Press ALT + F8 to see a list of keyboard shortcuts
Manage Discussion

Module 9: Q&A Discussion Board

All Sections
3 unread replies. 3 replies.

Use this Q&A discussion board to share your insights and post your questions as you engage with the material from this module. Feel free to engage in a healthy discussion with your peers. 

It is not mandatory to participate in this discussion and there is no grade associated with it.

To make this most useful for all participants:

  • Bookmark this page and return to it often during this module.
  • Search for your topic first (use the search box below).
  • If someone has asked a similar question, respond to that particular thread to keep the conversation contained. This will allow your fellow classmates to spot specific topics more easily and offer their insights.
  • Create a new thread if your question has not been addressed.
  • Start your post with a brief subject statement that specifies the topic or problem.
    • Examples:
      • Assignment 1.2, Question 3
      • Installing X issue
      • Sharing an interesting article about X
  • Post questions and topics related to this module’s content and be as specific as possible when referring to programme content or learning activities.
    • Example: In Video 3 at 4:10

Reply! Share your thoughts and insights.

Filter replies by unread    
Reply to Module 9: Q&A Discussion Board
    Attach
    Cancel
    • Collapse Subdiscussion Priya Jha

      Priya Jha (She/Her)

      1) Curious to hear from the cohort about your experience of productionising an Agent. Has anyone actually gone through the full journey from identifying the use case to getting an Agent into production? Did you follow something similar to the 90-day plan from the course - for example, starting with a data audit/clean-up and workflow redesign before building the Agent? What worked well, and what turned out to be harder than expected? 2) In Video 4: Trust & Compliance, Prof. Ian mentions that 80% of organisations have already experienced risky behaviour from deployed AI agents. Most of what I come across in the news tends to be around cases where an Agent has gone "rogue" or acted in an unexpected way. For those in the cohort who are already working with Agents in their organisations, I’d be interested to hear about your real-world experience. Have you come across situations where an Agent behaved in an unexpected or risky way? And what controls or guardrails did you put in place to manage it?
      Manage Discussion Entry

      1) Curious to hear from the cohort about your experience of productionising an Agent. Has anyone actually gone through the full journey from identifying the use case to getting an Agent into production?

      Did you follow something similar to the 90-day plan from the course - for example, starting with a data audit/clean-up and workflow redesign before building the Agent? What worked well, and what turned out to be harder than expected?

      2) In Video 4: Trust & Compliance, Prof. Ian mentions that 80% of organisations have already experienced risky behaviour from deployed AI agents.

      Most of what I come across in the news tends to be around cases where an Agent has gone "rogue" or acted in an unexpected way. For those in the cohort who are already working with Agents in their organisations, I’d be interested to hear about your real-world experience.

      Have you come across situations where an Agent behaved in an unexpected or risky way? And what controls or guardrails did you put in place to manage it?

      Reply to Comment
        Attach
        Cancel
        • Collapse Subdiscussion Edoardo Bertolani

          Edoardo Bertolani

          We are learning how to use AI models, agents and workflows, but will we eventually reach the point where we can use existing AI models to help us build and train our own AI model? In other words, can AI increasingly automate the technical work required to create another AI model, making model development accessible even to people without a deep technical background?
          Manage Discussion Entry

          We are learning how to use AI models, agents and workflows, but will we eventually reach the point where we can use existing AI models to help us build and train our own AI model?

          In other words, can AI increasingly automate the technical work required to create another AI model, making model development accessible even to people without a deep technical background?

          Reply to Comment
            Attach
            Cancel
            • Collapse Subdiscussion Nora Pataut

              Nora Pataut

              Maybe you could think of how to apply agentic AI around this? What would be the various steps required to get to building an AI model and with what outcome(s). Then for each step, define the inputs/outputs and look at what can be automated or not. Anything could be written as a workflow. Define model objective(s) > Data assessment > data set consolidation > data training set > build model > test > validation ...
              Manage Discussion Entry

              Maybe you could think of how to apply agentic AI around this? What would be the various steps required to get to building an AI model and with what outcome(s). Then for each step, define the inputs/outputs and look at what can be automated or not. Anything could be written as a workflow.

              Define model objective(s) > Data assessment > data set consolidation > data training set > build model > test > validation ...

              Reply to Comment
                Attach
                Cancel