We shape our tools and thereafter our tools shape us. – Marshall McLuhan

The spreadsheet is a tool, and it is also a world view — reality by the numbers. If the perceptions of those who play a large part in shaping our world are shaped by spreadsheets, it is important that all of us understand what this tool can and cannot do. – Steven Levy

Picture the office environment of 1980. Telephones, stacks of paper, the occasional calculator, and reams of data stored in impenetrable filing cabinets dominate it.

The logic of this office environment was essentially unchanged for the better part of a century. Calculators and typewriters transitioned from mechanical to electronic, but their basic functionality remained the same. Neither transition altered the daily functions of most businesses or their employees.

We are still struggling with this transition today. The typical office of 2025 still operates in a paper metaphor, and most operate according to industrial scheduling.

While communication and sharing have changed considerably, it has done nothing more than speed up the speed at which we receive, store, and process data. The world has shifted slightly, but we are still in the midst of an unfinished evolution.

For most people, this transitory period started around 1980. Computers existed at a distance in the early 80s for most companies. Perhaps you sent out data requests to one of the big computing firms, but the idea of doing this yourself at speed was something completely alien to your thought process or workflow.

In the consumer space, some people started playing with personal computers, like the Apple ][, or one of the Radio Shack computers. People viewed these as toys.

However, one of those computers, the Apple ][, had a new application on it. This application allowed you to do advanced calculations on grids of numbers. It grew out of a Harvard business school practice of creating these grids manually and the insight of Dan Bricklin and Bob Frankston that this was an ideal task for personalized computing.

Today, we take spreadsheets for granted, but in 1980 they were an alien and revolutionary idea. Business was operating at full speed just to keep up with the day-to-day requirements of running itself.

Technology limited their ability to predict the future. Paper is an artifact of the past. Once you put it on paper, it has already happened.

If you can’t predict the future, you can’t plan. You can’t lay in for unexpected crises and emergencies. You are hostage to the whims of fate.

Most importantly, you couldn’t model alternative futures and build scenarios to prepare your company for the likelihood that the unexpected will happen. This all changed with VisiCalc and its instantaneous calculation of rows of figures.

This idea was so powerful and revolutionary that people started buying Apple ][‘s with their own money and bringing them into the office. Businesses that looked to the future installed these strange devices and explored ways to use them more efficiently.

AI’s integration into business routines mirrored VisiCalc’s. Within weeks of ChatGPT’s release in late 2022, people were already using it to increase their personal efficiency.

They did this with little understanding of its strengths and limitations. This led to some hilarious missteps and professional embarrassment when people assumed the computing infallibility they expected from existing systems only to discover ChatGPT didn’t work like that.

For two years, considerable debate surrounded AI’s influence on our lives, both personal and professional. I am still awaiting the VisiCalc moment when our ability to perceive the world shifts.

I have written many times before about our struggles to put tasks in front of tools. Generative AI is no exception. It’s the latest and greatest thing. Everyone wants to use it. No one knows what to do with it.

VisiCalc reshaped the paradigm of the possible. It changed our sense of time in very subtle but profound ways. Suddenly, an application linked the past, present, and future.

Most applications of AI we see today are merely substitutions for existing tasks. All too often, it doesn’t do these tasks well. Early computers seemed obtuse and without a clear, practical function until spreadsheets and word processors came around.

I still use AI in “legacy” ways. For instance, this blog will get run through ChatGPT, which will offer suggestions for structure and content. I used to have human partners read through drafts, but they are usually much too busy to read my drafts.

This is not transformative. I am substituting ChatGPT for my busy friends. This has changed the speed and efficiency of my writing, but it has not significantly altered my workflow (yet).

Everyone seems to think that AI has changed everything, but no one seems to be able to identify what has changed. It probably took several years in the early 80s for the impact of spreadsheets to sink in. We had to play with them for a while to figure out how they changed our thinking.

I heard a speaker refer recently to AI as either a “toy, tool, or teammate.“ I don’t like the implication of separation between the three. We should play with our tools and our teammates in order to create innovative ideas. Therefore, everything should be a toy.

VisiCalc let us play with numbers in ways that were impossible before. AI lets us play with ideas in ways that were impossible before.

The dominant Generative AI tools today don’t make that very obvious. They reflect existing metaphors. It’s not obvious to most people how they synthesize information and so they look like more sophisticated versions of legacy information retrieval systems, like Google search.

We are still operating like we’ve been operating for the last quarter century. Sure, ChatGPT and other large language models give us more useful results than a Google search has done, but that’s not because they represent a new modality for thinking.

VisiCalc changed how we think. It grew out of the computational capabilities provided by access to PC computing power.

AI is not VisiCalc in this metaphor. AI is the PC. Generative Large Language Models are engines. We are still building cars around them to attach them to transformative action in our business and personal lives, much like VisiCalc transformed the utility of the Personal Computer.

Scott McCloud refers to these kinds of transformative changes as “durable mutations.” He describes durable mutations as lasting shifts in how we think and create, not just passing trends. Spreadsheets did that for the PC. We are still looking for AI’s durable mutations of how we work and live.

The thing that makes Generative AI unique is its ability to synthesize information at scale. This enables play on a vastly larger scale than the labor-intensive tools of the last 40 years have let us do.

The revolution that VisiCalc ushered in was systems that automated calculation. AI automates creativity.

Durable mutation will come from our willingness to play with boundaries in what we think is possible in the ideas space, just as VisiCalc encouraged us to play with boundaries in the business space. AI allows for unstructured combinations of ideas. That is where its power lies.

Instead of employing technical people to create macros for data, we can skip that step and let AI augment our vision in ways that show us the possibilities rather than constraints. We must allow ourselves to play in order to complete our unfinished digital evolution.