What is the “Next Wall” in AI Adoption? A Reality Check Reveals Management Blind Spots

What Lies Beyond the “Initial Success” of AI Adoption

Many companies are advancing the adoption of generative AI like ChatGPT and Claude, achieving a certain level of operational efficiency. However, as they move beyond the initial “trial” phase and transition into the “scaling phase”—where AI is permeated throughout the organization to generate sustainable value—new walls are emerging. Based on the latest reality surveys and our experience supporting over 38 clients, we explain the blind spots for management in this phase and concrete solutions.

Three Major Challenges in the “Scaling Phase” Revealed by Survey Data

A recent survey on corporate AI adoption (second half of 2024) revealed that among companies that have adopted AI in some form, only 18% consider themselves successful in full-scale, organization-wide utilization. The following three challenges emerged as the background to this.

1. Widening Skill Gaps and the “AI Literacy Polarization”
During initial adoption, employees with high IT literacy typically lead the charge. However, when trying to spread it across the entire organization, significant disparities arise among employees in understanding, from basic AI usage to prompt design and output validation. The survey showed that 67% of respondents reported a clear polarization between “employees who can use AI daily” and “those who can hardly use it,” revealing this as a real obstacle to organizational adoption.

2. Difficulty Standardizing Personalized “AI Workflows”
Because each employee uses AI in their own unique way, effective methods and best practices are not shared or standardized within the organization. Inefficiencies arise, such as one employee automating contract reviews with Claude while another performs the same task manually. 52% of surveyed companies cited “standardizing effective AI usage methods internally” as a challenge.

3. Opaque Cost Management and ROI Measurement
As AI usage becomes dispersed across individuals and departments, it becomes difficult to grasp the total monthly API usage fees and tool subscription costs. Furthermore, while individual users may feel time-saving benefits, only 31% of surveyed companies could quantitatively measure the return on investment (ROI) for the organization as a whole. Many remain stuck in a vague state of “it’s probably effective because we’re using it.”

The Need for a New Role: The “AI Adoption Manager”

To solve these challenges, I strongly recommend to client companies the clear establishment of a role tentatively called “AI Adoption Manager.” This role, distinct from the traditional IT department, fulfills the following four functions.

1. “Discovering, Standardizing, and Deploying” Internal Best Practices

Regularly collect effective AI use cases emerging from each department, and document and template them as internal standards. For example, reconfigure excellent prompts created in the sales department so they can be applied in the marketing department, promoting cross-functional knowledge flow.

Concretely, create an “AI Use Case Recipe Book” on an internal wiki like Notion or Confluence, accumulating the following items in a standard format:

• Business problem to solve
• Recommended AI tool (Claude / ChatGPT / Cursor, etc.)
• Specific prompt examples
• Expected time-saving effect
• Output verification points
• Relevant internal data/formats

2. Designing a Tiered Skill Development Program

Design a tiered program based on employee proficiency, not a one-size-fits-all AI training. The three-tier approach we practice is:

Level 1 (All Employees): Basic chat interface usage, handling confidential information, basic prompting (5W1H)
Level 2 (Department Leaders / Frequent Users): Advanced prompt design, tool integration (Zapier / Make.com), basic API utilization
Level 3 (Expert Users): Task automation using code-generation AI (Claude Code / Cursor), creating custom GPTs/Assistants

This categorization allows for effective allocation of limited training resources. Monthly costs, when using external instructors, can start from around $3 to $7 per employee per month for Level 1 training.

3. Cost Visualization and License Optimization

The AI Adoption Manager visualizes and optimizes company-wide AI-related costs. Specifically:

• Monitoring API usage per department
• Consolidating duplicate tool subscriptions (e.g., unifying separate ChatGPT Team plans across multiple departments)
• Proposing downgrades for high-cost licenses with low usage
• Exploring open-source/in-house development alternatives

One client company, by establishing this role, successfully reduced monthly AI-related costs by 23% while increasing the number of active users by 35%.

Redesigning Business Processes to be “AI-First”

In the scaling phase, it’s essential not just to “fit” AI into existing processes, but to redesign the business processes themselves with AI capabilities as a premise.

Practical Example: “AI-fying” Meetings

Traditional meeting process: Set agenda → Hold meeting → Create minutes → Share action items
After AI-First redesign: Pre-meeting AI analysis of agenda & background materials → Short dialogue session → AI auto-generates minutes & action items → Auto-distributes to stakeholders

With this redesign, one client reduced the average time required for a regular meeting from 120 minutes to 45 minutes. They also fully automated the man-hours for minute-taking (approx. 15 hours/month). Tools used: Claude (via API) for minute generation, integrated with Slack API and Google Docs API for distribution. Monthly cost is about $20 in API fees.

