Building an AI Tech Stack Isn’t Just About Technology

Welcome to Leading Disruption, a weekly letter about disruptive leadership in a transforming world. Every week, we’ll discover how the best leaders set strategy, build culture, and manage uncertainty, to drive disruptive, transformative growth. For more insights like these, join my private email list .

Let me get this out of the way: I’m not a technology expert.

Nonetheless, business leaders confronted with today’s complicated generative AI landscape often ask me for advice about technology:

“What technology do we need in order to be able to best use generative AI?”

“How do we build our tech stack as an organization?”

“Can we possibly maintain what we’ve created when everything changes on a dime?”

How can I possibly advise these leaders?

Here’s the good news:

The answers to their questions are not about tech. In fact, when thinking about the stack, don’t approach it with a “technology first” mindset.?

The Generative AI Tech Stack Process

Before you even get into the technology weeds when developing your organization’s AI tech stack, it’s vital to consider and tackle a few key things first:

1. Identify Your Priority Use Cases

Ask yourself: What do you actually want to do with generative AI? What are the use cases? If you know what the most important use case is then you can build your tech stack based on those use cases. There’s a risk of entering what I call “use case hell” because there are so many possibilities, so do your best to determine which ones hold the most value and possess the potential to create a significant competitive advantage to your organization.

Once you have a solid list of these priority use cases, it’s time to classify them into similar categories. Figure out which ones will be relatively easy lifts that you can do right now with publicly available tools. Especially for use cases, center on content creation—marketing content, sales proposals, code editing, and image design. Using off-the-shelf tools like jasper.ai , Adobe Firefly , and Github Copilot with clear guidelines and training can deliver immediate returns.?

The AI tech stack is continuously evolving, and improvements are happening rapidly. By tackling the low-hanging fruit first, you can start reaping benefits while still anticipating upcoming advancements. (Keep reading for more details on building a proprietary AI tech stack.)

2. Upskill Your Workers

Work with your HR team—and with your risk and legal teams—to identify all the policies and training that’s going to be needed as you build your AI tech capability.?

Upskilling, in particular, will be essential. For instance, coders who are no longer burdened with writing the most basic code may be tasked with writing more complex code, but their ability to understand what is good code may not be as strong. They might not have developed these skills yet. Meanwhile, the workers who are going to do really well in this environment will be those with a slightly higher expertise level—who can formulate good prompts, understand what quality looks like, and be able to leverage that into a deeper level of analysis and innovative work. The most junior people will not necessarily have those skills, so being able to upskill them very quickly is going to be pivotal.

3. Create a Center of Excellence

What’s a center of excellence? Simply put, it’s a community of practice, or a cross-functional team that combines business perspectives with technical expertise. This centralized team should be able to look at existing infrastructure—including all the tech and data—and know how it all works together. I can’t emphasize just how important this collaboration is to your organization’s generative AI efforts.?

If you leave it to just a tech team, they won’t have the context necessary to understand the business use cases and create the best possible platform. And, of course, if you leave it to the business side, they won’t have a strong enough understanding of how the technology works to be reasonable in terms of what to expect. A cohesive approach among departments and stakeholders ensures that the tech stack is employed with a deep understanding of the business's needs and potential implications. It also helps in managing expectations, making realistic demands, and achieving successful outcomes.

4. Integrate With Other AI Models

More customized use cases, such as putting a generative AI front end on top of proprietary company data for internal use, requires advance planning that should begin now. Examples include connecting internal support documents and FAQs to customer chat history to create better customer support. Or it could be connecting your customer relationship management, marketing automation, and financial systems to connect customer engagement to account profitability.?

Although it is technically possible to build your own foundational model from scratch, let me be clear: Don’t do this. It’s so difficult to create something from scratch, and unless you are willing to put in a hefty investment of time and resources in order to construct a model that is completely proprietary, it’s not worthwhile.

Instead, it may be worth simply waiting for these types of capabilities to be built into the enterprise systems you already have.?

In the meantime, ensure that your data architecture and data itself is updated, organized, and curated in a way that allows for easy integration with future generative AI models. There are countless APIs that are already written out there. Use them. Got technical debt? Pay it.?

Here’s a closer look at the generative AI tech stack for organizations:

5. Train and Maintain

Finally, some classic business advice: Make sure you have a process in place. Allocate reasonable budgets, set achievable timelines for development, and emphasize continuous maintenance to adapt to changing data and capabilities. Because the technology is ever-changing, and the interactions are constantly changing, the training the model is also going to change the level of quality responses you get. A step further, constantly maintaining your organization’s generative AI tech stack is going to be an important part of that.

Your Turn

What kind of generative AI technologies are you using on a regular basis? If you are using publicly available tools, how are you using them to protect your company's data? If you’re trying to leverage your organization’s internal data, how are you approaching that?