How to Evaluate AI Software Purchases (Plus Mistakes To Avoid)

Editor's Note: The following blog was taken from our June Monthly Software Report. Haven't checked out one of our Monthly Reports yet? Create a free Vendr account, or if you already have one, access your report here.

‍

Artificial Intelligence (AI) is transforming our world at unprecedented speed. Curious how fast AI is advancing? Just ask ChatGPT. According to the AI-powered service, global spending on AI jumped 27% from 2022 to 2023. Amid all the excitement, it's crucial for software buyers to distinguish between AI hype and essential AI components and to understand what additional factors to consider when buying AI software.

We sat down with one of Vendr’s Principal SaaS Consultants, Raven Howard, to learn what buyers should look for when evaluating AI software, how to recognize common pitfalls, and how to make the best AI purchasing decisions.

Is Their AI Built From The Ground Up Or Simply an Add On?

The first step in evaluating AI software is understanding the core technology. Is the AI integral to the product, or is it just an added feature? True AI companies have built their models from the ground up, often within the last few years, focusing entirely on AI. For instance, companies like OpenAI are deeply rooted in AI innovation, whereas other companies integrate with OpenAI to enhance their own offerings.

It’s also important to understand the Large Language Model (LLM) that is used in the product to mimic human intelligence. Examples of this are GPT4 and Claude. 

“When evaluating software, buyers should ask what AI model the software supplier is using,” Raven says. “From there, you can determine if you have access to that same model, or if the AI is proprietary to that provider, which can offer additional value.”

The outcomes of using AI are a direct reflection of the data you put into it. Injecting the right context into the right problem is essential for delivering value, and understanding how the AI you’re using works is crucial for achieving these outcomes.

An example of this is Glean, a Vendr Verified supplier. While a user could easily copy and paste text into ChatGPT and ask ChatGPT to summarize the content, Glean’s AI summaries are more powerful, because the AI can search 100 of your company’s apps to uncover additional relevant content and data.

Prepare for Additional Security and Legal Due Diligence

When you're buying AI, it's a good idea to set aside extra time for security and legal reviews. You’ll most likely need to disclose your software usage with AI suppliers, which can raise the risk of information leaks. Because of this, it's important to know what data the AI accesses and how your data is protected.

For context, AI purchases take 21% longer than non-AI purchases. In Q2, the median close time for Machine Learning and Artificial Intelligence transactions on Vendr was 63 days, compared to 50 days for all deals. This highlights the complexity and rising costs associated with these new AI deals.

For more information on how AI is impacting cloud and security risks, check our Vendr’s May Monthly Software Report.

Don’t Rush Into Usage-Based Pricing

AI is an expensive offering, so it’s rare to find unlimited pricing plans. 

The most common AI pricing models include usage-based credit systems or capped responses. Understanding these pricing models is crucial. For instance, suppliers may advertise a per-license model, but in reality, they most likely cap usage at a certain number of uses per license. 

“When it comes to usage-based pricing, I always recommend asking for a trial, so you can better understand what your usage will be before signing a contract,” Raven says. “I also recommend starting with a pay-as-you-go model, when available. Avoid committing to a high volume unless you have data to support that you’ll reach that usage level.”

New technology comes with new pricing models and emerging trends. For instance, OpenAI typically doesn’t offer discounts unless you purchase 5,000 licenses or cover 70-100% of your employees. Many customers also report difficulty in obtaining pricing quotes from OpenAI due to its high demand. If you need assistance with your purchase, you can always work with Vendr to simplify your procurement process.

Avoid These 4 Mistakes When Buying AI

Overlooking Hidden Costs

AI is expensive due to the required processing power and cloud hosting fees. Make sure you understand all potential costs associated with the AI software, including any extra charges for exceeding usage caps.

Not Negotiating for a Discount

Once you have pricing, don’t hesitate to negotiate for a discount. Many sales teams are incentivized to sell their company’s new products, so you might be able to secure more flexible pricing thanks to sales incentives, especially if it’s an add-on to an existing contract.

Falling for Hype Over Substance

AI is surrounded by hype, with many products touting AI features that may not deliver significant value. Focus on clear use cases where AI can genuinely enhance your operations.

“When buying AI products, don’t get distracted by bells and whistles,” Raven says. “Ask yourselves the same questions you would with non-AI purchases: will this create a better way for me to do what I’m already doing, or will it help me do something new that I can’t do yet?”

Ignoring Legal and Security Risks

Overlooking legal and security considerations can lead to significant and costly issues down the line. Always ensure that the AI software complies with your organization’s legal and security requirements before making a purchase.

It's also crucial to ensure your company's infrastructure can support the AI you’re buying. Key questions to consider include: Can you quickly test multiple LLMs? Can you identify data issues? Can you maintain data governance with these models? How does the AI impact your customer data?

Final Thought

Navigating the AI software market requires careful consideration and a clear understanding of your needs and the technology. Whether you’re purchasing AI software for the first time or the hundredth, Vendr can help. We not only save users time and money on software, but our expert advisors, like Raven, can simplify your purchasing process every step of the way.

‍

If you liked this article, more data and insights are ready for you. Create a free Vendr account, or if you already have one, access your report here.

‍

‍

‍