AI & Management: Predicting the 3-Year Future | Dentsu Digital

Mr. Satoru Yamamoto profile
Majored in artificial intelligence (AI) under Professor Yutaka Matsuo of the University of Tokyo. Established Data Artist Co., Ltd. in 2013, and merged with Dentsu Digital in 2023. Utilizing AI and big data, we provide numerous digital marketing services such as automatic generation of advertisements, prediction of advertising effectiveness, CRO and SEO. He has appeared on many media such as TV programs, and has given presentations at seminars at companies and universities. His major books include “How to create logic that sells” (Senden Kaigi) and “AI x Big Data Marketing” (Mynavi Publishing).

CAIO drives AI strategy at the company-wide level

In January 2025, Dentsu Digital created a new position called CAIO, which is still rare in Japanese companies. Mr. Yamamoto explains the background as follows.

“In order for Dentsu Digital to continue to provide value in the future, it is necessary for all employees to achieve high productivity and performance while benefiting from AI as a matter of course. Therefore, up until then, I had been leading AI initiatives in the position of executive officer in charge of AI, but I have created the position of CAIO as a control tower for deploying AI as a company-wide strategy.”

Mr. Yamamoto has three main missions as CAIO. The first is to find ways to use AI that match the capabilities of the company, and to predict the future and suggest the best way to use AI to increase corporate value. Second, we should propose specific requirements to AI provider companies and work together to increase value. The third step is to support company-wide implementation.

With this aim in mind, in conjunction with the establishment of CAIO, we also formed a cross-sectional organization called “AI Native Twin” that aims to standardize the implementation of AI across the company. Additionally, an AI committee has been established, with a product subcommittee, project subcommittee, and governance subcommittee.

Under this structure, we will cover all areas of strategy, tactics, and combat related to AI. “We work with each department to develop a concrete action plan, monitor it daily, and make corrections if there are any areas for improvement,” Yamamoto said. They are also actively disseminating information as PR.

Dentsu Digital assigns chief positions to priority areas of reinforcement, so there are only two positions with a “C”: Mr. Yamamoto’s CAIO and CSO (Chief Sales Officer). However, in a typical company, collaboration with the CIO (Chief Information Officer), CTO (Chief Technology Officer), CDO (Chief Digital Officer), etc. is required.

Regarding relationships with other executives, Mr. Yamamoto uses an analogy to cooking, explaining, “If AI is a chef, data is the material that AI handles. No matter how good the chef is, he or she cannot peel a potato with his bare hands, so a foundation, or kitchen, is needed to process the ingredients correctly.The kitchen is the domain of the CIO or CTO.The CIO’s role is to decide how to process the ingredients.”

The domestic Dentsu Group, to which Dentsu Digital belongs, launched the “dentsu Japan AI Center” in July, with Unit 2 in charge of developing AI solutions, Unit 3 in charge of developing AI agents, and Unit 4 in charge of data collection. Here, Unit 4 plays the role of CDO, and is asked to prepare data that can be used securely.

Three areas of application in line with technological evolution

Dentsu Digital has a wide range of businesses centered on digital. Among them, Mr. Yamamoto introduced efforts to utilize AI in the advertising and marketing field.

The use of AI at Dentsu Digital is being rolled out in stages in three areas in line with the evolution of technology. The first is creative, the second is marketing planning, and the third is new experience design. “It’s expanding from 1 to 2 to 3 in line with the order of AI evolution.”

For the first creative generation, we will release “∞AI Ads” in 2022. It is said that the advertising effectiveness improvement rate for clients who have introduced the service has already reached an average of 150%. When we conducted an internal experiment in which young copywriters were provided with an AI that had learned the know-how of experts, performance improved significantly. Still, the highest score was for humans. “It seems like there is a conflict between AI and humans, who want to go even further after seeing AI’s answers,” Yamamoto said.

The second field, marketing planning, research, and operations, is an area of ​​application that has been created as a result of the advent of the Reasoning Model and agents, which have made complex thought processes possible. For example, when it comes to research, they are able to move forward with planning by adding information from the web through deep research, and by using agents to incorporate in-house information. In the end, they have to be creative in deciding who to send the message to and with what kind of message.In this process, they create an AI persona that has learned the Dentsu Group’s survey data, and then develop ideas while having conversations with that persona.This is becoming more and more popular.

The third area, experience design, is a relatively new area. For example, by making the basic model multimodal, you can use “any to any” to interact in any mode, such as audio, image, or video, and create new experiences such as virtual try-on. Rapid progress is being made in this field. The company is also considering trends in new devices, such as the Ray-Ban Meta Glasses, a smart glass developed by Meta in collaboration with Ray-Ban.

Prepare more than 50 AI tools and encourage the use of AI for value creation

In order to safely and effectively promote such diverse AI applications, CAIO is working to create an environment. “The development of AI is exceeding expectations. In particular, generative AI in marketing is progressing faster than expected,” said Yamamoto. He says that creating an environment within the company is important in order to lead customers.

