Expert: AI Forcing 'Build vs. Buy' Rethink
The rapid evolution of AI is changing the economics of software development, forcing companies to re-evaluate when to build proprietary technology versus buying vendor solutions. In a recent podcast, an expert advised buying for non-differentiating features and building only for unique competitive advantages. He warned that while AI lowers development costs, it can increase operational costs and technical debt.
- The total cost of ownership for a custom-built AI solution over five years can be between $2 million and $3 million, factoring in salaries for AI/ML engineers which range from $200,000 to $300,000 per person. In contrast, buying a commercial solution can cost between $200,000 and $700,000 over the same period. - Building a custom AI project can take between six to eleven months, from nailing down requirements to deployment and stabilization. Purchasing and implementing a pre-built solution can be significantly faster, with a timeline of four to eight weeks. - According to research from Keystone.AI and Harvard Business School, the productivity increase from generative AI developer tools could add over $1.5 trillion to the global GDP by 2030. - Gartner reports that 85% of AI projects fail to deliver on their promised results, often due to the wrong "build vs. buy" decision. However, successful AI implementation can lead to a 20-25% increase in EBITDA, according to McKinsey. - A hybrid approach is becoming increasingly common, where companies buy vendor solutions for standard functions and build custom AI for unique, differentiating capabilities. - The "buy" decision is not without risks, including vendor lock-in, the possibility of a vendor going out of business, or their product roadmap not aligning with future needs. - While building an AI solution offers maximum customization, it also comes with hidden costs such as project management overhead, team training, and significant technical debt if not executed properly. - The decision to build is often justified only for the 5-10% of companies whose competitive advantage relies on a proprietary AI approach that no commercial solution can offer.
