Fin

Fin meetups are designed to help you stay up-to-speed with the latest AI best practices, and to connect with like-minded leaders and AI experts. Subscribe to our Luma calendar and join a future meetup in a city near you! luma.com/fin

Watch their full conversation here: youtu.be/AVVKO9cbRtw

Building an AI Agent for one company is straightforward. But making it work across every team, system, and dataset in any enterprise is where things get challenging. At our recent Fin x @Snowflake meetup in San Francisco, Jordan Neill (COO, Fin) and Baris Gultekin (VP of AI, Snowflake) broke down why context is the real bottleneck. As models improve, the challenge shifts to helping agents understand the nuances of each business (think data, language, and definitions) so they can operate reliably at scale. You can watch the full conversation at the link in the replies, and if you’d like to join a future Fin meetup, subscribe to our Luma calendar.

Learn more about Fin Procedures here: intercom.com/blog/the-harde…

Complex queries are a small percentage of your support volume, but they take up most of your team’s time, and leave the biggest impression on customers. AI Agents have automated simple queries, but as soon as processes become complex and multi-step, things get harder to build, test, and maintain. You're typically told you need a forward-deployed engineer and weeks of engagement just to get started. And then more waiting every time something needs to change. We built Fin Procedures so your team owns this end to end, from creation to testing to monitoring. And we're not stopping there. We've been continuously shipping improvements (for example: AI-powered procedure review, failure reporting, version history, and email simulation) with more on the way. Since launching, Procedures has handled over 1.5M conversations across industries, with a ~28.9% proven improvement in CSAT. And now Procedures is available to every Fin customer.

“This is now the highest performing, fastest model for customer service on the market.” We recently launched Fin Apex 1.0, our new flagship model, purpose-built for customer service. Apex is driving higher resolution rates, fewer hallucinations, and faster responses. Normally, you have to make tradeoffs between quality, performance, and latency. With Apex, we didn’t have to. In one case, a customer’s resolution rate jumped from 68% to 75% overnight. We’ve never seen improvements of this scale before. Apex is live for all @Fin_ai customers now. There’s nothing to turn on – you’re already on it.

The AI Group building Fin introduced a new form of attention: Low-Rank Key Value attention. It reduces KV cache memory by ~45–53% while improving model performance and training efficiency, by sharing a common KV base across heads with low-rank per-head specializations. Read all the details in this article from @JamesONeil21 : fin.ai/research/low-r…

Watch the full conversation here: youtu.be/-53JZG9-t20

Fin meetups are designed to help you stay up-to-speed with the latest AI best practices, and to connect with like-minded leaders and AI experts. Subscribe to our Luma calendar and join a future meetup in a city near you! luma.com/fin

On March 31 we had a full room of CX and product leaders and builders in London for our Fin x @linear meetup on delivering high-quality AI experiences. Together with @brian_donohue (VP Product, Fin) and @matthijswolting (Engineering, Linear), we explored what it takes to move beyond demos and build AI systems people actually trust in production. One theme came up again and again: quality isn’t just a metric. It’s a combination of signals including resolution rates, CSAT, and the instinct you build from reviewing real conversations and outputs over time. We also got into the trade-offs that come with shipping AI at scale. From faster responses reducing trust, to new features temporarily impacting performance  and why oversight has to be built in from the start, not added later. Thanks to everyone who joined, asked thoughtful questions, and pushed the conversation forward. If you’d like to join a future Fin meetup, subscribe to our Luma calendar to see what’s coming next.

Claude Code can now set up @Fin_ai and Intercom on your website – without ever leaving the command line. We built an agent-friendly CLI that Claude can use to get started, configure your setup, and ultimately even maintain Fin over time. The goal is simple: make deploying an AI customer service Agent as seamless as possible. Fin is the only self-serve AI agent for customer support, and it’s never been easier to get started. Watch @brian_donohue and @brian_scanlan (and @claudeai) set up Fin on a live website in under two minutes, using just a few commands in the CLI.

Over the past month we've delivered complete control over your conversation quality in Monitors, launched the best-performing model for customer service in Fin Apex 1.0, shipped Fin CLI for autonomous setup through your Agents, updates to Procedures, Helpdesk, and more. Here's the full rundown, with real demos and why each change matters for your team and your customers. → Fin Apex 1.0 (1:00) - Our first custom model, purpose-built for customer service. Apex 1.0 outperforms every frontier model on resolution, hallucination, and speed. → Monitors (3:10) - AI-powered QA that automatically reviews @Fin_ai 's conversations so you catch issues before they reach more customers. → Fin CLI (6:21) - Brian shows how AI coding Agents can now set Intercom and Fin up autonomously, from install to live. → Procedures (8:43) - Introducing updates including versioning, rollbacks, and failure reporting. → Fin over Discord (10:45) - Built specifically for how support works in Discord. Public answers, private threads, and control over what Fin answers (or doesn’t) → Intercom Helpdesk updates (12:21) - Workflow version history, changing ticket types, and WhatsApp voice note replies. Everything shown is ready to use – watch Brian D and Brian S show you what’s new:

LLMs decide what languages and libraries we use. They’ll also decide what products a business runs on. They'll prefer ones that work with them. Every business now needs a CLI. fin(.)ai/cli

Fin meetups are designed to help you stay up-to-speed with the latest AI best practices, and to connect with like-minded leaders and AI experts. Subscribe to our Luma calendar and join a future meetup in a city near you! lu.ma/fin

Watch the full conversation from Jordan and Baris here: youtu.be/AVVKO9cbRtw

“It’s always easy to build demos and POCs, but the real world is messy.” That tension defined our recent Fin x Snowflake meetup in San Francisco, where we explored what it really takes to deploy Agentic AI in the enterprise. Jordan Neill (COO, Fin) and Baris Gultekin (VP of AI, @Snowflake) dug into the “iceberg” behind production systems – why demos are easy, but real-world deployment depends on shared context, strong evaluation loops, and architectures that can handle messy, unpredictable environments. As Agents move from internal tools to enterprise-wide systems, reliability (not raw capability) is what makes the difference. The shift is already underway. The technology is here, but adoption now depends on trust, feedback loops, and organizations rethinking how work actually gets done. Take a look at the highlights below, and if you’d like to join a future Fin meetup, subscribe to our Luma calendar.

Fin meetups are designed to help you stay up-to-speed with the latest AI best practices, and to connect with like-minded leaders and AI experts. Subscribe to the Luma calendar below, and join a future meetup in a city near you. lu.ma/fin

Trust in AI isn't a feature. It's something you have to earn. That was the thread running through our conversation with @AnthropicAI at Fin Labs Paris, and it's harder than it sounds. Jordan Neill (COO, Fin) and Ryan O'Holleran (GTM Lead, EMEA, Anthropic) dug into the gap between what AI can do and what teams actually deploy. The technology isn't the bottleneck. Confidence is. When AI represents your brand, reliability, predictability, and knowing when not to answer matter as much as capability. Including the decision to say "I don't know." Want to be in the room for the next Fin meetup? Subscribe to our Luma calendar.

New on the Fin AI research blog, an in-depth write-up by Mariia Matskevichus on how the topic detection powering Fin Insights was built, combining the quality of LLMs for real dialog with the scalability and predictability of traditional ML clustering methods. Link below!