Learn AI Agent Through Implementation
Let's learn through building AI agent about best practices, things to avoid, tools to use, LLMs that are best and share learnings.
There have been a lot of discussion around AI Agents, anti-patterns and how to go about building one. Then there is discussions on SaaS will be replaced by AI or developers replaced by AI. The best way to have opinion on this is to try something out so I decided I want to build an AI Agent.
Now that I know agent needs to be built - what should be the agent doing? Initial thought that came to my mind are two options
HR Agent
Agent will attempt to automate HR processes
Customer Support Agent
Attempt to automate support activities
Also, this is an opportunity to learn and provide reference implementation that take care of 1
Guardrails to control the outputs
Role based access control to agents
Controlling and limiting tool access to LLM only for needed information
Definiting and validating prompts against multiple LLMs
Auditing agent actions
Observability on agent performance
What are next steps?
First one, decide which agent to build. Please vote below to share your thoughts
Second, need to detail out high level requirements. Leave your comments in case any other use case suggestion or bullets of requirements to be covered. Feel free to join to create Agent and publish your findings.
Lastly, I will post final requirements and agent that will be implemented by coming weekend.
When do we expect this completed?
I plan to do this as a weekend activity as and when I get time - hoping 2-3 weekends may be enough to build a close to production ready agent. I would keep giving updates each weekend on progress made through comments, chat and possibly an article to detail my learnings.
Let’s build an agent and learn together! Happy Learning!
Spread the word
Reference to LinkedIn article