Over the past few years, my team has been on a journey to develop and scale a multi-model, multi-modal Generative AI framework for business, aiming to accelerate productivity and unlock new insights from data.

The work was recently recognised by Google Cloud CEO Thomas Kurian during the “Gemini at Work” global digital event opening keynote.

I have documented the journey over a series of blog posts.

As we close the year, our Generative AI framework has reached another significant milestone (see the dashboard screenshot below).

Agentic AI

Specifically, we have passed 7,000 unique users. This is important, as my company has approximately 9,000 employees (excluding contractors), some of which do not have access to IT services. At the start of the journey, we predicted that 7,000 was the viable audience for Generative AI, meaning we have reached peak adoption.

This has delivered an estimated return on investment of $2,419,808, predominately through efficiency gains and cost avoidance. I shared more details regarding specific success stories in the article “Generative AI - Value”.

A major part of this success has been our marketplace, which allows any employee to securely create and share a chatbot (known as a custom persona), including specific business context and the ability to ingest data (unstructured and structured) from multiple source systems.

We now have over 1000 unique chatbots, all built on the same underlying cost-effective architecture. These chatbots support individuals and teams, covering the entire value chain from Research & Development through to Commercial Sales and Marketing.

As highlighted by the slide below, this private, secure and compliant foundation leads us to our next area of opportunity, specifically action and learning.

Agentic AI

Currently, our Generative AI framework responds to prompts and executes predefined tasks. In the future, we expect it to make decisions, plan and execute actions, and even learn. The industry has coined the term “Agentic AI” to describe this evolution.

We have already launched our first version of this capability, which we call “Workflows”.

Agentic AI

In short, workflows allow any user to declaratively (no code or specialist knowledge required) create chained events using pre-approved components. These events can be triggered in real-time, promoting experimentation and continuous improvement through iteration.

For example, a workflow could accept input from an external system or data source via an API. This input could be ingested and processed by multiple AI models, including self-verification and/or summarisation of the output. The workflow could then dynamically trigger a downstream action based on the input, including an API call to an external system or data source.

In our early testing, we have been able to create some very interesting workflows, pulling data from systems such as SAP, and analysing the data using a reasoning model, such as OpenAI o1 and Google Gemini 2.0, producing a structured output that includes text and images.

We believe this capability will trigger the next wave of innovation regarding the use of Generative AI, further increasing the overall return on investment. It also presents a far more modern, decoupled and less brittle approach to process automation, where the workflows are programmatically defined (reusable) and can dynamically adapt based on an ever-evolving landscape.

We are not alone in this thinking, with Google demonstrating similar capabilities within their AI Studio and as part of Project Mariner.

enterprise architecture ea enterprise_architecture artificial_intelligence ai machine_learning ml generative_ai agentic_ai large_language_model llm natural_language_processing nlp transformer data prompt_engineering openai chatgpt gpt o1 microsoft google gemini gemini_pro imagen imagen_2 imagen_3 ai_studio project_mariner bard lifeinwork