Mapping LLM Integration Levels to Levels of Autonomous Driving, it does not have to be all or nothing!

I thought it would be interesting to compare LLM integration into your day to day business workflows to Levels of Automobile Automation.

BTW Follow me on YouTube https://www.youtube.com/@alfrednutile

Here are the levels of Automobile Automation:

• Level 0 - No Automation: The driver performs all driving tasks.
• Level 1 - Driver Assistance: The vehicle can assist with either steering or acceleration/deceleration, but not both simultaneously.
• Level 2 - Partial Automation: The vehicle can control both steering and acceleration/deceleration in some situations, but the driver must remain engaged.
• Level 3 - Conditional Automation: The vehicle can handle all aspects of driving in certain conditions, but the driver must be ready to take control when prompted.
• Level 4 - High Automation: The vehicle can handle all driving tasks in specific circumstances without human intervention.
• Level 5 - Full Automation: The vehicle can perform all driving tasks under all conditions, with no human intervention needed.

Right now there seems to be a lot of “all or nothing” mindset when it comes to working with LLMs and I think each tasks you face day to day might benefit differently or not at all from where this tech is right now.

The challenge—and opportunity—lies in identifying which of your daily tasks could benefit from LLM assistance and at what level

Level 0

Sometimes this is it! You might try and use it to write cover letters for job postings only to find out that Grammarly was enough. But you tried. You pasted a few bits of your writing style into a chat window, pasted the job posting and said “Write me a cover letter in my voice for this job” and what came out was embarrassing.

Great stop right there for that task but keep an open mind to other tasks where LLMs might be more helpful.

Level 1

One example of this level we can think about it and me writing this article then giving I write to the LLM to review. I can give it some examples of my voice and then ask it to just update any places it could use more “meat on the bones” etc.

I might not like what it outputs or maybe I do. Maybe I use it as is or maybe I compare the two and just move some bits over. But I might end up with something a little better than what Grammarly could have done. Using examples of my tone it does a good job of not sounding like an alien from another planet trying to sound human and professional.

Level 2

Let’s take the above and take it up a notch. Let’s say I have a dozen articles in my www.larallama.io collection that I wrote, and then I start a chat with a bullet list of points I want to cover in a blog post and I tell it have at it!

Maybe just maybe it can pull it off. Maybe it can do some great tweets or LinkedIn “tweets”?

Anyways maybe it keeps the tone, or ten prompts later it pulls it off. That is great - it took the bones of my story and really fleshed it out for me, potentially saving time and enhancing the quality of the content.

Level 3

Ok like above we are building confidence and we even setup a system like www.larallama.io to collect data from the week. News about Laravel, news about RAG systems, news about LaraLlama.io itself. And then I tell it “Every Friday Write a News Letter and Email me a copy”

Great, but then I check it and if it looks good I post it to the mailing list using [INSERT YOUR MAILING LIST SERVICE HERE]

And that is great, I have a moment to check it, even tell it to try again etc. I am not 100% trusting it yet but getting closer.

This level of automation can significantly reduce your workload while maintaining quality control.

Level 4

Then comes Level 4! Level 3 proves itself week after week. Readers are writing back how impressed they are! They are even sending emails to LaraLlama that become suggestions for next articles and building reports on customer feedback.
It just does not end!

So we turn on FULL AUTOMATION!!! The system now sends the emails every Friday without my proofing, based on the consistent quality it has demonstrated.

Level 5

Ok my example ended above this one just goes over the top. Let’s list out what just happened above in Four to see how crazy this could get then wrap this up.

• One, the system is sending emails every Friday about news of the week.
• Two, the system is reading customer feedback to adjust each week to that feedback and preventing the infamous trolling of the system that would lead it to writing about some obscure topic.
• Three, the system is also taking suggesting based on the stats of the articles read!

This level of automation represents a significant leap in AI-assisted content creation and management, though it's important to note that human oversight may still be necessary in some capacity.

Wrapping it Up

And that's how far we can potentially go with automation for our different day to day work tasks. However, it's important to remember that not all business processes are ready for any level of AI integration. For instance, tasks like helping a customer get a quote on a new web feature still require a human touch (though this might change in a few years). Similarly, creating engaging videos about innovative ideas for LaraLlama.io remains a primarily human-driven process. (For now 🤔)

The key takeaway is that LLM integration isn't an all-or-nothing proposition. There's a wide spectrum of possibilities between no automation and full automation. The challenge—and opportunity—lies in identifying which of your daily tasks could benefit from LLM assistance and at what level. Some processes might be ready for Level 4 or 5 automation, while others might be best served by Level 1 or 2 support.

As we move forward, it's crucial to approach LLM integration strategically. Start small, experiment with different levels of automation for various tasks, and gradually increase the AI's role as you build confidence in its capabilities.

By understanding these different levels of LLM integration, you can make informed decisions about how to leverage this powerful technology in your day-to-day business workflows, potentially boosting productivity and innovation in ways we're only beginning to explore. (That is such an LLM worded sentence 🤦‍♂️)