Summary of Key Points
Artificial Intelligence (AI) is significantly reducing the cost for companies (customers) to develop their own software, making it more attractive to develop in-house rather than purchase externally. This poses a far more lethal threat to software companies than competition from other firms. Once customers start developing software in-house, these companies no longer even have the opportunity to bid on projects. However, software companies are not without options; they can find their own irreplaceability in the AI era by accumulating industry-leading experience, establishing high-quality delivery systems, and becoming providers of essential AI infrastructure.
1. AI Has Changed the Cost-Benefit Calculations for Customers: What Was Unattractive Before Is Now Appealing
Why did customers prefer to purchase software externally in the past? It wasn’t because they didn’t want to develop it themselves; rather, the cost of maintaining a team that could create only 90% of the desired functionality was too high—hiring engineers and providing long-term support was much more expensive than buying ready-made solutions. But with AI:
- AI-assisted coding tools (such as Copilot) have doubled the speed at which engineers can write code, reducing the need for manpower;
- Large models can handle many routine tasks (like data organization and basic feature development), eliminating the need for a large number of senior engineers;
- Now, customers can create systems that meet their needs for half or even less of the previous cost. As a result, they are re-evaluating every external software solution: “Can we develop this ourselves?”
2. Real-World Examples of Customers Cutting into Software Companies’ Business
The article provides several striking examples:
1. An AI company lost a major client: An AI firm with annual revenues in the tens of millions had a significant customer that paid millions in subscription fees each year. The company assumed the client wouldn’t develop the software in-house due to the complexity, but as AI capabilities improved, the client required less external support and decided to develop it on its own, resulting in a loss of several million in annual revenue.
2. Starbucks Replacing External Systems: Starbucks spent $400 million annually on software (such as inventory tracking from Microsoft and equipment maintenance from IBM). Now, using AI-assisted development, it is re-evaluating all its contracts and no longer purchasing external solutions for tasks it can handle internally.
3. An Insurance Company Developed Its Own CRM in Two Months: The American company Curative used AI tools to create its own customer relationship management (CRM) system in just two months, saving $600,000 in annual SaaS subscriptions.
These examples show that it’s not necessarily competition from other companies; rather, the entire software industry is seeing customers taking a direct cut into the market share.
3. There Is Still a “Window of Opportunity,” but Don’t Mistake It for an Immunity Period
Software companies have not yet felt the fatal threat because customer IT departments lack experience with AI products (such as how to design AI agents or address issues related to large models). Therefore, they continue to purchase software from vendors. However, this window of opportunity will not last forever:
- Customers will gradually gain more expertise (by using vendor products);
- They may also poach talent from software companies (especially engineers with knowledge of AI technologies);
- Once customers overcome these shortcomings, there will be no reason for them to rely on software vendors anymore. Software companies must act quickly and plan their strategies during this period.
4. Three Paths for Software Companies to Survive: Focusing on What Customers Cannot Do Themselves
To avoid being abandoned by customers, software companies should focus on areas where they have a competitive advantage:
1. Accumulate Industry-Leading Experience: While customers understand the specifics of their own business, they often lack knowledge of best practices in the industry (e.g., how other companies manage conflicts between sales and production departments). Software companies can extract these best practices and turn them into products or services that customers cannot develop on their own.
2. Establish High-Quality Delivery Systems: Just as in the ERP era, customers lacked implementation experience and had to pay for vendor assistance; in the AI era, they will need professional delivery capabilities for self-developed AI solutions. Software companies can become providers of these essential services.
3. Become Providers of Essential AI Infrastructure: Behind AI agents are many complex technologies (such as intent recognition, knowledge base management, and permission checks) that are difficult for customers to implement from scratch. Software companies can create reusable tools that make it easy for customers to integrate AI into their systems, making them dependent on these services.
In Conclusion
AI will not eliminate the demand for software; instead, it will create new opportunities. However, software companies must shift from selling mere systems to providing expertise, services, and essential infrastructure to maintain their position in the AI era. The real threat lies in the possibility that customers no longer need them at all when they develop software in-house.