Meet Fin, the Customer Agent That Delivers Perfect Experiences

For most consumers, poor customer experiences are still the default. Customers are subjected to long wait times, handoffs that lose context, and a transactional experience. Even the very best businesses can’t provide excellent customer experiences consistently, because the economics are hard, and humans are innately inconsistent.

For the first time, we have a technology in AI that makes perfect experiences possible, for all customers of every business, at all times, at all customer journey stages, and in all channels.

Agents can do many of the time-consuming tasks that humans do, often better than humans can, making new experiences possible. They can shrink a support process that might have taken days or weeks so it delivers a great outcome instantly. They can deeply understand a customer’s entire relationship with your company from day zero to year ten and bring those insights to every touchpoint.

Agents can do this because they are becoming:

• More autonomous, completing work in the background while people focus elsewhere. 
• More proactive, acting on signals before they’re even asked to. 
• More directive, increasingly knowing what should happen next and guiding people on what to do.  

At Fin, our mission is to bring perfect customer experiences to everyone. Today over 12,000 customers including Anthropic use Fin to transform their customer experience. 

The highest-performing Agent for customer service

Fin resolves 76% of customer queries on average. It uses your business knowledge and context and works across your systems to resolve complex queries end to end, from billing disputes and technical troubleshooting to refunds and appointment bookings. Because Fin understands your business goals and customers, it can go beyond service and operate across the entire customer lifecycle. It can help new customers get set up and proactively reach out if someone gets stuck in your product. 

Fin delivers that same level of service wherever your customers are, across Messenger, Voice, Slack, WhatsApp, email, and more.

Fin is also fast to deploy, so you can start seeing value quickly without limiting how far you can take it. As you give Fin more knowledge, context, and access to your systems, it can take on increasingly complex work. A number of Fin customers are already achieving resolution rates of 90%.

And all of that performance comes in a product your customer service team can run themselves.

Pioneering companies run their customer experience on Fin

Along with Anthropic, pioneering companies like Kalshi, Solidcore, and Rocket Money use Fin to solve customer problems. Anthropic’s team chose to implement Fin instead of building their own Agent in-house. 

Today, Fin is involved in 80% of Anthropic’s inbound queries, and, at a 79% resolution rate, is handling 63% of their total support volume end to end. 

The impact goes beyond the metrics. By taking on more of the customer service workload, Fin has freed Anthropic’s specialists to focus on higher-value issues and the work where human judgment matters most.

“Our team is now able to invest in skills they wouldn’t otherwise have time for when they were firefighting large volumes,” said Larrow. “Now they can get deeper with customers and cross-functional teams, and really build out what a successful support career looks like.”

Two beliefs shape what makes Fin different

Industry-leading performance comes from specialized AI

We believe that ultimately, customer trust will be won or lost on AI performance, and the best performance will come from vertically specialized AI. This shapes the technology behind Fin.

That matters most as the work gets harder: when an Agent needs to understand the customer’s situation, follow the right policy, use customer data, and take several actions to reach the right outcome.

You should be able to trust and fully control your Agent

We believe businesses need total control over their Agent system and must be able to trust it to do the right things, because the Agent will be their actual customer experience. This shapes how customers experience Fin as a product.

A powerful Agent shouldn’t require months of implementation or a team of technical specialists to deliver value. It should be fast to get live and easy for customer service teams to run themselves. They should be able to understand what it is doing and keep improving it as the business changes.

Accuracy and reliability, even when the work gets complex

For AI to meaningfully change the economics of customer service, it has to do more than handle simple questions. It needs to perform well enough to take complex, time-consuming work off the human team, without compromising the experience customers receive. Fin is built for that work.

If a customer reaches out about a damaged order, Fin can locate the order in your system, check an image the customer sends, decide whether a refund should be issued, ask for human approval if the amount exceeds a limit you set, and then keep the conversation moving through to resolution.

If a customer loses a credit card, Fin can freeze the card and arrange a replacement rather than handing the conversation to your team and leaving them to complete the work.

Fin’s performance comes from technology built specifically for customer experience. At its core is Fin Apex, a family of proprietary models that perform best for this work. These models are trained, tested and evaluated on what matters most: using a business’s knowledge accurately, following its policies, understanding the customer’s situation, and recognizing when more information is needed.

The first generation of Fin Apex models outperformed leading frontier models on resolution rate, hallucinations, and speed. Apex 2.0, builds on that foundation with 8.6% more confirmed resolutions (cases where the customer explicitly confirms their problem was solved) and 16.5% fewer hallucinations than Apex 1.0.

Fin’s AI team continuously tests and improves these models, and every improvement becomes part of the product, so every customer benefits automatically, without additional tuning.

That is what helps Fin resolve even the most complex customer issues at enormous scale. Fin resolves more than 10 million conversations a month, and handles a complex, multi-step query more than once every second.

A powerful Agent that delivers results in days

High-performing AI often comes with a long implementation project, outside consultants, and a long wait before the business sees a return.Fin doesn’t work that way.

