pay nothing if Fini does not reach 90 percent resolution in 90 days.

Vendor details

Canonical URL

https://www.usefini.com

Category

Customer support agent

Subcategory

Support — AI agent

Funding status

Venture backed; Fini Technologies Inc. is backed by Y Combinator and Matrix Partners, and its site gives no round sizes.

Company status

independent

Use cases & customers

Primary use cases

customer supportrefundsKYCplan changes

Target customers

support teamsSMB

Deployment options

SaaS

In practice

Your support team spends its day on refunds, KYC checks, and plan changes. Fini automates those actions directly, so routine requests resolve without a rep.

A customer wants a refund now, not a ticket number. Fini's agent processes the refund itself, closing the loop in the conversation.

Account changes like a plan upgrade bounce between support and billing. Fini handles plan changes as an action, so the customer's request gets done in one place.

Agentic Index coverage score

9.0 / 14 capabilities · 64%

Integrations & Tool Calling Full

Setup on an existing helpdesk takes one click through OAuth, with listings verified in the marketplaces of Zendesk, Intercom, HubSpot, Front and Gorgias, plus Salesforce, Freshdesk, LiveChat, Slack and Discord. Fini connects backend billing, CRM and internal systems so the agent takes real actions, and a worked example calls Stripe customer search, charge listing and refund creation before closing the ticket.

Sourceusefini.com/product/integrationsread 2026-09-05

Workflow Orchestration Full

A named orchestration layer decides the path and carries it out. The LLM Supervisor is layer two of the architecture and chooses the best path between retrieve, reason, reply and act, using conversation context, custom rules and tone instructions, and access to enterprise systems including CRM and order management. Beneath it, work routes to specialized skill modules.

Search and Retrieve finds information, Write Response drafts replies, Follow Process runs internal flows such as refunds, cancellations and verifications with automated rule enforcement, and Take Action triggers API calls into business systems with no human handoff. Fini's worked examples show the chain end to end, checking a refund policy, verifying a delivery timeline by API and starting the refund in one pass.

Sourceusefini.com/product/platformread 2026-09-05

Knowledge Grounding & RAG Full

Knowledge Atlas is a named knowledge structure Fini maintains. It takes in the customer's knowledge base and historical tickets as a graph, scans resolved tickets to find gaps against that graph, writes the missing article and commits the update, so the retrieval structure persists and repairs itself instead of being assembled for each conversation. Answers are grounded in it, with the reasoning traced back to approved internal knowledge.

Sourceusefini.com/knowledge-atlasread 2026-09-05

Human Oversight & Guardrails Full

Guardrails are layer one of a published five layer architecture, sitting ahead of the action instead of reviewing after it. A real time safety check blocks harmful, off topic and policy violating content, masks personally identifiable information such as phone numbers and emails, and enforces company rules including refund policy, so only safe, valid queries move forward.

Inside the response module every generated reply is checked for accuracy, tone and safety before it is sent, which Fini describes as every answer being reviewed by an LLM, not only generated by one, and the Follow Process module enforces policy automatically on refunds, cancellations and verifications.

A published case study names four checks, Polite Language, Competitor, Topic escalation for complex cases such as refunds, and Truth Check verifying accuracy before responses reach customers, with output rails using LLM judges to flag, correct or escalate to a person.

Sourceusefini.com/product/platformread 2026-09-05

Security, Identity & Governance Full

A SafeBase trust portal publishes GDPR, ISO/IEC 27001 and SOC 2, with the SOC 2 report, ISO certificate, penetration test report, DPA and cyber insurance available on request, along with named product controls for audit logging, access monitoring, data erasure, credential management and password security, and a subprocessor list. PCI DSS Level 1 and HIPAA with BAA appear as homepage badges but are missing from the trust portal's certification list, and ISO 42001 appears on no Fini page.

Sourcesecurity.usefini.comread 2026-09-05

Observability & Auditability Full

Voice, chat and email run on one reasoning layer sharing the same policies and a single audit trail. Every answer is classified, scored for confidence and checked against policy, with the result written to an audit log, so the policy behind a given reply can be recovered, and the trust portal lists audit logging and access monitoring as product controls. Fini does not say whether or in what format the audit log can be exported.

