The Distribution Blog

Navigating the AI Landscape: Distinguishing the “Locals” from the “Tourists”

November 3, 2023

Table of Contents

We recently received a great question from a newsletter subscriber: “Do you have any personal recommendations for staying on the front edge of AI and distribution?”

AI is the talk of the town these days. It makes sense distributors are trying to sort through what AI means for their businesses.  

With AI exploding in popularity, it seems like every other software company now claims to be “AI-powered.” With so much hype, how do you distinguish genuine innovators, “AI locals,” from those jumping on the bandwagon, “AI tourists?”  

We started helping distributors find their footing with AI in 2018, so we’ve seen enough to quickly identify the differences between the two groups. Let me be your tour guide to navigate the AI landscape in distribution.

Spotting AI “locals”

AI “locals” are companies that were early adopters and innovators in the AI space. They’ve used AI to solve real business problems from the onset, or they embraced the paradigm shift at the first opportunity and did the work to understand its capabilities and limitations.  

In 2018, when our CEO, Benj, was studying data science at Harvard, he realized distribution companies — including his family’s dental supply business — were at risk of getting squeezed out by technology. He wanted to help the distribution industry use technology to thrive.  

He teamed up with engineers and long-time distributors to create a product to make selling easier. AI was the tool that made it happen. Distributors are especially well-positioned to benefit from AI for several reasons:  

  • They have access to massive amounts of data, perfect for training AI models
  • They’re not restricted by regulations like in healthcare, allowing them to use their data freely
  • The risks of using AI are minimal, unlike in fields like social media; and there’s a significant gap between the technology they use and what’s possible with AI, leaving room for transformative growth.

AI “locals” understand AI is just a tool and the real value of any software solution is in its ability to solve a problem, like driving organic growth or transforming a reactive sales rep into a proactive one.

Understanding AI “tourists”

AI “tourists” are companies that have rebranded to appear as AI innovators without the foundational work. They may have a nice tan, but they’ve only just arrived at the AI beach party. Here’s a case that reveals why it’s vital to look beyond the hype:

During an investigation into a healthcare organization’s use of five AI tools, MIT Sloan found that several AI tools in the healthcare space performed “extremely poorly” in pilots despite lofty promises and headline-grabbing claims of the companies that sold them. MIT Sloan argues that understanding the “ground truth” of an AI tool is essential to assess its quality and how it may benefit or harm decision-making in an organization.

“Ground truth” is information known to be true, based on evidence. In AI, it refers to the data used to train the algorithm. For example, to suggest complementary items to customers shopping on a distributor’s website, the algorithm must be trained on the distributor’s product data. By training on product data, AI can learn to recognize that Brand A’s chemical hose is a suitable substitute for Brand Z’s chemical hose and that when someone adds a floor scrubber to their cart, they should also purchase cleaning solutions or accessories to go with that floor scrubber. The quality of ground truth is essential for the accuracy of the AI tool. For distributors, understanding and questioning whether the “ground truth” of AI software is indeed true is vital.

In the MIT Sloan study, one company claimed its AI tool was better than humans at diagnosing cancer. A healthcare team planning to trial the tool was shocked to find that the AI tool’s ground truth was not aligned with professional standards. The AI tool was tested with a single mammogram, whereas professionals would usually use more extensive data for diagnostics. The tool’s performance seemed impressive but was based on inadequate “ground truth” because the company was unwilling or unable to invest in the high-level data collection and analysis that good AI requires. Tools that rely on superficial or weak data can lead to poor decisions and significant risks.

Some “tourist” companies aren’t using AI at all. Instead, they’re doing basic data analysis, adding a bit of context and attractive visualization, and calling it AI. If the technology they’re marketing isn’t iterative — if it doesn’t analyze large amounts of data and then use what it has learned to become smarter and more capable — then it’s not really AI; it’s just standard data analytics or business intelligence software. True AI should be able to leverage all of your company’s data and use more than a few parameters to make intelligent decisions. And if you’re a distributor, that means massive amounts of data and thousands of parameters.

How to determine if an AI vendor is a local or a tourist

There are several questions you can ask to determine if a prospective AI vendor is an AI local or a tourist. Anything you can glean about the amount of data the vendor collects and stores, how they analyze it and how they use that analysis to solve problems will give you a much better idea of whether they’re tourists or locals.  

Do you build your own AI models or use third-party services?

