How a 16-Year-Old Company is Assisting Small Businesses in Adopting AI

Among all the discussions about whether artificial intelligence is experiencing a bubble, the supply chain and logistics sectors are emerging as genuine arenas for its application. Companies like Flexport, Uber Freight, and numerous startups are actively developing diverse applications and attracting prominent clients.

While AI technology is enabling Fortune 500 companies to enhance their profits (and rationalize layoffs to Wall Street), its appropriate use is also benefiting smaller enterprises.

Netstock, a company specializing in inventory management software since 2009, is contributing to this trend. They recently launched a generative AI-based tool called the “Opportunity Engine,” which integrates into their existing customer dashboard. This tool extracts data from customers’ Enterprise Resource Planning software and leverages that information to provide regular, real-time recommendations.

According to Netstock, this tool is saving businesses significant amounts of money. On Thursday, the company announced it has delivered one million recommendations so far, with 75% of customers receiving suggestions from the Opportunity Engine valued at $50,000 or more.

Despite its enticing potential, one of these clients — Bargreen Ellingson, a 65-year-old family-owned restaurant supply firm — was initially hesitant about implementing AI.

“Old family businesses typically don’t embrace significant changes without scrutiny,” said chief innovation officer Jacob Moody in an interview with TechCrunch. “I couldn’t just walk into our warehouse and proclaim, ‘This black box is now in charge of management.’”

Instead, Moody introduced Netstock’s AI as a tool that warehouse managers could “choose to use, or not use,” likening the approach to “carefully testing the waters” of AI.

Moody notes that the AI helps prevent errors by analyzing the numerous reports his staff uses for inventory management. Although he admits the AI-generated summaries are not infallible, he stated it “filters signals from the noise” efficiently, particularly during off-hours.

A more significant change that Moody observed is how the software has enhanced the effectiveness of some of Bargreen Ellingson’s less-experienced warehouse staff.

He mentioned an employee in one of Bargreen’s 25 warehouses who has been with the company for two years. Although this employee holds a high school diploma and lacks a college degree, the time required to train him in all of Bargreen’s inventory management tools and forecasting techniques is substantial.

“But he knows our customers, and he understands what he loads onto the truck daily. With this AI-driven insight, he can quickly assess whether a suggestion makes sense or not,” he explained. “This gives him a sense of empowerment.”

Netstock co-founder Kukkuk acknowledged the reluctance towards new technologies, especially since so many offerings essentially consist of average chatbots bolted onto existing platforms.

He attributes the early success of Netstock’s Opportunity Engine to several factors, including over a decade’s worth of data from collaborations with retailers, distributors, and light manufacturers. This data is securely protected to comply with ISO standards and serves as the foundation powering the recommendations. (Kukkuk mentioned that Netstock employs a mix of AI technologies from both the open source community and private firms.)

Customers can rate each recommendation with a thumbs up or thumbs down, and the models are further refined based on whether the customer implements the suggested actions.

While this type of reinforcement learning can occasionally produce bizarre or negative outcomes in social media, Kukkuk asserts that his focus is on different incentives.

“I’m not concerned with gaining views, you know? Platforms like Facebook and Instagram prioritize that since they want users to engage with their content. Our focus is on: ‘what is the outcome for the customer?’”

Kukkuk is cautious about broadening interactions due to the limitations of current generative AI technologies. Although it may be beneficial for clients to discuss with Netstock’s AI regarding the rationale behind a recommendation, he fears it could result in accuracy issues.

“Navigating this is tricky; the more flexibility you give users, the more freedom a large language model has to generate inaccuracies,” he noted.

This context illuminates the Opportunity Engine’s design on Netstock’s standard customer dashboard. The recommendations are prominently displayed yet can be easily dismissed; this is not a scenario where Google Docs overwhelms users with numerous AI features.

Moody expressed his appreciation for the AI’s unobtrusiveness.

“We’re not allowing the AI engine to make any inventory decisions without human review and approval,” he emphasized. “If and when we reach a stage where humans agree with 90% of the suggestions, we might consider granting more control. But we’re not there yet.”

This marks a hopeful beginning at a time when many enterprises struggle to effectively deploy generative AI.

However, if the technology improves, Moody does express concerns about its implications.

“Personally, I am apprehensive about what this means. There will be significant changes, and none of us truly knows what this will look like at Bargreen,” he remarked. This could potentially lead to fewer data science roles. Yet, he believes that even if such employees transition from the warehouse to corporate offices, safeguarding knowledge remains vital.

Bargreen requires individuals who “thoroughly understand the underlying theories and philosophies to rationalize how and why Netstock makes particular recommendations,” ensuring that they do not blindly pursue incorrect paths, he concluded.