As a note, this blog is purposefully removing a lot of nuance, otherwise we’d be writing a novel and not a blog.
We’ve been meeting with potential users over the past few weeks to get feedback on our product. We started to notice a question popping up, and decided we should write about it.
The question: can’t I just use ChatGPT to do this?
The short answer: no. But let’s ask ChatGPT directly and see what it says.
ChatGPT (and OpenAI) uses an LLM or large language model, which is a type of generative AI. The focus is on “language” in that phrase. LLMs can predict and generate text based on prompts, after being trained on large amounts of data. While tools like ChatGTP are really impressive, they can’t run causal models, or mathematical models that pull out relationships in a data set. That isn’t their purpose. What ChatGPT can do is help you with getting a high level understanding of your data. LLM
Causal AI, which Dacture uses, focuses on targeted data sets and makes predictions based strictly on those data sets. The data sets are generally much smaller than what’s required for LLMs, and are specific to your business. The results allow for a more nuanced predictive model that applies directly to your business and use case.
Examples of LLM/ChatGPT use cases:
- Better performing and more realistic feeling chat support
- Improving outbound email communication for Sales
- Clustering and classifying large data sets
- Summarizing large portions of text
- Pulling out high level observations from data sets
Examples of predictive modeling/causal AI use cases:
- Improving sales conversion rates for your product
- Risk modeling, like a company wanting to predict how higher interest rates may impact sales or time to close a deal
- Fraud detection
- Using transaction history to predict customer buying behaviour to better target customers
- Increasing revenue by identifying cross-sell/up-sell opportunities
- Improving customer retention by identifying at risk customers earlier, and understanding what will keep them
If you’re interested in coding your own predictive models, ChatGPT can get you started with that, though it may take some trial and error to get exactly what you want, as there are multiple model types and caveats to consider when building the models.
If you’re interested in making predictions based on your data and the nuances of your business without getting bogged down in code, then maybe you’d be interested in what we’re building at Dacture.
As always:
- If you want to learn more about what Dacture can do your for organization, discuss use cases, or see a demo, you can schedule something with us
- You can email us at info@dacture.com if you have questions