Building an Automated Report Generation Pipeline

For routine report creation tasks like sales reports, project updates, and accounting reports, building an AI-centric automation pipeline is highly effective.

Example pipeline we built and operate:

1. Data Collection: Automated data fetching from Google Sheets / Salesforce, etc. (automated via Make.com)
2. Analysis & Draft Creation: Input data and templates into Claude API for automated analysis and draft generation
3. Verification & Editing: Responsible person reviews the generated draft, makes edits if needed (edit history managed via Git)
4. Distribution & Sharing: Auto-convert final version to specified format (PDF / presentation) and auto-distribute via email/Slack to stakeholders

A company that introduced this pipeline reduced monthly routine reporting work from about 40 hours to about 5 hours. The initial build cost was approximately $3,300 (if outsourced), but the monthly time-saving effect translates to about $5,300/month, meaning the ROI can be recouped within a month.

Three Key KPIs to Prioritize in the Scaling Phase

In the AI adoption scaling phase, setting KPIs more directly linked to management, rather than just “usage time” or “user count,” is crucial.

1. Business Process Automation Rate

Measure the “percentage of business processes fully or partially automated by AI” per department. Set concrete, task-specific metrics, e.g., “invoice processing automation rate” for accounting, “quotation creation automation rate” for sales. Setting incremental goals, like improving by 5% each quarter, is effective.

2. AI-Dependent Task Continuity Score

Evaluate AI-dependent tasks reliant on specific employees on a three-tier scale: “Red” (high dependence), “Yellow” (medium), “Green” (low). The AI Adoption Manager works from “Red” tasks to standardize and document them, reducing key-person risk.

3. Departmental Visualization of Return on Investment (ROI)

Visualize AI-related costs (tool fees, training, development, etc.) and the reduced man-hours (time × hourly rate) they enable, per department. A simple formula we developed:

Departmental AI ROI = (Reduced Man-hours × Hourly Rate) / (AI-related Costs + Allocated Implementation/Training Costs)

Performing this calculation quarterly allows for reviewing usage methods or providing additional training to departments with low ROI.

Practical First Steps: Three Actions You Can Start This Week

Here are concrete actions executives and CTOs can start this week to move towards organization-wide AI scaling.

1. Conduct 30-Minute “AI Usage Reality” Interviews

Conduct short, roughly 30-minute interviews with leaders from each department. Questions:

• What AI tools are currently used for which tasks?
• What are the particularly effective use cases in your department? (Include specific prompt examples if possible)
• What is the biggest challenge in current usage?
• What tasks would you automate “if AI could do it”?

Based on these interviews, create an organization-wide AI adoption “map.” Whiteboard tools like Miro or Figma are recommended.

2. Launch a Monthly Cross-Departmental “AI Study Session”

Nominate 1-2 “AI Adoption Champions” from each department and start a monthly study session. First agenda:

• Share use cases per department (5 mins × number of departments)
• Identify common challenges
• Decide action items for next time (e.g., “Adapt sales department prompts for marketing”)

The facilitator of this session becomes the de facto first step towards an “AI Adoption Manager.”

3. Create an “AI-Related Expenditure” List for Cost Visualization

Collaborate with the accounting department to visualize all AI-related expenditures. Categories:

• Tool/Service subscription fees (ChatGPT, Claude, Midjourney, etc.)
• API usage fees
• Related books/training costs
• External development costs

Updating and sharing this list quarterly prevents unconscious cost creep and enables data-driven discussions on ROI.

Conclusion: Scaling Phase is a Shift from “Management” to “Ecosystem Building”

The AI adoption scaling phase requires building an “AI adoption ecosystem” within the organization, going beyond mere tool implementation management. This is a mechanism where excellent use cases are naturally discovered, shared, and evolved, with the key being a balance between centralized management and individual autonomy.

Through supporting over 38 clients, I’ve observed that companies successful in this phase position AI not just as a “convenient tool,” but as “infrastructure that enhances the organization’s very learning and adaptive capabilities.” Having moved beyond the initial “trial” phase, now is the time to squarely face the “next wall” in AI adoption to build sustainable competitive advantage.

The first step begins with objectively understanding what stage your company’s AI adoption is at and what challenges it faces. I hope the frameworks and practical actions introduced in this article serve as a helpful aid in that process.