Dentsu Digital currently uses more than 50 commercial AI tools in its operations. The aim is to encourage the effective use of AI by giving employees the freedom to use each AI tool to its strengths and characteristics.

In utilizing AI technology provided by vendors, Mr. Yamamoto believes that a cooperative relationship with AI vendors is important, as stated in his mission. The most advanced AI comes from Big Tech in the United States, and its evolution is rapid. This is why we believe that a good relationship with Big Tech is more important than ever.

In order to strengthen our relationship, we actively share with AI vendors how we use AI in our company, receive advice on how to use it, and receive information on the latest usage methods. The company also sends feedback on features and needs.

“We have a system in place to communicate specific requirements for the necessary technology and products to platformers and follow up,” Yamamoto said. “Development costs money, people, and electricity. We have to make something good with limited resources. That’s why we want to convey that everyone will be happy if we have something like this.”

Regarding in-house awareness and education, in addition to building a foundation, we hold company-wide seminars multiple times a year and set up an AI consultation desk to help employees learn the basics of how to use AI. However, that alone would be at the level of using AI to create emails. “We want to use it in a way that leads to value creation,” said Yamamoto.

Therefore, we first held an ideathon for the entire company. Approximately 500 ideas were collected, and in addition to selecting the winner in the final selection, we created a collection of ideas organized by genre. “It’s packed with hints on how to use it in your work and what you can offer your clients.What I want to promote is not just learning how to use it, but thinking about what it can be used for.”

Governance system: Risk management that separates data and AI

In parallel with activities to support company-wide introduction of AI, we also perform management. “We strictly control who can use which AI,” said Yamamoto. AI tools are advancing rapidly, and some of them are experimental, so they have set rules for each AI, such as allowing them to be used experimentally but not for commercial use, not including data related to privacy, and not allowing data that is not open to be included in an AI whose learning process is unknown.

AI governance is not just about who has access to which AI. As Dentsu Digital’s CAIO, Mr. Yamamoto works closely with the domestic Dentsu Group and is also involved in formulating group-wide policies. “Dentsu Digital uses AI in a relatively large number of cases, so we always inform the governance committee of events that occur at our company and constantly check to see if they are covered by the current guidelines,” he explains.

Data security will be handled by a separate committee. “Even before the AI ​​boom, we had taken measures to consider whether data was considered personal information or whether data could be shared among members working for competing companies.As an extension of this, in the case of AI, the Data Technology Committee has established rules such as not allowing primary data to be directly input into machine learning.”

In order to increase the effectiveness of its governance system, Dentsu Digital also carries out multi-layered technical risk management. The foundation of this is ensuring the transparency of AI models. The company says it always checks with AI tool providers about the source data for learning and what kind of checking system they have in place.

In addition to hallucinations, we will also pay attention to the latest trends, such as the new concept of “Potemkin understanding.” Hallucination refers to a case in which the AI ​​produces a wrong result, whereas Potemkin understanding refers to a case in which the AI ​​misunderstands a concept.

As for these countermeasures, Mr. Yamamoto says, “We will use the guardrail functions provided by the AI ​​provider as a basic premise.We will cover them as much as possible by creating a framework like a standard implementation process.”

Mr. Yamamoto also raised an aspect of AI that is rarely talked about as a risk: environmental impact.

“Currently, the focus of risk discussions is the risk of telling lies and the risk of personal information being leaked.However, there is also a risk that continuing to use (AI) will place a burden on the earth,” Yamamoto points out. AI models require huge amounts of computation. It is said that querying AI consumes 10 times more energy than searching.

“Generally, we judge models only from the perspective of economic rationality, such as looking for a model that gives a good answer quickly and discontinuing it if the cost is high. If we want AI to do the same thing, we can customize the model or distill it to lower costs and reduce the environmental impact. Measure how much our marketing activities are impacting the environmental impact. I want to move further in that direction.”

Predict how management will change with AI three years from now.

In fact, Mr. Yamamoto himself does not know whether the CAIO position will become permanent in the future. “Ideally, it would be successful even if it does not become widespread. To achieve this, top management should declare that the use of AI is a company-wide imperative.” He expressed the idea that creating a CAIO position would be effective in order for AI to cross the chasm.

Mr. Yamamoto offers the following advice to CIOs and other executives: “It is a major premise that we will use (AI). It is the manager’s job to predict what AI can do now and how it will change in three years.”

In order to accurately predict what will happen three years from now, the first step is to break down your company’s business processes. Even if you allocate resources to develop it in-house, the cost may be wasted the moment a platformer launches it, which is the current state of AI. After understanding business processes, we make adjustments while looking at resources, such as where business improvements can be made, how much improvement can be made with current AI, and which business areas should be invested in. “By breaking down business processes, how much efficiency will increase by what we use and where, and how much will the value we can provide increase?We should proceed with this in mind in the current phase.Isn’t it important for management to draw this out in high resolution?” (Mr. Yamamoto)