Original subtitles

- [Instructor] Now I'm going to show you

the Agent AI ecosystem

and how to group the various products

available in the market.

This is an important lesson for you to learn

to make build-versus-buy decisions,

and in designing the tech stack needed

for building Agent AI products and solutions.

There are many types of products in the Agent AI ecosystem.

Let us learn them one at a time.

Agent AI begins with the LLM,

so the many LLM solutions are an important part

of the Agent AI ecosystem.

Within this, there are open source and closed source LLM.

You can refer to my course,

No-Code Generative AI For Business Professionals

on LinkedIn Learning

to learn how to pick the right LLM.

An LLM can chat

or generate images or code or video on a prompt.

In essence, an LLM is a neural network.

It stores relationship of words

to other words from its pre-training,

and all it can do is a recall of words

or generate images in a similar fashion.

It cannot run a computer program.

This is where the Agent AI needs tools

to automate tasks.

Tools can be as simple as an internet search

or a web builder,

or it can be a video or image generator,

or it can even be complex tools

that helps the LLM go past predicting words

to help the Agent AI run simple tasks

to achieve the complex tasks requested of it.

There's a category of products evolving

called tool aggregators that provide an interface

to access multiple tools with a simple API interface.

This has led to a new service called Tasks

or Tools As A Service.

What do UX designers do?

They design web and mobile front ends

and the user experience flow.

They do A/B testing to arrive at the best option to proceed

with the right design for a product.

A UX designer product in the Agent AI ecosystem

does exactly this.

It automates this UX design, A/B testing,

user research and tests using agents.

There are products that make it easy

for enterprise to build agents

by offering a drag-and-drop interface

to a set of predetermined prompts

that integrates with the LLM,

like Google Vertex or OpenAI or ChatGPT.

This is a growing area

and one where you can contribute

by bringing your business acumen to shape these tools

to work for your business use case.

Data stores is an important component

of the Agent AI ecosystem.

LLMs need access to corporate data.

The data store can be used

to store embeddings in vector databases.

This is needed for the LLM to do mission translation

or the data store may be a multimodal data store

with the ability to handle multiple modalities of data.

Think of customer data that is stored in text and voice.

Agent AI has set of products

that offer it the capability to remember

what is important for it to improve itself.

This can be short-term

or long-term memory offerings

that allow for storing and recall of the memory.

Self-critique or reflection.

Its own results is an important aspect of Agent AI

and there are products

that offer this capability for Agent AI

so that they can improve towards better results.

The Agent AI can be designed

to auto-critique itself or critique using prompts,

or it can use a human-in-the-loop approach.

There are several products that offer no code

or low-code interfaces to build Agent AI products

for enterprise use cases.

We will be building two Agent AI as an exercise

with two different no-code, low-code agent tools

that are built using different approaches.

This ecosystem is evolving

and you can choose an LLM, a tool,

a memory storage product, and a reflection tool

and begin building out your Agent AI.

You should also watch the tool aggregators

or low-code drag-and-drop products

and how the ecosystem is expanding

and consolidating with new offerings.

Having a clear understanding

of the various tools in the ecosystem is important for you

to be able to make the right decisions for build versus buy,

and in picking the right solution provider,

and to begin designing

and launching Agent AI products and systems.