![The AI Threat Only Visible from Inside the IT Industry -- Can the Body-Shopping Business Survive? [Technovate Thinking #10]](https://megumirai.com/wp-content/uploads/2026/04/tct_eyecatch_10.jpg)
This is part 10 of a 12-part series documenting what I learned from the “Technovate Thinking” course at business school.
Last time, I wrote about building an app using no-code tools and AI APIs. This time, I’m tackling the assignment from the second half of Session 5: “How is AI being used in your business domain, what’s expected going forward, and what threats does it pose?” Having just built an AI app with my own hands, I felt grounded enough to produce an analysis rooted in reality, not hype.
- 1 The Current State: Generative AI Has Stalled at “Getting Started”
- 2 Four Areas of Expectation
- 3 Threats: Three Issues That Could Shake the IT Industry’s Foundations
- 4 A Leader’s Job Is to “Articulate the Threats”
- 5 The View You Only Get by Getting Your Hands Dirty
- 6 Next Up: The Final Assignment — Solving a Real Business Problem with Tools
- 7 Books Referenced in This Article
The Current State: Generative AI Has Stalled at “Getting Started”
As of 2025, generative AI is spreading rapidly across the IT industry. Microsoft Copilot, ChatGPT, various OpenAI-based tools — many people are already using them for intellectual labor support like drafting documents, replying to emails, and summarizing meeting notes.
But there’s a massive gap between “getting started” and “truly leveraging.”
Security restrictions on feeding business data into AI. Copy-paste constraints between internal systems and external tools. Usage limitations driven by personal data protection. “Want to use it but can’t” and “can use it but with restrictions” — that’s the reality for most enterprises.
There are cases where AI is partially deployed — automated FAQ responses, sales collateral draft generation, and so on. But many remain stuck at the PoC (proof of concept) stage and never make it to production. Companies issue press releases saying “We’ve adopted AI,” but whether the front lines are truly using it effectively? That’s a different story entirely.
Four Areas of Expectation
I identified four domains where AI will become deeply embedded in business operations going forward. These predictions are based on years of firsthand experience in the IT infrastructure background.
① AIOps — Automating and Elevating IT Operations
Log monitoring, anomaly prediction, automated recovery — the domain known as AIOps (AI for IT Operations). The idea is to learn patterns from massive log data and detect anomalies at speeds beyond human capability.
Consider a concrete scenario: at 3 AM, a server’s response time degrades to 1.5x normal. A human operator would take several minutes just to notice the alert and begin investigating the cause. But AIOps could detect the subtle upward trend in response times before the outage occurs, infer the cause from similar historical patterns, and automatically execute scale-out or failover — all potentially within seconds.
Since this domain directly relates to IT operations, it’s the one I can imagine most vividly. I’m convinced that AI will qualitatively transform 24/7 monitoring as we know it.
② AI Integration into the Development Process
Code generation, bug fixing, test automation. AI coding assistants like GitHub Copilot are already part of many developers’ daily workflow. But going forward, AI will push further upstream — into automated design document generation, test coverage completeness checks, and automated security vulnerability detection.
What’s especially worth watching is the expansion of the API economy. As I experienced with the “My AI Coach” app last time, the style of combining existing services — AppSheet, Zapier, OpenAI API — to create new applications will become increasingly mainstream even in professional development. We’re shifting from an era of writing code from scratch to one of assembling APIs to generate value.
This fundamentally changes the developer’s role. Simply “being able to code” won’t be enough. The question becomes, “Which APIs should I combine, and how, to generate value most efficiently?” — in other words, architectural design capability becomes the differentiator.
③ AI Agents for Administrative Work
Schedule coordination, contract review, expense processing — routine decision-making tasks will increasingly be handled by AI agents.
Take expense reimbursement, for example. Today it’s a human relay: “photograph the receipt, manually enter the line items, manager approves, accounting verifies.” But an AI agent could: automatically read the receipt via OCR, match it against past reimbursement patterns to auto-classify the account category, check for policy violations using rule-based logic, and auto-approve if everything checks out — handling the entire workflow end-to-end.
Contract review is similar. From a contract spanning dozens of pages, AI can instantly extract and highlight clauses unfavorable to your company, changes from previous agreements, and legally risky language. The lawyer’s manual review becomes just a final confirmation step.
A future where most “routine judgment tasks” shift from humans to machines isn’t far off. In fact, the awareness that not replacing them is the real inefficiency is just a matter of time.
④ The Generalization of Recommendations and Personalization
The recommendation concepts I learned in Sessions 3 and 4 aren’t limited to e-commerce or music streaming. They’ll become a general mechanism for “delivering the optimal proposal to each user” across virtually every business.
Think about internal talent allocation, for example. AI could match a project’s required skill set against employees’ past performance, evaluation data, and personal preferences, then present the optimal staffing candidates. Or personalized training programs: even people at the same level have different skill gaps. AI could analyze each individual’s weaknesses and recommend the learning content that’s most relevant to them.
The “My AI Coach” I built last time was precisely an example of personalization — not a generic fitness app, but one that delivered advice optimized for my body composition data, my goals, and my dietary preferences. This concept of “individual optimization” will spread across every area of business.
Threats: Three Issues That Could Shake the IT Industry’s Foundations
Everything above was about “expectations.” But honestly, the part of this assignment where I spent the most time was the “threats.” Talking about expectations is easy — anyone can do it. But with an IT infrastructure background, there are threats you can’t help but see. I didn’t want to look away.
Threat ①: The Collapse of the “Body-Shopping” Business Model
The IT industry has a deeply entrenched business model often called “body-shopping” or “staff augmentation.” You sell engineers’ and operators’ billable hours — monthly utilization times rate equals revenue. It’s simple, transparent, and has supported the IT industry’s revenue structure for decades.
