Explore the transformative potential of AI agents
In the last few years, we have seen incredible progress with Generative AI models. Thanks to these advances, we can use natural language to generate code, summarise reports, or even decide what action to take.
You might have heard the term “AI agent” before. This is when we take the power of generative AI a step further. Instead of prompting a chatbot and waiting for its response, you can create a custom agent that can take action and autonomously complete a task for you. Essentially, an AI agent is a bot that can solve a task and figures out the steps that are required. It can try to do this independently or ask for your feedback on how it should progress.
To power these agents, we use a combination of tools – with Large Language Models used to restrict the agent and provide it with the intelligence needed to make decisions. We also give agents access to APIs and sections of code, enabling them to use these resources when necessary. These are incredibly capable tools and allow us to do things that simply aren’t possible with ChatGPT. For some time now, Kainos has been exploring how agents could be used and what new approaches are needed to implement the technology.
How you can use agents in practice
There are lots of ways AI agents can be used. You can have one agent that is specialised in a specific task, or groups of agents that can communicate with each other and work together on a task.
Within Kainos, we have been exploring how AI agents can be used to improve content design for our customers – with multiple agents used to scan the content and identify improvements.
Since content is often stored across multiple webpages, we have also used AI agents to store information and become “experts” on a topic – which are then used to solve user queries.
But there are many more use cases. For example, you could ask an AI agent to compare the share price between two companies and store the graph on your computer.
In this case, the agent plans what it should do to complete the task – with code generated to retrieve the financial data, plot the graph, and then save the file locally on your machine.
For years, that sounded like science-fiction – but it’s not anymore. That’s a real example of how we can use AI agents right now. The technology is already here.
The benefits of agents
There are several advantages to using agents when compared to using Large Language Models (LLMs) on their own:
- Incredibly flexible – you can let them work out the steps needed; we don’t need to define every stage in the process – like you would when chaining prompts.
- Better results – by splitting a task among several agents, they are more focused than when you only use one LLM.
- Groups of agents – several agents can communicate with each other and decide which agent is best-suited for the task.
- Generate, check, and improve – an agent can generate the first draft, ask other agents to review it and provide feedback, and then use their responses to improve the content.
- Multiple models – you can use a mixture of models from different vendors, such as OpenAI, Anthropic, and Mistral.
- Humans are in-the-loop – you can intervene and tell the agents to go down a different path.
Other use cases
We touched on how agents can be used to analyse financial data, but that’s just one example. AI agents can also be used to draft posts on social media. It’s pretty rare for AI models to generate long-form content that’s perfect the first time.
Instead, we can use several AI agents to get better results. We could have a Writer agent that will generate the first draft. It then sends the content to a Reviewer agent, which checks that the post matches your brand guidelines and suggests improvements.
Finally, an Improver agent can be used to implement those suggestions and adjust the content to use your brand’s voice. By using multiple agents, rather than just one LLM, you’re more likely to get better results – as one LLM can become unfocused when given too much text to analyse.
Looking ahead to the future
AI agents are on track to revolutionise how we interact with data and accelerate much of our work. We can use groups of agents to aid with brainstorming sessions, review code for security and performance issues, or spot issues with write-ups and improve them.
We can already use frameworks to implement these agents, such as Microsoft AutoGen or LangGraph, and the technology will only improve. Interestingly, Anthropic have just upgraded their AI model and it can now control your computer – with more product announcements expected in the coming months.
Agents are incredibly flexible and can figure out the steps needed to complete a task. That opens up new possibilities for what we can do with software and how we interact with data, which is exciting to see.
Within Kainos, we have been exploring how AI agents can be used by our customers and accelerate their work. If you’d like to find out more about how the technology can support your goals, you can contact our team for more information. Click this link to drop us an email.