How to Pick a Generative AI Partner

(jittawit21/Shutterstock)

Back in July, I had a briefing from Dell on its impressive new generative AI program which is tied tightly to Nvidia for that technology but uses Dell’s extensive knowledge of how to use this innovative technology for customer outreach. The program is robust, scalable and clearly had a lot of thought put into how it was put together. It would not be appropriate yet for someone looking to use generative AI to increase their internal productivity since Dell’s strength and knowledge was almost exclusively focused on customer engagement.

This is not unusual. Vendors often begin selling a technology before they have truly embraced it themselves and, right now, Dell’s internal efforts are mostly focused on customer engagement, not enhancing rank-and-file employees. With respect to Dell, this means that asking them to do anything but customer engagement for generative AI would have them learning on-the-job which is typically expensive and less than ideal.

Let’s talk about picking a generative AI vendor this week.

Baseline Generative AI Knowledge

Before you even start on this path, you will need to spin up some internal people on the different generative AI platforms and how they are being used successfully. Without that baseline knowledge, even hiring experts to help you can be a problem because you have not yet built a knowledge framework upon which to measure the vendors you are considering.

(Marko Aliaksandr/Shutterstock)

Some of the most expensive, painful and failure-ridden projects were committed by companies that just didn’t understand the technology they were trying to deploy. The other reason for initial internal experts is that no external expert is likely to know the unique dependencies and problems with deploying innovative technology in your company. This knowledge must be part of the decision process, or the resulting technology choice is either excessively hard to deploy or undeployable at all.

If your people understand the technology, they are far better able to catch a vendor who is unqualified to bid on a project, let alone try to build it themselves at great expense.

Pick Vendors Who Have Demonstrated Competence

When I say “demonstrated competence,” I mean in the area(s) in which you plan to deploy generative AI. For instance, if you want to use it in HR, look for a generative AI vendor who has deployed this technology in its own HR organization. This assures you they have learned how to do this on their nickel, not yours, and should be better able to anticipate problems rather than just react to them, resulting in a faster, better, cheaper and higher quality result. Reference accounts are also a critical part of assuring competency, but if the firm has not done this in their own company, there is a good chance there are hidden problems in their approach that you will want to avoid.

If the vendor has employed it, you can also ask that your stake holders be allowed to interface with a safe (not tied to real data) version of the product as the vendor deployed it and to chat with the operators of that internal application to learn about problems you will want to avoid doing your own deployment. Often, reference accounts don’t feel comfortable with another company looking at how they handle core functions like HR, but the vendor should be more open to letting you chat with their people. In that exchange, you’ll want to confirm that the users have some input into future versions of the offering. If they don’t, you’ll likely want to pass on using that particular vendor because they’ll be less able to identify and correct problems before you see them.

Assure the Supply Chain

As noted above, Dell’s program is heavily focused on Nvidia hardware which is fine for most of the western world but could be a problem in some eastern countries currently hit with U.S. technology sanctions. And since there are concerns about AI being compromised to create a security problem for customers, it would be wise to choose solutions that are from vendors who operate freely in your geography so that future government actions won’t jeopardize the completion of the project, maintenance or future upgrades. Generative AI deployments tend to take a while, and you do not want yours to stall because of parts unavailability.

(cybrain/Shutterstock)

Data

One of the biggest expenses in a generative AI deployment is training the AI. In this case, a vendor like IBM watsonx, which already has a massive amount of data in certain targeted verticals, has an advantage if your project is in one of IBM’s verticals. Having training data can vastly reduce the cost of a deployment and speed up the implementation significantly. You may pay more to engage a vendor who has training data, but the extra cost is a fraction of what it would take to generate all your own training data.

Often, training a new AI System can cost several times what purchasing the hardware costs, so a vendor with relevant training data for your unique project would be invaluable.

Wrapping up:

One of the most painful lessons in my career was the client/server wave that took place when I first started working for IBM. A decade later, the research firm I was then working for discovered that most of those early client/server deployments were expensive failures that the firms, both buying and selling, were embarrassed to talk about.

The three things contributing to the failures were that the bidders did not understand the technology, the people getting the bid did not use the technology well internally and, frankly, the technology was not ready.

Generative AI is very new. It is not ready for everything we want to do with it yet. It is growing in capability very quickly and there is no doubt in my mind that, eventually, you will want to deploy it. However, your best bet for a partner is a firm that has deployed this technology internally, that has been using the technology long enough to understand it and develop reference accounts, and who has deployed the technology both internally and externally similarly to how you want it deployed.

According to Wharton, generative AI can provide between a 30% and 80% productivity increase but only if it is properly deployed and your employees use it effectively. The right vendor can assure both, so put your efforts into making sure you select a vendor that understands the technology, has deployed it internally similarly to how you want to deploy it, and who can assure their supply chain so that you aren’t left hanging if a critical part falls on the wrong side of politics.

About the author: As President and Principal Analyst of the Enderle Group, Rob Enderle provides regional and global companies with guidance in how to create credible dialogue with the market, target customer needs, create new business opportunities, anticipate technology changes, select vendors and products, and practice zero dollar marketing. For over 20 years Rob has worked for and with companies like Microsoft, HP, IBM, Dell, Toshiba, Gateway, Sony, USAA, Texas Instruments, AMD, Intel, Credit Suisse First Boston, ROLM, and Siemens.

Related Items:

The Three Approaches to AI Implementation

Should Employees Own the Generative AI Tools that Enhance or Replace Them?

Wrapping our Heads Around CES 2023 and Next Generation Applied AI