Most teams go live in days and start resolving customer issues straight away. Kalshi came to Fin three weeks before the Super Bowl, knowing record support volume was coming. On Super Bowl day, Kalshi handled 80,000 conversations in 24 hours.

But going live quickly doesn’t put a ceiling on performance. As you give Fin more knowledge, context, and access to your systems, it can take on increasingly complex work. Today, Fin resolves 80% of Kalshi’s support conversations.

Fin went live on Salesforce’s own help site, a very high volume site with hundreds of products that operates in almost every language, in just 15 working days and in a few weeks was resolving 65% of all queries it sees autonomously with a stellar customer experience score of 76%. 

Built for your team to manage, control, and improve

Getting Fin live is only the start. High-performing AI takes continuous iteration, and Fin puts that work in the hands of the people who know your customers best: your support team.

Fin is designed to be easy for support teams to manage themselves. You can configure its knowledge, behavior, and processes without engineers, and Fin also gives you a continuous loop to train, test, deploy, and analyze its performance.

That ability to run Fin yourself comes with complete visibility and control. You define how Fin behaves, which actions it can take, and when your team should be involved. Before anything reaches customers, you can test it thoroughly. You can evaluate Fin against real customer scenarios before changes go live and have control over how and when those changes roll out.

Once Fin is live, you can evaluate conversations against your standards, with Insights and Fin’s CX Score showing where there are opportunities to improve the customer experience. 

Fin’s decisions and actions are traceable too, so you can understand what happened and why.

That work all adds up. Maintaining a high-performing Agent takes ongoing analysis, updates, and optimization, and managing all of that manually doesn’t scale. It should be done by an Agent too, which is why Fin has an Agent that runs your customer operations – Operator. 

Operator manages and optimizes Fin, so your team can focus on setting the strategy and designing the customer experience. It understands what is happening across your support operation, spots problems, finds opportunities to improve Fin’s performance, and acts on them.

Say escalations suddenly increase; instead of your team digging through conversations to work out why, they can ask Operator. It investigates what is holding Fin back, recommends what should change, and can build and test the fix itself before bringing it to your team for approval.

Fin and your team collaborate with shared context and insights

Even as Fin takes on more of the customer experience, there will still be moments where human judgment is needed. Some Fin customers are already resolving 90% of their conversations with Fin, which means the 10% that’s left for the human team is the most complex and high-impact work. 

The role of the support rep is changing from answering questions to working alongside and supervising Fin. What matters is that when a rep does step in to help, it never breaks the experience.

Fin can ask your team for information or approval, then carry on resolving the issue. A moment that requires human judgment doesn’t have to become a full handoff.

When someone does need to take over, they can pick up in Service Cloud with the transcript, a summary of what has happened, and the record updates Fin has already made. Fin uses Contact, Account, and Case information to understand each customer’s situation and keeps those records current as it works.

Shared insights also show you how Fin and your team perform together, so you can see service quality and performance across the whole operation. 

All of that means your customers can get help, and a consistent experience, whether they’re talking to Fin or your team.

Meeting customers where they are

Every channel comes with different expectations, and delivering the same quality bar across all of them can add complexity and cost. 

Fin adapts how it communicates on each channel so customers get high-quality, personalized service however they choose to reach you.

• On chat, WhatsApp, SMS, and email, Fin delivers instant, on-brand conversations.
• On Voice, Fin speaks naturally, understands pauses and interruptions, and responds quickly. It runs on Fin Apex Flash, a model built specifically for low-latency voice conversations, and can take action to resolve complex issues during the call.
• In Slack, Fin resolves customer questions directly in Slack communities and Connect channels, with threaded replies that work the way Slack users expect.

Keeping pace with the Frontier

AI is advancing at a breakneck pace. At Fin we are passionate about bringing our customers safely into this new world, both through our speed of innovation and education on the transformation they will need to go through internally. 

AI Agents are becoming more autonomous, proactive and directive, and so is Fin. 

• More autonomous, completing work in the background while people focus elsewhere. 
• More proactive, acting on signals and doing things before they’re asked to. 
• More directive, increasingly knowing what should happen next and guiding people on what to do.  

As of October, 416 new capabilities shipped this year alone, expanding what Fin can resolve and making it easier for teams to run and improve.

• Operator is now proactive with Routines and Connections, working across the tools your team already uses, including Slack and GitHub. 
• Evals and Releases unlock the testing and rollout practices of AI teams. 
• Skills let Fin resolve problems that don’t follow a set path to resolution, working out what to do as the conversation unfolds. 
• Fin Apex 2.0 and Fin Apex Flash 2.0, the next generation of our Fin Apex models deliver with higher resolution, better accuracy, and faster responses vs. the previous generation. 

These advancements pave a clear path to a fully agentic customer experience, where the vast majority of customer conversations are resolved by an Agent, and businesses can deliver excellent customer experiences at any scale.

Explore the latest Fin updates →

See Fin in action

Fin serves your customers. Operator helps run and improve Fin. Your team stays in control and focused on the moments where people add the most value.

See how quickly Fin can start resolving your most complex customer issues and delivering perfect experiences at scale.

To see Fin in action, check out this demo.