In 2026, artificial intelligence permeates nearly every facet of modern life, from generating creative content to automating complex business processes. Tech companies are locked in a race to build ever-larger and more capable AI models, even as concerns grow about the technology’s potential impact on jobs and society. Yet, even as the current AI boom feels revolutionary, its roots—and its potential pitfalls—were foreseen decades ago. As early as 1985, industry observers were cautioning against the hype surrounding artificial intelligence, advocating instead for a more pragmatic approach focused on software that adapted to the needs of its users. This early debate, between the promise of “thinking machines” and the practicality of “softer software,” continues to resonate today.
The seeds of this discussion were sown at the dawn of the personal computer era. Mitch Kapor, then chairman of Lotus Development, warned at the January 1985 Personal Computer Forum that artificial intelligence was poised to become “the most despised and abused [software concept] of the next year.” He characterized the impending rush toward AI as “lemming-like,” but also saw an opportunity for those who understood its true potential to deliver value to customers. This sentiment was echoed in a February 25, 1985, editorial in InfoWorld, titled “Awaiting AI Hype, Promise,” which questioned whether 1985 would finally be the year AI moved beyond academic circles to become a truly useful tool.
The InfoWorld editorial, penned by Editorial Director & Associate Publisher James E. Fawcette, captured a core anxiety that feels strikingly familiar in the age of generative AI: “What can AI really do?” Fawcette argued that the very term “artificial intelligence” was a barrier to practical application, laden with expectations of “thinking machines” and dystopian visions of computers controlling human lives. He identified two primary forms of AI misuse: “AI-hype”—programs making grandiose promises they couldn’t deliver—and “Rube Goldberg overdesign syndrome”—overly complex systems attempting to solve problems that didn’t require such elaborate solutions. Scrolling through LinkedIn today, one can easily find examples of both.
Bill Gates’ Vision of “Softer Software”
Fawcette noted that some industry leaders were already steering clear of the AI label. “Microsoft’s Bill Gates has coined the term ‘softer software’ to describe his vision of programs that will learn the user’s work patterns and help execute them,” he wrote. This concept, first articulated two years earlier in an August 29, 1983, InfoWorld interview, represented a deliberate shift in focus. Gates and Charles Simonyi, then at Microsoft, acknowledged the complexity of achieving true artificial intelligence, stating, “AI is a very complex goal… You need a philosopher to determine what AI is.”
Instead of pursuing the elusive goal of artificial intelligence, they proposed “softer software” as a more attainable and practical objective. Simonyi explained that this approach was empirical, meaning it would “modify its behavior over time, based on its experience with the user,” ultimately aiming to simplify tasks in the “real world.” Simonyi, a pivotal figure in software history, led the teams that created Microsoft Word and Excel, and is credited with popularizing features like pull-down menus, icons, and WYSIWYG editors—elements now ubiquitous in modern computing. His vision of “softer software” foreshadows today’s personalization engines and adaptive AI assistants.
As InfoWorld described it in 1983, “softer software” would “remember” a user’s preferences—such as double-spacing and right-margin justification—and automatically apply them, learning from observed habits. Simonyi predicted that computers would evolve into “working partners,” anticipating user needs and proactively offering suggestions, “molding itself based on events that have taken place over a period of time.” This vision is remarkably prescient, mirroring the functionality of contemporary software and AI assistants.
From Expert Systems to Excel Macros
Microsoft took its first steps toward realizing this “softer software” dream with “expert systems” designed to enhance its spreadsheet program, Multiplan. These early systems aimed to automate tasks like building formulas and analyzing data, making spreadsheets more accessible and powerful. Multiplan was eventually superseded by Excel, Microsoft’s next-generation spreadsheet software.
A 1985 InfoWorld review of the first Macintosh version of Excel (Windows 1.0 followed later that year) highlighted the “learn-by-example macro feature” as a key step toward fulfilling Gates’ vision. Reviewer Amanda Hixson noted that Excel macros allowed users to automate tasks without needing to understand programming, simply by recording their actions and replaying them as needed. This user-trained automation was a significant advancement, empowering users to customize their software experience. At the time, Lotus 1-2-3 was the dominant spreadsheet program, but Gates criticized Lotus Jazz, the company’s attempt at an all-in-one software suite, arguing it sacrificed integration for breadth. Microsoft’s focus on focused integration ultimately proved more successful.
Excel’s success wasn’t built on promises of magic, but on delivering “consistency, power, lots of features, and macros,” as Hixson wrote. The program’s ability to learn from user behavior and automate repetitive tasks resonated with a market tired of overly complex and underperforming software.
A Copilot, Not an Oracle
The 1985 InfoWorld editorial envisioned software capable of not only generating sales reports but also automatically creating accompanying charts. This once-futuristic scenario is now commonplace, with AI systems routinely drafting reports, summarizing meetings, and generating visualizations from simple prompts. These systems function as “agents,” executing multi-step tasks across various applications. The editorial concluded with a prescient observation: “We’ll secure our first taste [of AI] this year. Let’s hope some applications are as intelligent as the software algorithms used to implement them.”
Bill Gates, in 1985, wasn’t dismissing intelligence in software, but rather rejecting the inflated expectations surrounding it. By championing “softer software,” he envisioned systems that learned from users, adapted to context, and acted as partners—a copilot rather than an all-knowing oracle. This distinction remains crucial today, as developers grapple with the ethical and practical challenges of building AI systems that are both powerful and responsible. The ongoing debate over AI regulation, including the recent Executive Order from President Trump and the evolving EU AI Act, underscores the need for a thoughtful and pragmatic approach to this transformative technology.
As AI continues to evolve, the lessons from the early days of personal computing remain relevant. The focus should be on building tools that augment human capabilities, rather than attempting to replicate human intelligence. The future of AI likely lies not in creating artificial minds, but in crafting intelligent systems that seamlessly integrate into our lives, making us more productive, creative, and informed.
What are your thoughts on the evolution of AI? Share your comments below.