Why Ask This: Knowing whether the company uses its own AI or a third-party service like OpenAI or Azure can give you insights into their expertise and control over the technology.

Good Answer: “We have an in-house team that builds and fine-tunes our own AI models, allowing us to tailor the software specifically to the needs of our customers.”

Not-So-Good Answer: “We rely solely on third-party services like OpenAI and Azure. Our focus is more on implementation rather than developing AI technology.”

If you build your own AI, how do you ensure quality data?

Why Ask This: Quality data is crucial for effective AI. You want to know how they prepare and clean the data they use.

Good Answer: “We use automated scripts and manual reviews to clean and standardize data. Techniques include handling missing values and normalizing numerical data.”

Not-So-Good Answer: “We use the data as it comes in and make minor adjustments. Our system is robust enough to handle some inconsistencies.”

How do you transition your AI from testing to live customer use?

Why Ask This: A smooth transition from testing to live use is crucial for a reliable service.

Good Answer: “We have a multi-stage process that includes extensive testing in a controlled environment. We use CI/CD tools like GitHub Actions and MLOps tools like Kubeflow to ensure a smooth transition.”

Not-So-Good Answer: “We have a basic testing phase and then we roll it out. We rely on ongoing monitoring to catch any issues.”

If you use third-party AI like OpenAI, how do you make it work for your customers?

Why Ask This: You want to know how they’re adding value and not just serving as a middleman between you and a third-party AI service.

Good Answer: “We integrate OpenAI’s models with our own data systems, using tools like Langchain to ensure the insights are relevant and specific to our customers.”

Not-So-Good Answer: “We plug into OpenAI’s API and use their default settings. It’s generally good enough for what our customers need.”

Are you planning to build your own large-scale AI models?

Why Ask This: This gives you an idea of their future plans and commitment to AI technology.

Good Answer: “Yes, we’re planning to build our own large language models to offer more personalized and secure services.”

Not-So-Good Answer: “No, we find that existing solutions like OpenAI and Azure meet our needs. We don’t see the need to invest in building our own models.”

How do you protect our data if you’re using third-party AI services?

Why Ask This: Data security is paramount, especially when third-party services are involved.

Good Answer: “Before sending any data to third-party services like OpenAI or Azure, we implement robust data anonymization and encryption processes.”

Not-So-Good Answer: “We rely on the security protocols of third-party services like OpenAI and Azure. They’re reputable companies, so your data should be safe.”

The investigative report in Sloan Management Review also offers some practical advice. When engaging with AI vendors, the researchers recommend “having open conversations about their “ground truth” selections, their logic behind those choices, and any trade-offs they considered.” If the vendor dances around your questions, and if they aren’t transparent about the data and metrics they use to make their claims, you should see it as “a serious red flag.”  

For distributors, it’s especially important to find an AI vendor that has built a distribution-specific solution. Even if they do a good job of collecting data, using key parameters and establishing a legitimate “ground truth,” a generic solution is likely to fall short of solving a distributor’s unique problems. AI solutions must be built for a specific use case; if not, it’s known as “weak AI.” Beyond being able to handle your large product catalog, there should be an omnichannel element to their AI tools, and they should be useful for both ecommerce and offline sales.  

Welcome to the world of AI … stay a while!  

True AI-powered tools can help distributors thrive, but make sure you’re doing your research and spotting those touristy red flags before you commit resources to a particular solution. A real conversation with a potential vendor can go a long way in determining whether a company is a tourist or a local. Are they pulling from their own experiences with AI? Or are they regurgitating buzzwords and talking points? Are they passionate about the problem-solving capabilities of AI? Or are they just trying to make a sale?

Are you curious how we responded to the newsletter question we shared in the intro? Here’s our answer:

“We have data scientists who keep up with the latest developments in AI and a team focused on understanding the needs and challenges faced by distributors.

Our data scientists are always tuned into the broader world of AI and what visionaries like Sam Altman, Andrej Karpathy and others have to say. But the rest of us are fully dialed into the world of distributors. We’re all about identifying their pain points, understanding their goals and figuring out how AI can simplify their lives.”

Our team is happy to answer any of your questions or talk with you about how others in the distribution industry are using AI. We’ve been helping distributors find their footing with AI since 2018. If you’re interested in speaking with us, request a conversation here and someone from our team will get in touch.

Check out our other blog articles

Ready to make Proton your secret weapon?