Claire

AI middleware & assistant platform

Client

Serapion owned Product

Team size

6 - 7

Time frame

2 years

Technology used

The History of Product

CLAIRE STUDIO & Serapion

Claire Studio was born at Serapion when we realized early on that language models - GPT-3.5-Turbo and GPT-4 at the time - were capable of utilizing tools that they were given, even though this was not yet natively supported at that time. From that discovery, we built up a fully featured platform that abstracted away a lot of the complexity and repetitive work that was necessary to build up an AI assistant, from LLM vendor integration independence, context management and tool integrations, over to branding and deployment.

The Vision

WHERE IT ALL BEGINS

We built Claire Studio with one idea in mind: Make it as simple as possible for businesses to create, deploy and subsequently monitor an AI assistant. AI should be approachable to our customers, with little to no need for technical knowledge or any own developer resources.

By combining the creation of chatbots, with pre-built no-code integrations to manage knowledge bases and other common tools, as well as fine-grained controls for branding, we are able to offer an end-to-end experience from creation to production.

The Challenge

UNTAGLING THE KNOT

While platforms to build conversational experiences have been available on the market for a long time, they either focused on classical, intent-based systems with no world knowledge or context awareness or did not enable brands to roll out experiences they control.

Key challenges included:

Setting up an agent harness that maintains context & integrates with a wide variety of tools.
Developing a system to dynamically spin up infrastructure for integrations, like vector stores for retrieval-augmented generation.
Design of a custom branding solution that allows a chatbot to be branded and styled in any way our customers intends
Create a solution that enables our customers to monitor and control AI assistants in production, including alerting, budgeting and cost reporting.

The Solution

HOW WE MADE IT WORK

The solution relies on an international collaboration between different locations from Serapion, product ownership and strategy. While development was ongoing, we were already envisioning not only how to adapt Claire to ongoing market developments and advances in research, but also to make the product as flexible as possible to suit our customers' use cases, while keeping the system intuitive to use.

Staying Ahead

Introduction of features that utilize emergent capabilities of new AI models
Support for multiple AI vendors to prevent lock-in, enable state-of-the-art models, and meet regulatory requirements
Claire Studio functions as a harness for increasingly powerful agents acting on behalf of end users.

Customer Experience

Experience designed for non-technical users with minimal knowledge on the operation of AI systems
No-code deployment of extensions, including knowledge bases, web crawling, general-purpose integrations, and other commonly used integrations
Strategic consulting to help customers deploy AI solutions and optimize workflows

The outcome was a stable and productionized self-service application that not only supported companies to rapidly explore generative AI use cases, but also to build industry-leading assistants in various domains, including finance, legal and education.

The Value

IMPACT THAT MATTERS

We demonstrated measurable value by enabling non-technical staff to rapidly explore and validate generative AI use-cases. Our customers provided measurable benefit to their users through intelligent software solutions, and obtained actionable insights through reporting features.

Benefits delivered:

Non-technical staff can build and try out AI assistants in minutes
Large amount of extra knowledge can be made accessible with no technical know-how
Integration into existing customer systems, from code snippets to standalone applications, with authentication through company identity providers
Access to LLMs across jurisdictions, including EU-only processing & data residency, with insights and controls for scaling AI assistants confidently

Skills That Stand Out

CORE STRENGTHS

Product Thinking

With full ownership over the system and its evolution, Serapion adapted and evolved a product in a rapidly evolving market

Domain Expertise

Our extensive experience across customer projects, prototypes, and academic research enabled us to provide a framework for deploying AI assistants with confidence.

Embracing Technology

Serapion built a platform incorporating state-of-the-art AI advancements, while using said advancements in the engineering process itself in a way that kept engineers in control of software quality.

Flexible Architecture

The system has been built with extensibility in mind. This enables new add-ons to be developed and deployed independent of the core system.