​AI Tech Circle​

Welcome to your weekly AI Newsletter! Read and listen on AITechCircle:

This newsletter has quickly become a trusted go-to for me and many others in the AI community, offering practical insights that can be applied immediately to elevate your work or business.

Before reading this week’s updates, consider sharing this resource with a friend or colleague who could benefit from these valuable insights!

Today at a Glance:

• Navigating the AI Adoption journey – The first part, Adopt
• Generative AI Use cases in Financials Industry, Generating Financial Reports
• AI Weekly news and updates covering newly released LLMs
• Courses and events to attend

Navigating the AI Implementation Journey

This is a continuation of conversations at GITEX, the region’s largest technology conference; several questions and discussions were exchanged while interacting with many folks during the GITEX. I have covered typical questions asked last week, and I can relate to those we covered in our earlier edition: How do I get started with AI? Where can AI deliver the most value? And How do I scale and operationalize AI?

​

This week, I wanted to address another set of questions received from an audience during the event. Most of the folks I met were trying to explore how to drive or intake AI from the ERP application’s point of view.

Here are my two cents on this topic.

Organizations typically need to go through stages of adopting and integrating AI and Generative AI, starting with simple adoption and moving toward advanced development.

It’s similar to how quickly someone can take on a pilot project in the organization or division.

You can go through the three phases: Adopt, Buy, and Build, each representing a different approach to leveraging AI.

The diagram outlines three stages for organizations to adopt AI, moving from simplicity to deeper integration derived from Deloitte Forum 2023 Gen AI and McKinsey’s Take, Shape, and Make concept:

​

This approach balances simplicity with increasing complexity and strategic business impact as organizations evolve in their AI journey. From using embedded AI features in existing SaaS applications and extending capabilities with specialized AI Services / Generative AI to creating custom AI models, the progression emphasizes a shift from simplicity and cost-efficiency to maximizing business value and ensuring governance.

This structured path helps organizations align their AI strategy with their growth objectives.

To start following the above approach, refer to the detailed article “Simplified Architecture Was Designed to Take up Generative AI​ in Cloud Applications.” This architecture followed the approach of Just Enough or Good Enough Enterprise Architecture (EA).

​

Let’s start with the three-part series covering the three stages of the AI Implementation Journey.

1. Adopt (Takers): Leverage AI features embedded in existing SaaS applications, focusing on ease and low costs.
2. Buy (Shapers): Expand AI capabilities using a Generative AI portfolio, aiming to extend functionalities.
3. Build (Makers): Develop and train custom Generative AI models, focusing on business value and governance

The first part will be covered in this week’s edition, and the remaining two will follow during the next two weeks.

Adopt (Takers):

This is an easy intake and quick way to start the AI journey in the organization. You only need to explore your current footprint of an application.

For example, if you are using Oracle SaaS applications, which have released 100+ AI, Gen AI Agents features as part of the standard product offerings, your task is to go through them individually, map them to your organization’s benefits, and start using them.

Let’s explore it further. Generative AI in Fusion applications targets three areas: 1 – Assisted Authoring, 2 – Suggestions, and 3 – Summarization.

Users retain complete control when using generative AI in Fusion Applications. These AI-driven features activate only upon user request via the AI Assist button. Content generated within the application can be reviewed, modified, or disregarded as needed.

Here’s a simple breakdown of how the process works:

Source: Unleashing the potential of generative AI directly within Oracle Fusion Cloud Applications Suite

Oracle Fusion AI Agents: AI agents integrate large language models (LLMs) with other tools and technologies, enabling them to perform complex tasks traditionally handled by humans. These agents interact with their surroundings to collect information, identify the necessary steps to meet specific goals, and act in designated roles. They can plan, access various tools and data, make autonomous decisions, and even collaborate with other AI agents to achieve their objectives.

Source: An Introductory Guide to Oracle Fusion AI Agents

​Oracle’s initial RAG (Retrieval-Augmented Generation) agents in the Fusion Applications are just the beginning. The AI Agents are categorized into Supervisory, Conversational, Functional, and Utility agents who work together to achieve specific tasks or outcomes. In practice, these agents interact, leverage tools, retrieve data, make decisions, and coordinate efficiently to accomplish shared tasks.

For example, Agents for hiring managers can streamline recruitment by documenting essential requirements, such as desired skills and experience, to aid hiring decisions. It also reviews job postings generated by GenAI systems to ensure accuracy and relevance.

A utility agent can do a specific function or tool and is activated by other agents to complete tasks, such as querying a database, sending emails, performing calculations, or retrieving documents.

​

The Buy (Shapers) and Build (Makers) parts will be covered in the next two weeks.

I’d love to hear your insights and experiences regarding the route you’re taking on your AI journey.