Sourceusefini.comread 2026-09-05

Memory & State Persistence Not documented

The platform has five named layers, Guardrails, LLM Supervisor, Knowledge and Action Modules, Live Feedback and Evaluation, and a Traceability Layer. None is a memory layer, and the only state named is conversation context feeding the LLM Supervisor within a request, held for the session, not across sessions.

Past tickets are taken in as a knowledge source, and the adaptive learning loop trains on real interactions, which is learning absorbed into a model, not state an agent keeps. A Fini Labs blog post mentions memory that respects policy, but the architecture lists its components without it, and the documentation at docs.usefini.com is gated.

Sourceusefini.com/product/platformread 2026-09-05

Deployment & Data Residency Partial

Fini runs as a hosted service on Google Cloud Platform, as its trust portal names, with backups and disaster recovery covered. The Enterprise plan lists data residency options, but no region list, selection surface, VPC, single tenant or on premises option is named, and Scout, the Slack coworker, states EU data storage in Sweden for its own workspaces. So residency is offered as an option on the top plan without the mechanism shown.

Sourceusefini.com/pricingread 2026-11-19

Prebuilt Agents, Templates & Packs Partial

Fini sells one support agent, ready for production from day one, that learns from docs, tickets and the knowledge base without manual tagging, plus Scout, a separate coworker in Slack, and Agent Handshake, a capability of the agent for customers' own AI agents. Sophie, the prebuilt agent once named for refunds, duplicate charges and cancellations, no longer appears on the site. There is no template library, agent catalog or marketplace, and the skill modules the agent routes to are internal architecture rather than assets a customer selects.

Sourceusefini.comread 2026-11-19

Triggers & Channel Coverage Full

One agent runs across voice, chat and email on the same reasoning and policies, with real time voice under 100 milliseconds of latency handling bookings end to end, and intake arriving through the connected helpdesk, Slack or Discord, or from customers' own AI agents through Agent Handshake. API and webhook integrations are listed, and Scout watches tickets and feedback for spikes and raises them in Slack. More than 130 languages are supported.

Sourceusefini.comread 2026-11-19

Model Flexibility & Routing Not documented

One model provider is disclosed, and customers have no choice of model. The trust portal's subprocessor list names OpenAI as the only model provider, with Supabase, AWS, Microsoft Azure and Google Cloud as infrastructure rather than models. Model version control is listed as an Enterprise extra, which pins versions rather than choosing between them, and there is no model selection, routing or bring your own model option.

Sourceusefini.com/pricingread 2026-11-19

APIs, SDKs & MCP Extensibility Partial

An API Library and docs at docs.usefini.com cover the platform, with an authentication page under the API reference, but the documentation sits behind an access code, so its scope is not readable. Agent Handshake lets customers' own AI agents, such as Gemini, contact support and request refunds or exceptions, which is an outside caller driving the agent, but no endpoint, authentication scheme or agent card is published for it. No SDK or MCP server is offered.

Sourceusefini.com/agent-handshakeread 2026-11-19

Testing, Debugging & Optimization Partial

Fini runs a live optimization loop but no gate before release. Live Feedback and Evaluation, layer four of the architecture, validates outputs automatically, detects errors and contradictions and learns from corrections, and every response is reviewed by an LLM for accuracy, tone and safety before sending. Scout tests the support agent from Slack, but how those tests are scored or compared is not described, and there is no test environment, simulation harness or scored evaluation of a proposed change before it reaches customers. The published accuracy and resolution figures are Fini's own.

Sourceusefini.com/product/platformread 2026-11-19

Browser & Computer Use Not documented

There is no browser control, hosted session or computer use. The Take Action module triggers API calls into business systems, with Stripe and Zendesk named, and the Follow Process module runs internal flows such as refunds, cancellations and verifications through automated rule enforcement rather than by operating an interface. Scout reads tickets, reviews and social threads but changes nothing without a named approver.

Sourceusefini.com/product/platformread 2026-11-19

The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded

Product updates

·MCP / tool calling / APIPartially Verified

Fini launched Fini MCP, a hosted MCP server at mcp.usefini.com that connects a Fini workspace to ChatGPT, Gemini and Cursor as a custom connector, to Claude through the Fini plugin, and to Claude Code, Codex and the terminal through the fini-skills package. From a chat, teams can build a support agent from help center content, tickets and API docs, turn endpoints into actions, replay past conversations as tests, pull reports and set plain language monitoring alerts. Reads run freely, while knowledge, prompt, rule and tag changes land as drafts in Review and wait for a person to approve them.

