Generative AI Engagement Pyramid for growing AI Inference

Welcome to your weekly AI Newsletter from AITechCircle!

I'm building and implementing AI solutions and sharing everything I learn along the way...

Check out the updates from this week! Please take a moment to share them with a friend or colleague who might benefit from these valuable insights!

Today at a Glance: 

• Increasing the adoption and effective use of AI and Generative AI technologies in various use cases or business processes
• Generative AI Use cases repository
• AI Weekly news and updates covering newly released LLMs
• Courses and events to attend

How to Grow AI/Generative AI Inference?

This week, I read about Chet Holmes’s Buyer's Pyramid. It was the first time I saw this pyramid, and the source of this is 'The Larger Market Formula.'

Why this got my attention, I am linking this to the situation of Generative AI Landscape, which is going on the market where we all started realizing sooner enough that whatever investment is

Going on the AI Infrastructure is about what & how we will use, and we need to have an AI Infrastructure in place to do the Inference.' 

In the AI field, what does inference mean? Let's get clarity in simple terms.

'Inference is the process of running live data through a trained AI model to make a prediction or solve a task.'

Definition of Inference:

In artificial intelligence, especially machine learning and Generative AI, “inference” refers to the process by which a trained model makes predictions or decisions based on new, unseen data. After an AI model has been trained on a large set of data during the training phase, it reaches a state where it can apply what it has learned to actual tasks - this application phase is inference.

In generative AI, inference is the stage where the model generates coherent, contextually relevant outputs tailored to specific prompts or queries after being trained on vast datasets.

Few examples:

1. Customer Support Chat Bots use inference to understand and respond to customer queries effectively, providing relevant solutions based on the conversation’s context
2. Medical Diagnosis AI Assistance helps diagnose diseases by interpreting medical images, where inference allows the model to provide potential diagnoses based on visual inputs
3. Personalized Recommendations in e-commerce, Generative AI can suggest products personalized to the user’s tastes and previous interactions, achieved through inference.

Linking to the Buyer’s Pyramid:

Let's consider the increasing levels of interaction and commitment to Generative AI within an organization or industry, which mirrors the structure and intent of Chet Holmes’s Buyer Pyramid.

(The % do not represent actual data; they are meant to align/represent the Buyer’s Pyramid)

60% Not Problem Aware:

At the bottom of this pyramid, 60% are unaware of what purpose they will use AI Infrastructure for; in other words, what use cases or areas in the organizations can lead to using Gen AI / AI.

Vendors are releasing the latest models and capabilities and investing billions in AI Infrastructure, but most organizations don't know for what purpose they will use this infrastructure.

Action for this: You and me, we all, need to invest time, resources, and talent to find out the target ' AI/GPU Inference' where to use this AI. Massive works need to be done. 

This leads us to read, observe, and hunt for use cases in any business process where AI/Gen AI can be utilized. 

This could lead to an entirely new operating model or a wholly new enterprise architecture for the organization, which would help us infuse the AI.

20% Problem Aware:

This twenty percent has identified a few areas and is waiting to see if someone else will do it. Then, we will take action. This group is looking for others' success stories or lessons learned. 

However, they also need to be moved from looking into it to let them know what Gen AI could do in their business; in other words, they need to be aware of the Inference Use cases or how to use it. 

This is the most challenging part as of now.

17% Information Gathering:

I am categorizing these seventeen percent into those who have selected a few pilot use cases for Generative AI and are going for POCs/ MVPs; they are investing resources and money to move in the direction where they can get some outcome.

3% Buying Now:

The three percent group has actually bought or is buying it, has already implemented the use cases of Generative AI into business functions, or may have started some business solely focused on AI. 

Call for Action: 

There is a long way to go, and massive efforts need to be diverted to how to apply Gen AI and AI to business. When we start looking at this one, how to use it, it brings a million-dollar question: 

Why use it, and what business benefit will I get from it?

Therefore, I believe each Generative AI Use case needs to be divided into these 7 areas to show benefits to the organization.

This is how I think, and that's how I have started documenting/ curating several AI Use cases as a repository. You are welcome to contribute and share. By sharing your insights and experiences, we can learn from each other.

Weekly News & Updates...

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

1. SmolVLM-2, the newest model running entirely on-device and in real time, offers a strong foundation for healthcare applications. With proper domain-specific training, small models enable offline inference of 2D (ECGs, CXRs) and even 3D modalities (CT, MRI) while remaining HIPAA-compliant by design. The model weights (256M-2.2B parameters) and their conversions are open source.
2. The AI assistant is called Microsoft Dragon Copilot for clinical workflow link
3. Baidu Inc. has released ERNIE 4.5 & X1. As a deep-thinking reasoning model with multimodal capabilities, ERNIE X1 delivers performance on par with DeepSeek R1 at only half the price. Meanwhile, ERNIE 4.5 is the latest foundation model and a new-generation native multimodal model. link
4. Aya Visio is an open-weigh model that connects the world through language and vision. Aya Vision adds breakthrough multimodal capabilities to the multilingual 8B and 32B model, expanding to 23 languages. link

The Cloud: the backbone of the AI revolution

• How RSS Hydro, OCI, and NVIDIA Can Protect Energy Infrastructure with AI Flood Modeling link
• World Leaders Put Spotlight on Sovereign AI at NVIDIA GTC link

Generative AI Use Case of the Week:

Several Generative AI use cases are documented, and you can access the library of generative AI Use cases. Link

Favorite Tip Of The Week:

Here's my favorite resource for the week.

“Large Language Models in 2025 – How much understanding and intelligence?

Potential of AI:

OpenAI has launched NextGenAI, a first-of-its-kind consortium with 15 leading research institutions dedicated to using AI to accelerate research breakthroughs and transform education.

Things to Know...

One of the best interviews with Allan Dafoe, Director of Frontier Safety & Governance at DeepMind, was conducted by a host of 80,000 hours. Where he "explains one of the deepest patterns in technological history: once a powerful new capability becomes available, societies adopting it tend to outcompete those not. Those who resist too much can find themselves taken over or rendered irrelevant."

The Opportunity...

Podcast: 

• This week's Open Tech Talks episode 154 is "Generative AI Risks and Governance: What Business Leaders Need to Know with Terry Ziemniak." he is a Fractional CISO and Partner at TechCXO

Apple | Spotify | Amazon Music

Courses to attend:

• Long-term Agentic Memory with LangGraph from Langchain and DeepLearning

Events:

• Dubai AI Festival 2025, April 23-24, 2025, Dubai, UAE

Tech and Tools...

• Ollama Deep Researcher, Ollama Deep Researcher is a fully local web research assistant that uses any LLM hosted by Ollama

The Investment in AI...

• Anthropic has raised $3.5 billion at a $61.5 billion post-money valuation, led by Lightspeed Venture Partners.
• Norm AI , a Regulatory artificial intelligence (AI) agent company, has raised another $48 million in funding, bringing its total funding raised in the past 18 months to $87 million.

And that’s a wrap!

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

I would love to hear your thoughts. Please reply and share what you found most valuable this week. Your feedback means a lot to me.

Until next week, 

Kashif Manzoor

The opinions expressed here are solely my conjecture based on experience, practice, and observation. They do not represent the thoughts, intentions, plans, or strategies of my current or previous employers or their clients/customers. The objective of this newsletter is to share and learn with the community.