So you want to use AI in your business, but you don’t know how to get started.
Well, here’s the short answer: prepare your data.
The longer answer?
An AI application is only as good as the data it’s built upon.
According to Business Insider, ChatGPT was trained on ~300,000,000,000 words, or 570 GB of data.
For context, 300,000,000,000 seconds translates to ~9512 years!
Of course, if you’re reading this you probably aren’t trying to build your own foundational model like OpenAI.
But even the simplest AI workflows require proper data to perform consistently.
It’s almost like exposure therapy. The more good data the model is trained upon, the more resistant it is to making mistakes.
Where do I get the data?
You already have it!
Here are some examples:
- Claims Data
- Emails
- Content
- Documentation
- Texts
- Phone Calls
- Etc.
Notice that the list doesn’t include rummaging through your EHR for something like patients’ demographic information (though that’s certainly a viable source).
Pretty much anything you think of can serve as data, especially in healthcare where a lot of that data is unstructured.
Unstructured data is basically unorganized data. It’s data that’s accumulated without a scaffold or schema.
A perfect example would be a transcription of a patient-conversation. It can include everything from dates, to facts, to numbers, and more.
But it’s not like patients give you information in the exact order and quantity you need to update their medical record.
In other words, lots of different types of information, yet very difficult to store in something like a relational database (which is basically a table). And it’s these kinds of databases that many applications are built upon (including EHRs).
The manual process of taking information from a patient-conversation and organizing it in an EHR is what many physicians get burnt out from.
Fortunately, LLMs are actually quite good at extracting information from unstructured data (like patient conversations), and organizing it so that it can be used for other purposes (such as maintaining an EHR).
In summary, every part of your business has a pool of data waiting to be used.
How do I get started?
A really simple way to get started is just by organizing the data associated with one SOP or process within your business.
For example, lots of medical and dental practices use Facebook and Instagram to get customers (using both organic and paid traffic).
So assuming you’re making any kind of content, each post, each ad, and each article is a piece of data that an LLM application can learn from.
If you wanted to design an AI workflow that repurposes your content, you would need a lot of examples of content you’ve already created.
Among those examples, you’d want to tell the LLM which content performed well, and which content didn’t, so that it can avoid the latter and produce the former.
Practically, you could do this by having a folder of your best-performing content, a folder of your worst-performing content, and a folder for everything in between.
This cuts down on the time it’ll take you or an AI developer to clean the data so that it can be used for the application.
TL;DR
- An AI application is only as good as the data it’s built upon
- You have data literally everywhere in your business
- A really simple way to get started with using AI in your business is to organize your data
That’s all for this week.
See you next Saturday.
Whenever you’re ready, here’s how we can help you:
- Newsletters: Our newsletters provide tactical information that innovative entrepreneurs, investors, and other forward-thinking people can use to scale their impact.
- Community: Coming soon! You’ll automatically be added to the waitlist by joining any of our newsletters.
