Skip to content
Getting Started With Generative AI
Tony OlsonSeptember 19, 20235 min read

Getting Started With Generative AI: 3 First Steps for Data Leaders

All Aboard the Generative AI Hype Train

Generative AI has taken 2023 by storm. SaaS companies across the globe are pouring in billions of dollars to enhance their digital products with various Generative AI features. Entire R&D teams are getting refocused on this capability.

If you feel like this is the next blockchain, don’t. This is not a technology looking for a use case. Instead, there are thousands of use cases out there looking for this technology (image creation, text generation, code completion, etc.)

If you feel like you are being left out, don’t. Many companies are just considering getting started with Generative AI (and AI in general). As they begin this journey, many data leaders feel overwhelmed with how quickly Generative AI has gone from concept to reality. It can be challenging to figure out where and how to begin as they sort through the use cases and applications to their own business. In this article, we offer three concrete steps to beginning your first Generative AI project.

Set Expectations

This Generative AI train is moving like a rocket, so fast it can be dizzying to keep up. The new buzzwords range from MosaicML and LLaMA to DALL-E 2 and Bard. Each comes with its own specialty performance and required architectures. This isn’t to mention all of the SaaS companies that will be embedding these types of Generative AI models into their new 2024 products.

When you consider the potential value Generative AI could bring to your organization, regardless of the speed of the current Generative AI market,  it’s important that you succeed now. Without early Generative AI wins, you risk disillusionment with the business, a lack of trust in the solutions you build, and a lack of innovation coming from your data team.

There is also a competitive advantage to be achieved with early Generative AI successes. We’ve seen some of our clients shift 25% - 50% of their R&D resources to embedding Generative AI into their products and organizations.

In order to succeed, it is important to set the following expectations with C-level stakeholders before you start:

  • Prepare for disruption. Things will continue to move fast. The decisions you make today will be different six months from now. It doesn’t mean that you are ill-informed today. It’s simply a byproduct of the current market and this rapidly evolving technology. Make sure stakeholders understand this. Prepare for this disruption by building flexibility into the changes that you make. When you are building your first Generative AI solutions, consider building them on a flexible platform (versus going the route of point solutions). For example, LLMs are becoming more niche. You will want a platform that allows you to easily swap niche models and products as part of your data products and solutions. 
  • Prepare for opinions. There are a lot of opinions out there such as architecture, deployment mechanisms, build vs. buy decision, and solution scalability. Developing relationships with a trusted advisor or guide in the industry can help you not only learn more about a variety of approaches but also provide perspective on which of the many options might work best for you. Finding additional guidance in this way can help you take in the opinions of others and construct your own mental model that you can clearly communicate and execute.
  • Prepare for a rollercoaster of emotions. The results of these solutions can be shocking! Even jaw-droppingly amazing. At the same time, they can also be laughably incorrect or just plain wrong. Make sure your stakeholders know that it's normal to see both at first. One technique we use to normalize these early highs and lows is to deploy a user feedback mechanism in the solution. Allow your users to easily inform you if the result was perceived as inaccurate or bad.

Select a Use Case for Generative AI

Selecting your first use case can be a difficult process. There are a lot of options out there and their importance differs from company to company. We coach our clients to consider the following criteria:

  • How visible is the use case? You want to have strong stakeholders at a variety of levels. If the use case is successful, what value does it bring to the company? The results should be meaningful.
  • Can you fail fast? If it’s not going to work, you don’t want to find out months into the future.
  • Can you build upon previous work? One of our favorite things to do is to build upon analytics solutions that have already been created and are being utilized by the business. You already know the data is right, so you can contain the scope of the engagement to just the application of Generative AI technology for the business to utilize.
  • What are the data security implications of executing this first project? Don’t try to come up with an entire Generative AI security architecture right away. Define what you are comfortable with for this first engagement with the understanding that things will change over time.

At Snow Fox Data, we’ve found success starting with Natural Language Query (NLQ) use cases. NLQ typically has strong stakeholders. Who wouldn’t love to ask a question against their own data? The results are meaningful because it saves analyst time and allows additional sharing of information across the organization in an easy and effective way. Additionally, NLQ use cases can be put into production quickly by only utilizing one table of data and establishing simple prompts.

Educate on the Results

Remember that you aren’t just educating your company on a solution. You are educating them on a new and disruptive technology. Work against the hype and toward the unique impact at your organization by sharing the highlights and lowlights of the effort.

We coach our clients to build presentations for executives and business leaders that include the following:

  • A summary of the use case. What were you looking to solve?
  • What worked?
  • What didn’t work?
  • What took more time than expected?
  • What tools did you use? What were their alternatives?
  • What were the costs? Make sure to balance costs vs. savings. Much of Generative AI’s business value can be found in efficiency and labor savings.
  • Does the solution scale?
  • What recommendations do you have for future Generative AI efforts? We make specific recommendations for people, processes, and technology.

Starting Your Generative AI Journey off Right

Remember that although the Generative AI hype train is real, it also can solve a variety of impactful use cases with enormous business potential. Organizational investment into this technology has already hit the mainstream, so now is a great time to explore the options. To get started, you need to set the appropriate expectations, select the right use case, and educate executives and business leaders on the results that will help drive results for your business.

You can also check out this post on


Tony Olson

I help companies make data-driven business decisions. I care about unbiased positive business outcomes through data initiatives. Whether it's through ML/AI or basic descriptive reporting, I believe the capability to execute data science initiatives gives companies a competitive edge and am passionate about helping companies build that capability.