​

Your feedback is invaluable!

Weekly News & Updates…

Last week’s AI breakthroughs marked another leap forward in the tech revolution.

1. OpenAI’s new research focuses on enhancing Continuous-Time Consistency Models called sCM, which aim to stabilize and scale AI model performance over time. By refining these models, OpenAI seeks to simplify the training process, reduce computational needs, and maintain model accuracy even as they scale. This breakthrough holds promise for creating more efficient, scalable AI systems. link​
2. Aya Expanse from Cohere is an open-weight, state-of-the-art family of models to help close the language gap with AI. Multilingual large language model serving 23 languages. link​
3. Hugging Face introduces HUGS (Generative AI Services), a solution for rapid AI application development with optimized, zero-configuration inference microservices. Built on open-source technologies like Text Generation Inference and Transformers, HUGS enables efficient scaling of generative AI applications within your infrastructure, simplifying the deployment of open models. link​
4. OmniParser from Microsoft is a screen-parsing tool that converts UI screenshots into structured data to enhance LLM-powered UI agents. Its training datasets include an interactive icon detection dataset from widely used web pages, annotated to identify clickable areas, and an icon description dataset that links UI elements to their functions. link​

The Cloud: the backbone of the AI revolution

• Serving smaller Llama LLM models in a cost-efficient way with an Ampere CPUs link​
• LLM inferencing with Arm-based OCI Ampere A1 Compute in OCI Data Science AI Quick Actions link​
• AI Hypercomputer software updates: Faster training and inference, a new resource hub, and more link​
• How to Accelerate Larger LLMs Locally on RTX With the LM Studio link​

Gen AI Use Case of the Week:

Generating Financial Reports with Large Language Models (LLMs). Large Language Models streamline report generation by integrating multiple data sources into structured, regulatory-compliant formats, cutting manual effort, reducing errors, and delivering timely insights for stakeholders and investors.

​

Favorite Tip Of The Week:

Here’s my favorite resource of the week.

• Open-sourcing of SynthID text watermarking tool through an updated Responsible Generative AI Toolkit from Google. SynthID uses a variety of deep learning models and algorithms for watermarking and identifying AI-generated content. link​

Potential of AI

• Meta’s Open Materials 2024 (OMat24) Dataset and Models. This includes the OMat24 dataset and the sAlex dataset. link​

Source: Meta Open Materials 2024 (OMat24) Dataset

Things to Know…

The article from Andreessen Horowitz explores AI’s role in addressing the “messy inbox” challenge, where traditional tools fail to manage an overwhelming influx of emails and messages. It highlights how AI apps can act as a powerful entry point (or “wedge strategy”) for companies, helping users better organize, prioritize, and respond to messages through tailored automation and intelligent filtering.

This strategy positions AI as indispensable for modern communication efficiency.

Source: The Messy Inbox Problem: Wedge Strategies in AI Apps

The Opportunity…

Podcast:

• This week’s Open Tech Talks episode 147 is “How organizations can start the journey of AI with Mark Wormgoor.” In 2022, after 25+ years in technology in the corporate world, Mark decided to start Tairi. Tairi originates from his passion for leadership in tech.

Apple | Spotify | Amazon Music ​

Courses to attend:

• Stanford released a 1.5-hour lecture on Building Large Language Models on Youtube​
• Practical Multi AI Agents and Advanced Use Cases with crewAI from DeepLearning​

Events:

• ​Future of AI, 6-7 November 2024, London, UK
• ​9th Middle East Enterprise AI & Analytics Forum, Nov 7, 2024, Dubai, UAE
• ​Global AI Show, Dec 12, 13, 2024, Dubai, UAE

Tech and Tools…

• ​CodeGPT empowers you with an AI copilot, which can be self-hosted into your infrastructure.
• ​Skyvern automates browser-based workflows using LLMs and computer vision.

Data Sets…

• Dataset for Evaluation of Extreme Weather Impacts in the USA: This dataset is a fusion of three data types (operations and maintenance tickets, weather data, and production data) that was used to support machine learning analysis and evaluation of drivers for low-performance at photovoltaic (PV) sites during compound extreme weather events.

Other Technology News

Want to stay updated on the latest information in the field of Information Technology? Here’s what you should know:

• Google is reportedly developing a ‘computer-using agent’ AI system, as reported by The Verge​
• VC megadeals are booming — and AI is surprisingly not the top category, as reported by TechCrunch​

Join a mini email course on Generative AI …

​Introduction to Generative AI for Newbies​

Earlier week’s Post:

And that’s a wrap!

Thank you, as always, for taking the time to read.

I’d love to hear your thoughts. Hit reply and let me know what you find most valuable this week! Your feedback means a lot.

Until next week,

Kashif Manzoor