The Evolution of AI over 80 Years

AI Advisory

Most people think of ChatGPT, Claude, and generative AI tools as "AI". There is much, much more to artificial intelligence than what most people are familiar with today.

This one maps technology classes instead: 28 of them, from 1943 to 2026. Each lane shows when the class was first formulated in research, when it became deployable in commercial production, and whether it is still doing work today.

Card titled The Evolution of AI over 80 Years, with two-tone bars where the pale segment is the research era and the solid segment is the commercially deployable era, and a median gap of 22 years.

On this page

Free Revenue Lifecycle Assessment

Connect with Marissa Wright to receive a free Revenue Lifecycle Assessment Report on your own business.

Book a consult →

Most people think of AI as a chat tool, or more recently, AI agents and coding tools. But ChatGPT arrived in 2022 and we are only just past one year for commercial AI agents and coding tools like Claude CoWork and Claude Code that can drive real business value.

To give you a better understanding of AI's history and the technologies involved, I lay out a timeline of various AI technology classes: 28 of them, from 1943 to 2026.

Each lane shows when the class was first formulated in research, when it became deployable in commercial production, and whether it is still doing work today.

Across the 27 classes here that reached commercial deployment, the median gap is 22 years. The range is wide: the transformer architecture reached production use roughly a year after publication, while reinforcement learning took 65 years.

The second finding is overlap. Older AI technologies did not stop when newer ones arrived. Constraint programming, rule-based systems and classical planning are all still in production, and several remain the better choice where a decision has to be accurate rather than close.

This interactive diagram maps each AI class and includes a plain-language explanation alongside the technical summary. I hope you find this as interesting as I do!

Other Posts

Post FAQ

What AI technologies can be used to drive business value?

One answer is all of them: some like speech recognition and natural language programming have been absorbed into other technologies. None of them are fully dead. However, tools like Claude, Google Gemini, and ChatGPT use about 18 of these 27 classes today.

Business value is gained from (1) creating new capabilities in your business, (2) improving your process scalability, and (3) affordability. We now have affordability. However, deploying licenses for a chat tool to your workforce will not drive significant value; integrating assistive and agentic AI tools into your business processes will.

Should I generally provide AI licenses for ChatGPT and Claude to all of my employees?

The problem with deploying licenses for these tools without integrating them into your business processes is that they generally:

  • Lack the context of your business processes
  • Lack effective cybersecurity controls
  • Are dependent on the user’s input for the quality of output
  • Can be a significant distraction to your employees
  • Can be misused and enable false information to be presented to your customers
  • Create significant costs without measurable business value

AI enablement for business needs to include controls for all of these factors.

How mature are autonomous AI agents in 2026?

Pilots are common and sustained production use is not. A substantial minority of large enterprises report at least one agent running in production, concentrated in banking and insurance, and far fewer report scaling agents across multiple departments. The binding constraint is generally safe and reliable access to live systems rather than model capability. Adoption figures for this class come from vendor and analyst surveys that define agent and production inconsistently, so the direction is more reliable than any specific percentage.