Before Replacing Workers with AI

Hiring People vs Using AI: Who Owns the Work?

Most people see "hiring employees" versus "using AI" as merely a difference in cost and efficiency. But the reality has a deeper dimension — that of Ownership.

Dimension	Hiring Employees	Using AI (Subscription)
Knowledge and Processes	Can be documented and transferred; stays within the organization	Resides in the AI company's model — you don't own it
Price	Negotiable with employment contracts	Adjusted at the AI company's sole discretion
When Systems Go Down	Can continue working (albeit slower)	Work stops immediately until AI returns
Data in Use	Stays within the organization's systems	Passes through foreign AI company servers
Continuity	Resignation → hire replacement with knowledge transfer	Service shutdown → must start from zero

When hiring employees, you entrust your fate to another person — but at least that person is within your organization, under Thai labor law. Using AI means entrusting your fate to a foreign company that has absolutely no obligation to you.

- Organizational Risk Perspective

Risk 1: Service Outage — When AI Goes Down, Work Stops Immediately

On March 2, 2026, Claude AI went down globally. Tens of thousands of users opened their screens to find nothing but error messages. During those 20 minutes, software developers worldwide had to stop working. Organizations that had embedded Claude into every step of their workflow found themselves completely paralyzed.

The Claude incident is just the latest example — ChatGPT, Gemini, and Copilot have all experienced outages before. The difference is that if one of your employees falls ill, others can continue working. But if your sole AI provider goes down, all work stops.

Risk 2: Policies Change, Prices Rise, You Have No Bargaining Power

AI companies are private entities that answer to no one. They can:

• Raise subscription prices at any time (OpenAI has adjusted pricing multiple times in the past 2 years)
• Change pricing tiers — features you rely on may move to more expensive packages
• Discontinue certain services without much advance notice
• Change terms of use that affect how you use data

Risk 3: Data Fed to AI — Is It Still Yours?

When employees (or systems) feed data into AI every day, what happens to that data?

Data Type	Risk	Level
Customer data (names, contracts, behavior)	Could be used to train models / PDPA risk	Very High
Internal financial data (costs, margins)	Competitive data leaked to third parties	Very High
Strategies and business plans	Confidential organizational data flowing to foreign cloud	Very High
HR data (salaries, performance reviews)	PDPA violation, employee personal data	High
General documents, templates	Low risk, but caution still needed	Low

Most AI Terms of Service include language allowing them to use input data "to improve services" — meaning your customer data, business strategies, and internal information could be used to train models that your competitors also use.

Furthermore, Thailand's PDPA Act requires consent for transferring personal data abroad — feeding customer data into AI processed in the US could constitute unauthorized cross-border data transfer.

Risk 4: Vendor Lock-in — Switching Is Harder Than You Think

When all your workflows are designed for a specific AI provider, switching to another isn't just changing a subscription — it means:

• Re-train all prompts — each AI responds differently; prompts that work well with Claude may not work with ChatGPT
• Output quality changes — employees must readjust, and work quality becomes inconsistent during the transition
• Integrations break — if AI is connected to other systems (ERP, CRM, email), switching requires rebuilding all integrations
• Conversation history is lost — context accumulated in the old AI system cannot be transferred

Risk 5: Geopolitical Risk — Available Today, Potentially Banned Tomorrow

The events of February-March 2026 clearly demonstrated that Anthropic was banned from US government use by the Trump administration over AI policy disputes — all within a matter of weeks.

For Thai organizations, geopolitical risks take multiple forms:

Scenario	Impact on Thai Organizations	Likelihood
AI company hit by US sanctions	Service suspended in the region	Low but not zero
AI company goes bankrupt or gets acquired	Data and pricing policies change immediately	Moderate
Thailand enacts new AI legislation	Some AI services may not be compliant	Moderate to High
US-China tech war escalates	Some AI services may be restricted	Moderate

AWS, Azure, Google Cloud — the infrastructure behind every AI provider falls under US jurisdiction. Any conflict between the US and other nations can impact your access to services at any time.

Framework: What AI Can Do vs What AI Should Not Fully Replace

Asking "Can AI replace people?" is the wrong question. The right question is: "Should AI assist with this task, or fully replace humans?"

✅ AI Can Help (Should Not Fully Replace)	⛔ AI Should Not Fully Replace Humans
Document summarization, translation, drafting	Tasks requiring legal accountability (signing contracts, budget approval)
Data analysis, creating report templates	Tasks handling confidential and customer data
Answering FAQs, tier 1 customer support	Mission-critical tasks with no fallback
Creating marketing content, social media drafts	Strategic decision-making and judgment calls
coding assistance, QA testing support	Tasks requiring accountability to stakeholders

Simple rule: If the task cannot stop or cannot tolerate errors — never let AI handle it alone. AI should be an "assistant" that helps people work faster, not a "substitute" where humans disappear from the process.

Recommendations for Executives: Use AI Wisely

The point is not "to use AI or not" — AI is a highly effective tool and should be utilized fully. The point is that the way you use it must not cause your organization to lose control.

The Solution: AI Assists People + ERP as the Core

Organizations that use AI sustainably and safely don't "replace people with AI" — they adopt a different model:

• ERP system serves as the core that stores your organization's data, processes, and knowledge — data stays in your hands, not on foreign servers.
• AI serves as an assistant that helps employees work 3-5 times faster, not as a replacement.
• Data stays in ERP even if AI goes down or changes policies — the business can continue moving forward.

Having an ERP system with its own database means that no matter which AI provider goes down, changes policies, or raises prices — all your data and processes remain within your organization, never held hostage by foreign technology companies.

Conclusion: AI Is a Tool, Not an Employee

In an era where AI improves daily, executives' desire to reduce labor costs is understandable. But the hidden risks behind AI go deeper than just price and performance:

• When AI goes down — work stops immediately
• When policies change — you have no bargaining power
• When data is fed to AI — you're not sure where it ends up
• When vendor lock-in is complete — switching costs are extremely high
• When geopolitics shift — services can disappear overnight

The question executives should ask themselves is not "Can AI replace people?" but rather: "If AI stops working tomorrow, can our organization continue moving forward?"