Over the past year, my team developed and scaled a multi-model, multi-modal Generative AI framework for business, which has been used to power a wide range of use cases, primarily accessed via a natural language chatbot.

I have documented this journey via a series of articles (links below).

In my last article, I shared six diverse business use cases and their outcomes, which contribute towards the wider return on investment, primarily focused on productivity acceleration and data insights.

Earlier today, the team launched the next major revision of the Generative AI framework, including a wide range of new features, as well as a new user experience for the natural language interface.

Outlined below is a summary summary of the new features.

  1. A refined and simplified user experience that includes user personalisation, smart recommendations (tooltips), and light/dark modes.

  2. The ability to start and maintain multiple conversation threads, each targeting a specific persona, model and context.

  3. New multimodal models that can now handle up to 2,000 pages of text or 100 minutes of video content. Text and video content can be linked or uploaded directly.

  4. New interactive user interface elements, including code blocks, tables and text, which can be easily exported, automatically selecting the appropriate file format (e.g., CSV for tables, etc.)

  5. Access to the latest and greatest foundational models from OpenAI and Google with no restrictions. This includes the Imagen v3 for advanced image generation.

  6. New APIs for developers, allowing the Generative AI framework to be programmatically integrated with any system.

  7. New native integrations, covering systems of record, such as SAP and Microsoft SharePoint.

  8. New getting started information, providing education material, frequently asked questions and examples for common use cases.

The image below highlights the new user experience (dark mode), including the customisable chat pane, tooltips and improved discoverability of custom personas.

Generative AI

Our custom persona capability (similar to OpenAI GPTs) also received a major upgrade.

Custom personas allow any user to create and publish their own Generative AI chatbot (no technical skillset required), leveraging the underlying framework. These custom personas can be published to individuals, teams or the entire company, secured (where needed) by modern authentication/authorisation (FIDO2 compliant).

A custom person can target any available model (e.g., OpenAI, Google) and include specific grounding and context, including large datasets from structured or unstructured data sources.

In the latest release, custom persona creation has been dramatically simplified via a step-by-step wizard that can be previewed in real time.

Generative AI

This new approach helps ensure every custom persona is configured correctly with optimal inputs to improve the quality and consistency of the output. It also includes native integrations to securely and privately ingest large datasets from systems of record, such as Microsoft SharePoint.

Finally, where appropriate, the custom personas can be posted to the marketplace, promoting discoverability, collaboration and reuse, which further increases the return on investment.

Generative AI

At the time of writing, 773 active custom persons, covering everything from business-specific expert systems to general productivity accelerators for common tasks such as knowledge sharing, summarisation, translations, etc.

Overall, our Generative AI framework continues to gain momentum, with more than 6,700 unique users (95% of the target audience) and approximately 1,900 logins per day. Conservatively, since launch (one year ago), this has saved approximately 35,564 hours (1,482 days) and delivered an estimated return on investment of $1.7MM through productivity acceleration and cost avoidance.

The image below highlights the dynamic dashboard, showing the growth over the past year.

Generative AI

Although we have seen incredible viral growth over the past year, we continue to believe this is just the beginning and therefore, we are excited to see what comes next!

enterprise architecture ea enterprise_architecture artificial_intelligence ai machine_learning ml generative_ai large_language_model llm natural_language_processing nlp transformer data prompt_engineering openai chatgpt gpt microsoft google gemini gemini_pro imagen imagen_3 bard lifeinwork