Using LLMs to discover automatable business models | Dabble.AI #3
Using LLMs to discover automatable business models | Dabble.AI #3
How close are we to seeing fully autonomous business models?
Can AI autonomously build and run a profitable business without employees? That’s the question the Dabble.AI project is trying to answer.
In this post, I’ll be outlining the steps I've been testing to generate automatable business models using LLMs. I also explain my working assumptions and the general thinking behind each step.
My wife sometimes tells me "Your mind is like a drunk baby." It's a funny line from the movie Office Christmas Party - so she's not being mean. She's just reminding me that sometimes the things I say can come out sounding random and disconnected. I know she's right - I often confuse myself. I'm trying to stay mindful of that in these posts.
To that end, I’m going to begin each post with a few questions I’m focusing on. But I won’t necessarily have answers. They are more about staying focused. Hopefully that will keep me from sounding completely incoherent. But no promises. 😵💫
Focusing questions
Can LLMs be used to discover automatable business models? Can they help me discover models that are unique and not totally obvious to everyone? How? What are the steps? What are the prompts? What LLMs, or combination of tools would do the best job?
My working assumptions
AI tools currently exist to enhance and optimize almost every facet of a company. The tools will get better and better until humans are only needed for oversight. Today, there is a human in the loop for most tasks. That won’t be necessary for much longer. I’d guess by 2033 for online only businesses. Possibly much sooner. At that point, AI will do all of the following things - autonomously.
Identify market opportunities
Create and test products based on market opportunities
Raise funding if needed
Create product marketing materials
Build and manage online communities
Test and execute marketing campaigns
Take and fulfill orders
Provide product support
Monitor and adapt to market demands
Iterate and optimize everything
But many offline businesses will also be fully automated. Even in cases where human capital is required, AI will be able to fully manage the workforce. For example, consider a future in-home nursing care business. AI will:
Post job openings
Interview applicants
Conduct background checks on applicants
Negotiate compensation based on labor supply and demand
Monitor that workers are showing up (GPS tracking - like what Uber does)
Monitor customer satisfaction. Think Amazon Alexa devices in the home that does advanced sentiment analysis on the interactions between the caregiver and the person receiving the care.
Adjust compensation or replace workers based on customer satisfaction scores
That was an in-home nursing service example but you can apply the same thinking to almost any other in-home services - plummers, electricians, painters, HVAC services - almost anything.
The technology to do everything above already exists. Now it’s just a matter of connecting the dots. That’s often easier said than done. Things like privacy concerns and other compliance nuances will create friction for sure. So, it’s going to take some time. But, if the current technological trajectory continues, this is just a matter of time.
I’m sharing these assumptions because they influence how I’m going about my business model generation research. I’ll expand in future posts. For now however, let’s get into the steps I’m testing and the tools I’m using.
Business model generation steps
Here are the general steps I follow to generate business models with LLMs.
Decide on a general business model.
Create a list of my personal values, experiences, interests, and skills.
Define very specific goals and objectives.
List your available resources.
Create a prompt using the stuff from steps 1-4.
Use the prompt to generate a list of business ideas using multiple LLMs.
Identify similar models from each LLM and combine them into a single business model.
Prompt each LLM to expand on the combined business model.
Now let me break each step down in more detail.
Step 1: Decide on a general business model.
In my last post, I provided a list of general online business models. The fist step in the process is to choose one of those models. I'll use that model as the basis for the business model I'm going to generate.
NOTE: I’ve tested prompts without doing this. It’s helpful for brainstorming but tends to make the results too broad.
Step 2: Create a list of personal values, experiences, interests, and skills.
In a future post, I'll break down my thoughts here in more detail. But the purpose for this step is two-fold. First, by including your personal values, experiences, interests, and skills, you're more likely to generate business models that you're passionate and knowledgeable about. Second, you're more likely to generate business models that might not be obvious to others. A good business model is NOT one that everyone else is already doing. Or one that anyone else can easily copy. As AI becomes more capable and mainstream, it will be increasingly easier for business models to be copied. So, you want to focus on trying to identify models that are unique, not obvious, and not easily replicated. Again, I'll go into more detail in a future post.
Step 3: Define very specific goals and objectives.
I define goals as outcomes I'd like to see happen, but don't have 100% control over. For example, I might set a sales goal of $1 million dollars. But I don't have 100% control over whether or not I’ll generate $1 million dollars because I can't control the actions, preferences, or decisions of others.
Objectives on the other hand are things I have 100% control over. For example, I can set an objective to write 10 blog posts. In the context of an automated business, I might want the business to generate 1M in revenue but I can't automate that outcome. I can however automate the process of writing 10 blog posts.
The main thing for this step is to be very specific about what you’d like to achieve (goals) and what need to get done (objectives).
Step 4: List your available resources.
This step is pretty straight forward. I create a list of the resources I have available to me. This includes things like capital to invest, time, skills, and relationships, etc. I’m very specific here. I don’t want to generate a business model that requires $500,000 in startup capital if I only have $10,000 to invest. Or for a market I know nothing about and don’t have any contacts in.
Step 5: Create a prompt using the stuff from steps 1-4.
I use the information from steps 1, 2, 3, and 4 to create a prompt. I’m not sharing the prompts I’m testing now for two reasons. First because I’m still testing and refining this process. Second, because I’m using information that others have provided me for testing purposes. So, it’s the secret sauce. Reread step 2 if you’re still wondering why I’m not sharing those prompts. 🤔
NOTE: I can create some example prompts with fake data if that’s of interest to enough people. If so, please leave a comment.
Step 6: Use the prompt to generate business models ideas using multiple LLMs.
After creating the prompt I test it using three or more different LLMs. My go-tos have been ChatGPT, Bard, and Claude.ai. But I’m also testing with:
Note: If there is a model you think I should be testing with. PLEASE leave a comment.
Step 7: Identify similarities and combine them into a single business model.
Using the results generated by each of the LLMs, create a single business that combines similar results. At this point you should have just one business model.
Step 8: Prompt each LLM to expand on the combined business model.
Using the final business model from Step 7, create a new prompt to expand on the model. For example, that prompt might start something like.
I'm working on a business model for an online business. Can you expand on what I have so far? My current model is:
[model from step 7 goes here]What's next?
I’ve been thinking a lot about game theory. Questions like: How would I define the game? How do you win? What are the rules? Who are the players? What are the constraints, paradoxes, risks, rewards, incentives, penalties, and strategies? Mostly about strategies. I’ll probably talk about that next - if I can do so coherently.
Thanks for following.