PragmWhere organisations still have an advantage
Unless you work in the public domain, open-source AI systems will not:
- Have access to your knowledge
- Understand your processes
- Understand what good looks like for you
That's where your advantage lies.
And you can, and should be, leveraging AI to increase that advantage.
With Generative AI it is now possible to automate more of your processes.
And to setup evaluations to trust it's doing a good job.
Hence freeing you and your staff up to focus on strategy. And management. Before setting more AIs to work.
Lets image an intern is joining your company to work through this.
The Intern ↔ AI Agent Parallels
Imagine you have an intern joining your team. You need to train them up. My guess is that you would:
- Tell them what they need to know
- Tell them what actions they need to perform
- Give them examples on how they should behave
You can think of an agent in exactly the same way.
And now imagine you have a team of interns joining. They're all responsible for different tasks. And you need them to communicate with each other based on the outcomes of each tasks.
That's where Agentic Systems come in.
A team of Agents that are constrained to different tasks. And can go back and forth with each other to complete a piece of work.
Going back to the example of coding, this might be a Product Manager Agent responsible for understanding the current code and assigning tasks, a Coding Agent responsible for writing the code and a Review Agent responsible for reviewing and suggesting refinements to the code.
And finally imagine when an intern (or team of interns) complete a task. You need to review their performance, tell them what they did well. And how they can improve for next time.
That's where evaluations and reinforcement learning comes in. And it happens on the fly. The Agents Self Learning System checks it's performance and improves itself. All of this can be visible to you. No need to wait for a meeting at the end of the week.
So agents can replace interns for well defined processes in your enterprise. Except they can get to work within minutes, not weeks. If you've defined for each point in the process:
- The knowledge that needs to be available
- The actions that can be taken
- What good performance looks like
Having these workflows mapped out is the key.
How do we get ourselves ready
I've worked in Enterprise. It can be a mess. Knowledge is held in different databases, different file systems, in people's heads. We don't always know who we need to go to for different tasks. We can feel a little lost when joining a new team. Unsure of what our exact role is and what actions we should perform. That doesn't help interns. And it won't help AI.
This is where the focus needs to be. Defining the workflows. Structuring the knowledge. Getting clear on what actions can be taken at each stage in your process, and what good looks like in each case.
As the development of AI accelerates this is still your moat. And with clear focus in these areas you'll be able to set AI to work for you. AI won't replace enterprise. But enterprise need to embrace AI. And those who adopt it first will win.