Value Architecture

Aligning Gen AI programs to business outcomes is our battle cry at Oraczen. We do this with a wider partner ecosystem that includes some niche Strategy and Domain consulting firms. Our recommendation is for an Enterprise to first arrive at a set of business outcomes that dove tail into your exiting Digital Transformation journey. We then bring these ideas to production quickly. To help Enterprises we have developed a comprehensive Value Architecture approach that essentially works on two dimensions

‍

Value Framework

The Value framework rates a task, activity, or even a process on a scale of 0-5, with 0 being no cognitive automation to 5 being completely autonomous AI driven outcomes. Each maturity level has an associated cost to achieve and a quantifiable business / technology benefit associated.As an example, one of the most transformative impacts of Gen AI is in providing Enterprise Search. At one end are retrieval systems that use no AI. At a higher level companies use some kind of classical AI and a semantic topology to achieve better results. LLMs can take this to higher levels with out of the box, RAG, Fine-tuning applied for even more contextual search capabilities.

Roadmap

The second is a timeline where we divide an AI implementation roadmap into Foundation, Growth and Go-Big stages.The idea of the foundation stage is to achieve quick value realization while doing the first iteration of the AI lifecycle elements described above. Here are some key activities and outcomes that were targeted for a Enterprise client

Establishing a cross functional team across business, IT, finance, security, legal that starts to understand AI value.
Creating a value management process for value tracking and reporting frame work. This process is used to evaluate a prioritized set of use cases enabled by Gen AI.
A brief 3P assessment that informs an Enterprise AI architecture (tech stack and data policies).
Building a “AI middleware” that allows Enterprise data to be shared with Gen AI platforms like Open AI securely and in a cost effective manner. Aegis AI has developed a “Tryst” platform that allows text documents to be tokenized and shared in a confidential and cost effective manner for Gen AI processing.
Create a co-sourced AI Center of Excellence that leverages Aegis AI factory for Data Science and MLOps

In this foundation phase the AI target use cases are centered around improved enterprise productivity – high impact and low complexity typically in Sales/Marketing and IT. The Aegis AI Use Case catalog offers foundation use cases around content creation and co-pilotsIn the foundation phase our emphasis is on Use Cases that can leverage the Gen AI ecosystem and starting to lay the stage with Enterprise Deployment guidelines

Security of data leaving your enterprise AI to train up the use case, all of that needs to be secured and be compliant.
Accurate – is the output accurate so that it justify the change management effort, the model gets it right. Enterprise environments are messy, duplicative, outdated data that solutions need to take into account.
Legal compliance with regulatory requirements like EU AI act or the New York AI act.
Transparency at every level (including auditability).

Growth Stage (6-18 months)

The idea of the Growth stage is to bring all forms of AI including traditional Machine Learning models to expand the value realization. Typically this requires a strong Data foundation and governance to be in place on a select Cloud platform. In this phase an expanded use case framework aims at realizing Enterprise value through Creativity and not just productivity. Expand to creative – own stuff, accurate, derivative of internal data structure and security needs to be well formed. So the existing maturity of the cloud and data foundation would determine the timeline, cost and targeted use cases.

Go-Big State

The Go-Big stage typically builds on AI maturity organization to transform key business processes and/or to build new ecosystem partnerships. Some of the examples include clients trying to simulate existing key business processes to better predict outcomes (ex: supply chain, sales/marketing promotion outcomes), completely automate a back office process (ex: medical claims) and is usually a multi phase journey.