This is the story of how a simple Slack chatbot evolved into a comprehensive AI platform that's saving our teams thousands of work days annually - and how we did it while keeping costs under control and security tight.

Nearly two years ago, in May 2023, we took our first step into enterprise AI with ChatITP, creating a Slack-based AI assistant. What began as an experiment in secure AI implementation has transformed into something far more significant. The idea of our own AI assistant was born from our commitment to explore, enable, and safely implement generative AI across our organisation. What started as a simple chatbot processing 4,000 requests in its first month has evolved into something far more ambitious: a comprehensive AI platform that's now the third most-used application in our company.

The ChatITP Evolution 

In the early days, ChatITP served primarily as a wrapper for various APIs and models, providing our team with a secure alternative to ChatGPT through Slack's familiar interface. While this approach successfully lowered adoption barriers, we quickly realised its limitations. As Glenn Bostoen, our Solution Architect and the driving force behind ChatITP 2.0, explains: "I started working extensively with AI coding assistants and saw immediate value, but our existing Slack-based solution felt restrictive. It was clear we needed something more flexible that could support our AI ambitions."

The team's vision went beyond just creating another chatbot. The goal was to build a secure, scalable platform that would democratise AI access within the company while maintaining strict control over data and costs. This meant creating an environment where employees could freely experiment with AI capabilities, integrate them into their workflows, and utilise multiple AI models for different use cases, all while ensuring enterprise-grade security and cost management.

Today's ChatITP represents a significant evolution in our approach to AI integration. Built on Open WebUI  and hosted securely on our Google Cloud infrastructure, the platform breaks free from the limitations of single-model solutions. Instead of restricting employees to one AI provider, ChatITP offers access to a carefully curated collection of industry-leading models, including OpenAI GPT, Anthropic's Claude, Google's Gemini, and Mistral’s models.

Making AI Accessible and Affordable

What sets this solution apart is its innovative approach to cost management. With individual licences running hundreds of dollars per user per year, scaling AI access across our company seemed financially impossible. We've implemented a pay-per-use model with careful budget controls. This approach democratises access to a diverse set of  AI capabilities while preventing runaway costs and model lock-in, making it possible for teams across the organisation to experiment and innovate without concern about individual licence fees.

The growth since our initial launch has been remarkable. ChatITP has become an integral part of our digital toolkit, with 62% of ITP colleagues now incorporating it into their daily workflow. The tool processes an impressive 4,500 requests weekly, having handled over 150,000 messages to date. This adoption has positioned ChatITP as our third most utilised productivity tool, ranking just behind Google Workspace and our password manager.

Empowering AI Mindset across Teams

The platform's versatility has enabled diverse applications across different departments. Technical teams are using the platform for advanced code generation and review, taking advantage of features like in-browser Python script execution and real-time code analysis. Design and architecture teams benefit from the platform's ability to generate and render diagrams through Mermaid, while content creators and strategists use it for everything from writing assistance to brand consistency checks.

One of the platform's most powerful features is its integration with our internal knowledge base. By connecting with our Confluence, ChatITP can provide contextually relevant information from our company documentation, making it easier for teams to access and utilise our collective knowledge.

While ChatITP began as an internal innovation, its potential extends far beyond our own walls. We're increasingly seeing opportunities to adapt this solution for clients who face similar challenges in their AI adoption journey. Many organisations want to experiment with AI but find themselves held back by a combination of licensing costs, security concerns, and implementation complexities.

"We've done the hard work of setting up a secure, scalable platform," notes Bostoen. "This could be a powerful solution for clients who want to democratise AI access while maintaining control." Our experience building and scaling ChatITP can help other organisations navigate their own AI transformation journeys.

Start with Curiosity, Build with Purpose

ChatITP represents more than just a technical achievement, it's a testament to our belief in empowering teams through thoughtful innovation. As we continue expanding its capabilities, we're focusing on several key areas of development. We're exploring deeper integration with internal tools and workflows, enhancing our knowledge base capabilities to make information more accessible and actionable, and developing automated development workflows that can streamline complex processes.

We're also working on expanding our multi-model support, ensuring teams have access to the best AI capabilities for their specific needs. This includes careful evaluation of new models and providers as they emerge, always with an eye toward maintaining the perfect balance between innovation and reliability.

From its humble beginnings as a Slack bot to its current status as a core enterprise platform, ChatITP exemplifies our approach to innovation: start with curiosity, build with purpose, and always keep humans at the centre of technological advancement. 

‍