Bears on: MCP / tool calling / API

View source
·Agent capabilityVerified

Agent Handshake handles support conversations where the other side is a customer's personal AI agent. Fini flags likely agents from message patterns and request speed, applies the same rule and account data however a request is reworded, checks every reply against Live Guardrails, and stops once an answer is final by escalating to a person or sending the agent back to its owner. Teams define which actions an agent may request, and anything that moves money can go to a human.

Bears on: Human approval / guardrails

View source
View all 2 Fini updates →Tracked since Oct 2026 · Verified from public vendor sources

Pricing

$0.89 per resolved ticket on Growth; $0.69 on Scale; $0.49 on Enterprise

Per resolved ticket, with a monthly allowance per plan

Free tier

Included quota

Unused allowance rolls forward one month. Conversations reopened within 72 hours are not billed twice, and greetings, spam and abandoned sessions are never billable.

What is public

Fini prices by resolution: a fee is charged only when its agent solves an issue end to end with no human handover, and escalations to the human team are free. Each plan covers the platform, implementation and a monthly ticket allowance, with no per seat fees. Growth is $0.89 per resolved ticket for up to 3,000 tickets a month, with Voice AI on the standard rate card, shared customer success and priority email support. Scale is $0.69 for up to 10,000, adding white glove onboarding, a dedicated AI engineer, a nightly Knowledge Atlas scan, a Slack agent with SSO and multi language support. Enterprise is $0.49 with unlimited tickets, adding a dedicated engineer pod with a 15 minute SLA, an uptime SLA, premium support, a custom helpdesk integration, admin and audit logs, data residency with a BAA and DPA, a switching cost credit of up to $50,000 and no price increases. Voice AI on every plan costs $0.89 per answered call for the first 10,000 a month, $0.59 up to 50,000 and $0.35 beyond, or $0.22, $0.18 and $0.14 a minute.

Billing mechanics

Plans bill in advance, monthly, quarterly or annually, and annual billing includes two months free. Usage above the allowance bills monthly in arrears. Prices are in USD, with contracts in EUR, GBP and AUD at the rate at signing. A flat per conversation rate is available on request, and the metric can be switched at renewal. Two and three year terms take 15% and 20% off on Growth and Scale and 20% and 25% on Enterprise, and upfront payment takes another 5%, or 10% on Enterprise.

Cost watchouts

Every resolved ticket bills, so spend rises as the resolution rate improves.

Variable cost rationale

Spend grows with ticket volume and with the share of tickets the agent closes. Higher plans lower the rate per resolution.

Overage / add-ons

Above the allowance each extra resolution bills at the plan's own rate, $0.89, $0.69 or $0.49, or per conversation at $0.69 on Growth and $0.49 on Scale, with no penalty rate and no minimum.

Sales call required

Mixed (some tiers require a call)

Free / trial

There is no free plan. Enterprise customers can run Fini free for 90 days on live traffic before the plan begins, against agreed resolution, CSAT and accuracy targets. Under the 90 Day Money Back Guarantee, eligible new enterprise customers pay nothing if Fini does not reach 90% resolution in 90 days.

Commercial notes

Fini is backed by Y Combinator and Matrix Partners. Its agent takes actions such as refunds through Stripe or Adyen, and Fini cites 98% accuracy and 2M+ queries processed.

Key ambiguities

Enterprise commitment discounts are 20% and 25% in one place and 15% and 25% in another.

Support SLA / resale

Compliance is included on every plan: SOC 2 Type II, ISO 27001, HIPAA compliant and BAA eligible handling, GDPR and CCPA. Customer data is never used to train foundation models.

Missing data

Fini does not state Enterprise overage terms or cancellation and refund terms beyond the guarantee.

Agentic Index verified 2026-11-16

Fini pricing, plans and cost →

Alternatives to Fini

Other products that do the work Fini does, and what each one is.

Fini alternatives and competitors →

Contact us

Found a vendor we missed? Have feedback on the index? We'd love to hear from you.