AI’s Next Phase: How Digi-Key Is Deepening Its Expertise With Digital Team Members

The author said that over many years in the technology sector, he has experienced the changes brought by personal computers, the internet, cloud computing, and remote work as they reshaped the way companies operate. He said these shifts have consistently been accompanied by anticipation, expectation, and considerable interest.

He explained that AI is similar in many respects to major past shifts, but the difference is the speed of change. Most new technologies take years to mature, he said, but AI feels like it changes every week.

He said that pace makes it tempting to make bold predictions about what 'jobs' will look like in 5 or 10 years. But the simpler question, he said, is what can be done with the technology right now, and Digi-Key is looking for the answer to that question in 'complexity.'

Through its e-commerce platform, Digi-Key sells more than 17 million products, and its teams process large volumes of information every day. The information they handle includes product specifications, pricing structures, trade compliance requirements, manufacturer data, and customer inquiries.

In this information-processing work, there is little difficulty due to a lack of expertise, but the bottleneck lies elsewhere. Highly skilled employees are spending too much time gathering information and relatively little time applying the knowledge they already have.

AI is presented as the means most likely to create the biggest change at this pain point. The speaker said he and his team refer to it as a digital twin intern, a concept that refers to an intern who is always willing to help. This digital twin intern is described as someone who does not avoid repetitive work and can sort vast amounts of information in seconds.

AI agents can potentially perform meaningful work, but that requires guidance and context, as well as review of the work by an experienced person, he said. It also requires the final decision of an experienced person. The view of AI agents is that they provide experts with a different starting point.

Previously, employees spent time searching through data across multiple systems, comparing information, and handling routine administrative tasks. By contrast, AI agents help experts by enabling them to begin work from a foundation of tasks that have already been substantially completed.

This kind of support leads to a reallocation of the time once spent on procedural work. That time can be used for situational judgment, problem-solving, and customer communication.

The important task is to think about what humans can do when AI removes some of the parts people find difficult.

The tasks AI can perform are referred to as low-value work, and the proposed approach is to hand these low-value tasks to a 'digital twin intern' or an AI agent. As a result, skilled employees can focus more on solving difficult problems, delivering customer service, and driving innovation.

The common perception of Digi-Key's AI strategy focuses on future possibilities and AI-driven changes in distribution. But Digi-Key's real focus is solving practical business problems in the present, and the purpose of that problem-solving is to save people time.

Over the past year, Digi-Key applied a 70-20-10 investment model. About 70% of its AI-related efforts focused on expanding operational capabilities, 20% on improving the customer experience, and the remaining 10% on experiments and long-term opportunities.

The resource allocation plan reflects a core value identified in the process of AI development. That core value is removing friction from everyday work.

Digi-Key is using AI agents. The areas of use are accounting, customer operations, and trade compliance.

In these work environments, employees spend a great deal of time gathering information and moving between systems. They also spend a great deal of time understanding complex work-related regulations. In many cases, the work itself is not highly difficult, but the volume is large and the surrounding environment is complex. Trade compliance was cited as one such example. The subject of tariff impact analysis is several million products, and evaluating tariff impact requires considering the complex relationships among manufacturers, manufacturing facilities, regulations, and customer requirements. As a result, the number of possible scenarios grows quickly.

AI can review vast amounts of information quickly and identify the most important details. For that reason, it is especially well suited to this kind of environment.

Earlier this year, Digi-Key introduced an AI agent into its accounting department and put it to work on live operations. The first success case involved support for processing customer remittance requests.

The AI agent reviewed emails, identified the requested task, and gathered related information. It then recommended the appropriate decision to employees and helped drive approval of the task, and about two-thirds of remittance requests are now automated.

As a result of this deployment, productivity improved, and the process also yielded additional insights and lessons. Before building the AI agents, Digi-Key had established infrastructure for responsible operations, as well as governance, security controls, and operational processes.

Through this deployment, Digi-Key identified where the technology is strong and where it is difficult. It also identified areas where human oversight is essential.

The lessons and insights from the initial experience can be reused, and the opportunities are not limited to a single use case. Digi-Key is rolling out more than 70 use cases across the company, with the level of adoption varying by case.

