The Smart Way to Buy AI

Key questions to consider

There are unique challenges to buying an AI tool, especially given the speed with which the technology is developing.

Read on to see the kinds of questions you should answer before committing to a particular AI solution.

You might also be interested in our previous article on how to make the most of working with any SaaS vendor you choose.

With AI, we recommend you start with something small and encourage your team to explore. The more people you have who know what these AI tools can and can’t do, the more eyes looking for appropriate places to leverage new technology.

The people currently doing the work know best which high hassle tasks they would love to speed up with technology. They usually have less understanding of what’s possible though, which is where hands-on experience can help.

Ask your team what they’re trying to accomplish with AI. Is it just speeding up tedious tasks, or is it about new ways of navigating and synthesizing information? Maybe they’ve heard a sales pitch which described features they would like to explore?

It’s smart to stay skeptical. Most people can’t tell the difference between realistic promises from AI solutions, and hypothetical what-ifs that are still far from fruition. By ensuring the team is carefully evaluating the potential, you can avoid an expensive mistake when the solution underdelivers.

What are the options?

There are many different categories of AI solutions that companies need to evaluate.

General  purpose chat interfaces like Copilot, Gemini, Claude Code, and ChatGPT are a good place to start. After those initial tests, it quickly becomes apparent that more specialized solutions offer important advantages.

Many vendors are already embedding AI into their existing SaaS solutions, which means they have access to all your existing data. However, like most all-in-one platforms, those AI solutions are not usually the best, and are not built to integrate multiple sources of data.

Specialized or custom solutions can quickly become expensive, and it can be hard to assess how well they might work. Also keep in mind that any commitment you make today you may regret three months from now when a new best-in-class tool takes the world by storm.

Flexibility and ease of configuration are often the most important criteria to consider, depending on your use case. Low cost to get started and switch when you need to are crucial considerations when any tool might fail unexpectedly.

IT questions to think about

When your procurement and IT teams work together to vet AI solutions, you often get better results. Failing in either category can introduce massive risks and costs to your organization. Collaboration can also help other business functions to understand why they face barriers to buying the tools they want.

Information security is one of the most important criteria to IT. The biggest concern today is that those who are not tech savvy might simply upload proprietary information into a public AI tool, and give up control by accident.

Not all tech vendors are as careful as they should be, so expect IT to ask some important questions about how your organization’s data will be protected.

Read through our SaaS blog post series to understand the rest of the technical questions your IT group may want answered before committing to a technology.

How does this tool work?

Some of the more technical questions might best be handled by either end users or your IT team. However, what data the system needs to work and how it uses AI to deliver results are important topics for everyone.

Some use cases for AI can exaggerate data quality issues to the point of being unusable. In those kinds of situations, your team is signing up for a lengthy data quality improvement project before they get any benefit from buying an AI solution.

In other cases, data quality problems cause AI to deliver occasionally absurd answers, and all you need is a good process to vet the results.

It’s important to understand the different components of the tool you’re buying. For example, is the tool primarily just a fancy interface for Claude, or are there other kinds of analysis and data transformation that are input for the AI model? Is the solution about taking an existing human process and allowing a computer to automate it, or is it about a new way of getting the work done?

Remember that for now, the cost of running AI models is still being highly subsidized by investors. As that changes, every AI tool in existence will get dramatically more expensive, as a function of what the AI models are doing, and how efficiently computing resources are being used.

If the vendor you’re working with is not carefully designing their systems, you might end up with a large cost increase when the investor money dries up.

What are the business factors for adopting AI?

Use the expertise of your procurement team to help you understand the business context and contract terms.

Certain kinds of AI solutions are easier to rip and replace – like say an internal tool with a handful of tech-savvy users – compared to other solutions like a supplier portal with 100s of external companies needing to be onboarded.  

Contracts can either lock you to a vendor or give you flexibility if a new and better option appears. If implementation has high costs, your organization may hesitate to pay those sunk costs again in the future. Additionally, some kinds of implementation costs are useful for the next project (say, data cleaning work), while others will need to be repeated for the next tool (like integrating your system with theirs).

AI offers new opportunities to streamline work and improve results. But it’s vital that organizations really examine their goals and options.

If you’d like us to help you vet a potential solution or identify which problems to solve with AI, feel free to hit the contact page and schedule an initial discovery call.