Benedict Evans: The notion that AI can make everyone a tool builder is a misconception

Benedict Evans criticized the optimism that AI will replace existing software and enable all users to build their own tools, arguing that it misunderstands the actual structure of enterprise operations. Modern large enterprises possess complex software ecosystems ranging from massive systems like SAP and Workday to hundreds of SaaS products, scripts, and spreadsheets, and the expectation that AI will instantly replace all of this is disconnected from reality.

The core issue is not the ease of writing code, but the difficulty of problem identification and solution definition. Most employees are not tool builders capable of designing tools to automate their work, and the workflows targeted for automation span multiple departments and systems, making them impossible for individuals to change arbitrarily. Institutionalized software is essential for auditing and security, and when work in free-form spaces (such as Excel) becomes repetitive, a 'Desire path' process occurs where it eventually transitions into new SaaS products.

According to experiences with enterprise AI deployment, even when all employees are provided with Copilot or ChatGPT, many fail to find ways to utilize them. This is a change management problem similar to the periods of Lotus 1-2-3 or web browser deployment. AI does not change the questions themselves, but it creates new options and shifts thresholds. When adopting AI, enterprises must make strategic decisions across three dimensions: build/buy approaches, operational impact scope, and business model changes, requiring organizational transformation beyond simple tool deployment.