Short Bootstrap on AI Agents or Large Agentic Models (LAMs)

AI is in breakout mode, with many initiatives happening left, right, and centre. One of the most exciting inventions is the emergence of Agents, also known as Large Agentic Models or LAMs. They are a jump forward from traditional AI and the current generation of Large Language Models (LLMs). This post is a short bootstrap or primer for people new to LAMs.

Wipro client use cases and examples can be found on our AI pages.

OK, David, just what Are Large Agentic Models (LAMs)?

LAMs are AI designed to chase complex goals and workflows with limited direct human supervision. Sounds scary, huh? Unlike regular AI systems designed to automate specific repetitive or routine tasks, agentic AI finds ways to understand, reason, and act on more complex instructions and scenarios. Or at least that’s the idea!

This moves the AI beyond simple search-style use cases and onto more Taskmaster-style actions (For people outside the UK, this is a TV show where contestants have to achieve weird and wonderful tasks without outside help). They are not the Terminator but more like the Universal Paperclips game (check it out if you haven’t come across it).

This is creating marketplaces for Agents. Check out Agent.ai and try some excellent agents there for free (click the link to get 100 credits to get started; in full disclosure, I get 50 credits if you use it).

Features of Large Agentic Models are:

1. Cool use cases

2. Autonomous goal-setting

3. Better reasoning capabilities

4. Decision-making processes

5. Strong language capabilities

6. Ability to connect with enterprise systems

It sounds like a movie: how do Agents work?

LAMs ingest data from various sources, apply reasoning to understand tasks and generate solutions, acting by integrating with external APIs to execute tasks and learning through a feedback loop to improve effectiveness.

You are scaring me now. Are the robots taking over?

LAMs can help people solve problems in new ways they may not have thought about.

It’s a long way from the Terminator, but Agents can offer a big step up from LLMs (search on steroids) and RPA (predictable systems integration). LAMs give us a potential step up in AI capabilities for tasks. They can increase efficiency by automating complex workflows and connecting various systems. This automation allows people to work on more strategic tasks.

OK, so what’s the use case?

We work on a wide range of AI projects at Wipro and see agents being considered in many areas. Typically, LAMs are used to automate simple tasks, find new areas of productivity, and add something extra to human problem-solving capabilities. Use cases include customer service, content creation, software development, healthcare, research, finance, etc.

Where is this headed? What happens to people?

It’s not the Terminator, more like the old Microsoft Clippy, that works. Increasingly, these sophisticated AI agents will likely collaborate with people in new and exciting ways, allowing us to focus on what we want, like doom-scrolling Instagram, building a new startup, or helping patients recover from illness. Take your pick.

Regardless, if we build these systems thoughtfully and consider ethical and responsible considerations, we can create AI agents that enhance human capabilities rather than replace them. LAMs are here to stay but very much in the early stages of growth.

Hopefully, they will improve how AI and people work together and drive change for the future.