A schematic diagram of the Fin AI Engine
The AI models purpose-built for customer service

Meet thefin-cx suite of models

The world’s best models for CX

The fin-cx suite of models have been purpose-built to handle the scale and complexity of customer service. [1]

Based on deep domain expertise, these models deliver unmatched performance and a customer experience you can trust.

Introducing the

first fin-cx models

Retrieve relevant content

fin-cx-retrieval model

The fin-cx-retrieval model scans a vast body of knowledge (help center articles, snippets, etc) and pulls out a small set of useful information.

Specification

  • 1.1 Understand intent behind the question
  • 1.2 Search across all available content sources
  • 1.3 Match on semantics, not just keywords
  • 1.4 Select top N most likely candidates

Rerank for precision

fin-cx-reranker model

The fin-cx-reranker model takes the retrieved content and scores each piece for relevance, accuracy, and usefulness in context. It then selects the final piece(s) for the LLM to use.

Specification

  • 2.1 Score relevance with respect to query
  • 2.2 Evaluate context match and resolution fit
  • 2.3 Downrank outdated or low-confidence sources
  • 2.4 Output: final selected content for generation

AI Team

Built by a world-class team of AI experts

01Building models of this quality is only possible thanks to Intercom's 50-person-strong, world-class AI group, led by Fergal Reid, and our decade-long experience in building customer service software.

AI Group Leadership
Avatar of Fergal Reid
Fergal ReidChief AI Officer
Avatar of Brian McDonnell
Brian McDonnellDirector, Engineering
Avatar of Mario Kostelac
Mario KostelacPrincipal Machine Learning Engineer
Avatar of Alexey Tarasov
Alexey TarasovSenior Manager, ML Engineering
Avatar of Fedor Parfenov
Fedor ParfenovStaff Machine Learning Scientist
Avatar of Pedro Tabacof
Pedro TabacofPrincipal Machine Learning Scientist
Avatar of Molly Mahar
Molly MaharPrincipal AI Designer
Avatar of Pratik Bothra
Pratik BothraPrincipal Machine Learning Engineer
Avatar of Rati Zivarwa
Rati ZivarwaDirector, Product Management
Avatar of Rob Clancy
Rob ClancyStaff Product Engineer

Performance

Widening the performance gap between Fin and the rest

02In production testing, the fin-cx models outperform every other model combination we've tested. We use them alongside Claude Sonnet 4.0 in production to deliver best-in-class accuracy and performance.

Performance uplift validated through out-of-sample backtesting and large-scale A/B testing in live environments.

Fig 2.A - Powered by fin-cx models
A schematic diagram of the Fin AI Engine
Fig 2.B - Fin wins every head-to-head test on resolution rate

Resolution rate based on independent testing conducted by Fin customers.

Fig 2.C - Fin's average resolution rate increases 1% every month
Fig 2.D - Fin Customer
A headshot of Angelo Livanos, Senior Director of Global Support at Lightspeed

“Fin is in a completely different league. It’s now involved in 99% of conversations and successfully resolves up to 65% end-to-end—even the more complex ones.”

Angelo Livanos,Senior Director of Global Support at Lightspeed

Our Guarantee

Performance backed by our $1 million guarantee

03Fin is the highest performing AI Agent for customer service, in part because of fin-cx models. But we know others claim to be the best also, which is why we back our claims up with the Fin $1million guarantee.

Learn more

Security

Customer-first approach to data privacy and security

Schematics of the Fin AI Engine

04We have invested heavily in data privacy and security. These models are trained on deidentified data from past Fin interactions.

The following existing customers are automatically excluded from our training process:

  • Customers that signed an MSSA
  • Customers that have a BAA in place with us
  • Customer using an EU/AU regionally hosted workspace
  • Any customer that was previously granted an opt-out