Data Action Layer April 4, 2023
Are you considering joining the growing ranks of companies adopting or interested in an AI solution? These days, there’s a bewildering array of software out there claiming to be powered by AI. To further complicate things, not all AI technologies are equally capable in every situation (and not all of them are actually AI!).
It’s important to know how to differentiate between the many types of technology that fall under the AI category in order to find your perfect match.
Here are six questions you should ask when evaluating solutions using AI:
The first question you should ask when evaluating an AI solution is surprisingly the most obvious: What type of AI technologies do you specifically use, and how are you deploying it to customers? Simply applying AI to a business problem may sound good, but it’s important to dig in and understand the why behind it. Do these workflows require AI to create value or is AI being deployed just for the sake of it?
There are different types of AI best suited for specific use cases. For example, large language models (LLM) and natural language processing (NLP) have different skillsets: extractive NLP algorithms can extract insights from unstructured data, and large language models like GPT and LaMDA have the ability to understand and generate human-like language. An NLP can complete a simple extraction task, but you’ll need an LLM if you want to summarize a document
So what do you need the AI to do? Automate tasks (like streamlining customer support through chatbots), provide insights (like analyzing customer behavior to inform marketing strategies), or make predictions (like forecasting sales or demand for inventory)? Knowing what you need solved and understanding how the AI will be applied can help you determine if it’s the best fit for your business needs.
AI models are constantly learning, but they need to be trained to produce the best output. It’s good to understand how much data their solution requires to train the model. Does their model require a lot of training data to get started, or can it learn from a smaller dataset? How long does it take to train the AI—a few days, weeks, or months? How often does it need to be updated (monthly, quarterly, or annually)? It's also worth asking about any ready-made options—are there any useful models available for immediate use? Understanding these factors can help you determine the time and resources needed to implement the solution and help weed out some of the more impractical offerings on the table.
A software provider who understands your problem, industry, and documents is critical for three reasons:
If your use case requires access to sensitive and business critical data, privacy and security are essential. Ask where the vendor is sending data—does it leave their system? What safeguards are in place? Can they accommodate your desired deployment model? It’s important to understand how the technology is being used, what data is sent through it and who has access to it.
Another key consideration when purchasing AI-enabled technology is how much oversight and maintenance the AI model will require. Will it run seamlessly in the background, or will it require adjustments and retraining over time? Will the vendor take that on, or will you have to? It's important to understand the level of involvement required to ensure that the AI solution remains effective and continues to deliver value to your business.
Asking questions about how you can grow with the solution is a good idea as well. For example, if you have another workflow that you would like to automate, can they accommodate it? Additionally, you’ll want to confirm that the vendor offers robust support, such as a dedicated team to provide guidance on making adjustments and retraining the AI solution to keep meeting your needs.
Be on AI-trend, but do it for the long haul
In the current climate and the rush to automate, remember that not all AI solutions are created equal. Asking the right questions will help ensure that you’re not just following the latest business trends, but making an investment that genuinely meets your business needs.
To learn more about how financial services and operations teams can implement AI into your business, check out our ebook, the Financial services guide to understanding AI: The technology behind data acceleration platforms. Get a complete AI glossary and take a deep dive into the different kinds of AI technology and their use cases for finserv.
Fine-tuning is not the only way to get relevant, domain-specific responses out of an LLM. Alkymi’s team of expert data scientists explain an alternate route.
Find out which type of automated document processing solution is right for you: data extraction, an IDP, or a complete business system for unstructured data.
We’re partnering with Portfolio BI, a provider of portfolio analytics and reporting solutions, to bring structured and unstructured data sources together.