Once the patterns are understood, new possibilities expand. In this context, recurring companywide questions are raised, centered on identifying tasks that involve gathering information across multiple systems and making decisions based on established business rules.

These tasks are often well suited to AI-assisted workflows. That makes it possible to expand adoption across the organization.

However, the most difficult part was the production deployment of the first operational agent. Digi-Key is now able to focus on scaling what it has learned, and the hardest challenge is not the technology but adaptation.

As AI technology continues to advance rapidly, what would have been unimaginable a year ago may soon feel routine when paired with capable agents. As a result, the biggest challenge is shifting away from the technology itself and toward helping people adapt to changed ways of working.

In the past, most work hours were consumed by administrative tasks. But as administrative tasks are gradually automated, a new question arises: how should the time created by automation be used?

Answering that question is difficult. In particular, spare time can easily be mistaken for idle time in the system, but real spare time is not idle time.

Where mornings used to be spent searching for data, that time now should be used to detect missing reports, help manufacturers find solutions, handle customer requests, communicate with customers, and think more deeply about immediate issues.

The shift in how time is used as a result of AI adoption is not unique to Digi-Key. Organizations across the board that are experimenting with AI are expected to face similar changes.

As a result, the central discussion is shifting from technology to culture and leadership, and to what it means to 'do a good job' once there is no longer a need to perform simple repetitive tasks. Once repetitive work is no longer necessary, the meaning of 'doing a good job' needs to be redefined.

Organizations that find the answer can gain an edge over those that merely buy tools. This shift is tied less to the technology deployment itself than to how an organization judges what counts as performance.

The No. 1 trend to watch in the AI market is the rapid progress of open-source AI models. Most people think of a few well-known companies when they think of AI, and the platforms of those well-known companies remain important.

That said, high-performance specialist models for specific fields continue to emerge. These specialist models are usually much cheaper.

The scope of discussion around AI is expanding beyond technical issues to include economic ones. As a result, the importance of questions related to the economics of AI is also growing. In this context, choosing the right model is posed as a question similar to choosing the right tool.

For some tasks, a general-purpose model may be appropriate. For others, a smaller model specialized for a specific function may be used. These specialized small models can deliver similar results on other tasks while offering a much lower cost advantage.

The scale of AI adoption in organizations is expanding. As AI adoption grows, the importance of deciding which model to choose is also increasing. The point is emphasized that the companies that create the most value with AI are those that understand which models are appropriate for each problem and make those decisions consistently.

The speaker said he is often asked what AI will look like in 5 or 10 years. In that context, he said, he is reminded of similar past discussions surrounding personal computers, the internet, and cloud computing.

This association leads to the idea that after a certain point in technological development, the technology itself recedes into the background. Eventually, once a certain stage is reached, AI is also presented as a case where use comes to the fore more than the technology itself.

Mentioning the use of the internet at work in a self-introduction did not become common, nor did describing a laptop as an innovative strategy. The internet and laptops became a natural part of the way work is done, and AI is seen as moving in the same direction.

Accordingly, it is suggested that the future's exclusive domain may not belong only to organizations with access to AI. That is because AI access is gradually becoming more widespread. Therefore, the key to differentiation lies in understanding how to apply the tools carefully, safely, and responsibly.

At the same time, the importance of distinguishing between areas where automation creates value and areas where human judgment is essential is also increasing. The argument is that competitiveness depends less on access to AI itself than on how it is applied and how the boundary between automation and human judgment is set.

Existing rules continue to apply throughout this process. The issues organizations must review are governance, security, spending, and accountability. The effect of AI is not the disappearance of those responsibilities; rather, it increases their importance.

The speaker said there are many discussions about AI, but he acknowledged that AI can assist with many tasks, including supporting information processing and helping respond to complex situations. Still, he was convinced that the most important variable is people, and that human expertise will remain central, with the judgment of seasoned employees especially remaining the most important part.

Source: TECHWORLD · Lee Gwang-jae
Original: https://www.epnc.co.kr/news/articleView.html?idxno=407741

References

• https://www.epnc.co.kr/news/articleView.html?idxno=407741