This model only works because “there are tasks that only humans can do.” But what happens when AI agents reach a practical level of capability?
Document creation — AI drafts it, humans just do the final check. Summarization — AI automatically summarizes meeting recordings. Translation — multilingual support is AI’s sweet spot. Test execution — AI handles everything from test case generation to execution to result reporting. When work that used to require three person-months can be done with a monthly AI subscription costing a few hundred dollars, the “sell by the person-month” business gets shaken to its core.
So what replaces it as a revenue source? I believe it’s “systems design capability” and “depth of business understanding.” Even when AI can handle the execution, the people who can design “what to have AI do” and “what system structure to use for running the business” will only become more valuable. What I realized building the AI Coach app was that using the tools themselves is easy — designing how to connect which tools to solve your specific problem is far harder.
Threat ②: AI Standardization Erases Differentiation
AI tools are available to everyone. Copilot, ChatGPT — subscribe and you can start tomorrow. This means that using AI for operational efficiency is no longer a competitive advantage in itself.
Suppose a company uses AI to improve its system incident response speed by 50%. That’s an impressive result. But if competitors adopt the same AI tools, they achieve the same improvement. AI creates differentiation not through “being able to use it” but through “what you use it for” — yet many companies still mistake the former for a competitive edge.
True differentiation lies in the ability to envision and design solutions grounded in deep understanding of the customer’s business. “Where are the bottlenecks in this customer’s workflow?” “How much business impact would AI-driven resolution of those bottlenecks deliver?” — only someone who understands both the technology and the business can answer these questions.
This is precisely why “Technovate talent” — people who understand both technology and business — are in such demand.
Threat ③: Organizations That Can’t Thoughtfully Collect Data Will Be Left Behind
As I learned in Sessions 3 and 4, recommendations and AI alike can’t do anything without data. But simply collecting data isn’t enough.
What I felt acutely while building the AI Coach app was that designing what data to collect, in what format, and at what timing is what decisively determines app quality. Because I structured body composition data into separate columns for “weight,” “body fat percentage,” and “muscle mass,” the AI was able to return advice citing specific numbers. If I’d recorded it qualitatively as “body composition: good,” the AI would have had nothing to work with.
The same thing happens in business settings. If customer interaction records are just free-text notes without structured data on response time, issue category, or customer satisfaction, no amount of AI adoption will enable meaningful analysis. Organizations that can’t design “systems for collecting data” will be left behind in the AI era.
Let me be specific. Instead of sales reports that say “Worked hard again today,” design a system that records “Visits: 3, Proposals: 1, Close probability: B” as structured data. Instead of filing customer support inquiries under “Other,” design classification rules and record them properly. This kind of unglamorous but essential data design should come before AI adoption, not after.
Risks You Can’t Ignore: Ethics, Hallucinations, and Accountability
Beyond the three threats, AI adoption carries unavoidable risks.
Privacy and ethics. When feeding business data into AI, will customers’ personal information be used in model training? Is there a risk of confidential partner data leaking? With legal frameworks still catching up, companies must design their own guardrails.
The risk of hallucinations. During the AI Coach development, the AI confidently suggested menu items that didn’t exist. If something like that were used for business decision-making, it could lead to irreversible mistakes. A system that doesn’t blindly trust AI output — and people who can verify it — are indispensable.
And accountability. If an AI-generated document contains errors, who’s responsible? If an AI-recommended initiative fails, who answers for it? “The AI did it” is not an excuse. The principle that humans bear final judgment and responsibility needs to be clearly established within organizations.
A Leader’s Job Is to “Articulate the Threats”
The strongest takeaway from this assignment was that talking about expectations is easy, but articulating specific threats is far more valuable.
“AI will improve operational efficiency” — anyone can say that. Lining up AI’s shiny possibilities in a presentation deck can be done in five minutes with ChatGPT.
But “the body-shopping business model won’t survive,” “AI standardization will neutralize existing differentiators,” “organizations that can’t design data collection will be dead in three years” — only someone on the ground in the industry can articulate threats like these as personal truths.
Moreover, articulating threats takes a certain kind of courage. Admitting your own business model is vulnerable is uncomfortable. But looking away only makes it too late. Confronting inconvenient truths, communicating them to the organization, and devising countermeasures — that, I believe, is the real job of a leader.
The View You Only Get by Getting Your Hands Dirty
Building the AI Coach app last time and analyzing AI threats this time were two sides of the same learning coin.
Because I built an AI app with my own hands, I gained a visceral sense of the boundary between what AI “can do” and “can’t do.” Because I wrestled with the constraints of free tools, I’m not seduced by the fantasy that “AI can do everything.” Because I struggled with prompt design, I can see the danger in the naive optimism of “just deploy AI and everything will work out.”
The essence of Technovate Thinking is understanding the potential of technology while coolly analyzing what structural changes it will bring to your business and formulating a response. Not just expectations but threats. Not just possibilities but constraints. Only by examining both sides can you build a strategy with your feet on the ground.
A business leader’s job isn’t just to talk about what technology can do. It’s to confront the structural changes technology brings and figure out how the organization should prepare. As I approach the final stretch of Technovate Thinking, I can feel this perspective taking root within me.
Next Up: The Final Assignment — Solving a Real Business Problem with Tools
Session 6 is the final session. The capstone assignment: “Following the Technovate Thinking problem-solving process, use free tools to design and implement a solution for a real problem in your own business.” Having analyzed both expectations and threats, what do I actually do about it? That’s what I’ll write about next time.
Books Referenced in This Article
