On October 9, 2026, Keisuke Tanaka of Trust said companies planning an AI Center of Excellence (AI CoE) should start by defining their target state and assessing their current position, rather than first setting up the organization, staffing, or tools. The CoE’s three roles are to set direction, establish an execution environment, and spread results across the company.

The design process has three steps: define the desired state and goals, assess the current state, then choose measures and set priorities to close the gaps. The desired state has two parts: how the whole company wants to use AI, and what the AI CoE itself should achieve. The company-wide vision should align with its digital strategy and describe, in business terms, where AI and data can create value and strengthen competitiveness, including revenue, profit, customer experience, and business models.

AI CoE goals can include centrally managing priorities for company-wide AI initiatives, providing teams with the environment and procedures to start new initiatives, and enabling one department’s results to be reused by others. Because digital technology advances quickly and business goals can change, companies should set a “North Star” to show their direction; that guiding point may also shift as technology changes.

Tanaka recommended assessing the current state only after defining the target. Setting criteria first narrows the assessment to relevant areas and helps identify issues without spending time on an exhaustive survey. He pointed to the Financial Data Utilization Association (FDUA)’s Financial Data Utilization Organization Checklist. Although designed for finance, it includes perspectives shared across industries. It has two levels—use, which produces results, and infrastructure, which supports that work—and six themes in total.

The three use-level themes are data used, organization, and business impact. For data, the checklist examines the accumulation and quality management of internal and external data, along with purposes of use and operating rules. For organization, it checks roles and responsibilities, cross-department coordination, activity evaluation, and efforts to build a supportive culture. For business impact, it asks whether goals are translated from company and department targets into initiatives, and whether teams prioritize them, agree with relevant departments, verify results, and share successful cases across the company.

The three infrastructure-level themes are data foundations, workforce development, and governance. The data foundation review covers functions for preparing necessary data as well as stable operation, scalability, security, and incident response. Workforce development covers talent profiles and development roadmaps, training content and methods, response to technology trends, and evaluation and incentives for skills acquired. Governance covers executive approval, documented policies and rules, and management arrangements for development and operations, security, privacy, and AI ethics. The assessment is meant to clarify gaps from the target state, not to produce a detailed evaluation report; taking too long risks the situation changing before results are ready.

On October 9, 2026, Keisuke Tanaka of Trust said companies planning an AI Center of Excellence (AI CoE) should start by defining their target state and assessing their current position, rather than first setting up the organization, staffing, or tools. The CoE’s three roles are to set direction, establish an execution environment, and spread results across the company.

The design process has three steps: define the desired state and goals, assess the current state, then choose measures and set priorities to close the gaps. The desired state has two parts: how the whole company wants to use AI, and what the AI CoE itself should achieve. The company-wide vision should align with its digital strategy and describe, in business terms, where AI and data can create value and strengthen competitiveness, including revenue, profit, customer experience, and business models.

AI CoE goals can include centrally managing priorities for company-wide AI initiatives, providing teams with the environment and procedures to start new initiatives, and enabling one department’s results to be reused by others. Because digital technology advances quickly and business goals can change, companies should set a “North Star” to show their direction; that guiding point may also shift as technology changes.

Tanaka recommended assessing the current state only after defining the target. Setting criteria first narrows the assessment to relevant areas and helps identify issues without spending time on an exhaustive survey. He pointed to the Financial Data Utilization Association (FDUA)’s Financial Data Utilization Organization Checklist. Although designed for finance, it includes perspectives shared across industries. It has two levels—use, which produces results, and infrastructure, which supports that work—and six themes in total.

The three use-level themes are data used, organization, and business impact. For data, the checklist examines the accumulation and quality management of internal and external data, along with purposes of use and operating rules. For organization, it checks roles and responsibilities, cross-department coordination, activity evaluation, and efforts to build a supportive culture. For business impact, it asks whether goals are translated from company and department targets into initiatives, and whether teams prioritize them, agree with relevant departments, verify results, and share successful cases across the company.

The three infrastructure-level themes are data foundations, workforce development, and governance. The data foundation review covers functions for preparing necessary data as well as stable operation, scalability, security, and incident response. Workforce development covers talent profiles and development roadmaps, training content and methods, response to technology trends, and evaluation and incentives for skills acquired. Governance covers executive approval, documented policies and rules, and management arrangements for development and operations, security, privacy, and AI ethics. The assessment is meant to clarify gaps from the target state, not to produce a detailed evaluation report; taking too long risks the situation changing before results are ready.