1985: when AI learned to think like an expert
Written by Eve Oliver, Marketing Manager at CSG Computer Services Ltd and a CyberFirst ambassador as part of the National Cyber Security Centre [NCSC].
As CSG celebrates 40 years in business throughout 2026, we’re taking a step back in time to explore what life was like for AI and technology the world the year we were founded. Today, we’re looking at a defining artificial intelligence theme from 1985: the rise of expert systems.

What were expert systems?
In the mid-1980s, artificial intelligence wasn’t about chatbots, image generation, or large language models. Instead, AI researchers were focused on a different challenge:
Could a computer replicate the decision-making ability of a human expert?
The result was the development of expert systems, AI programs designed to solve complex problems by applying rules and knowledge from specialists in specific fields.
These systems worked using structured “if–then” logic. For example:
- If a patient has symptoms X and Y → suggest diagnosis Z
- If a system configuration includes component A → require component B
This approach allowed computers to mimic reasoning in a way that felt ground-breaking at the time!
By 1985, expert systems were already being used commercially. One of the most famous examples was XCON, developed for Digital Equipment Corporation. This AI system helped configure complex computer orders with high accuracy, replacing a manual process that was prone to errors.
At its peak, systems like XCON demonstrated that AI could deliver measurable business value by:
- Reducing errors
- Automating complex decisions
- Improving operational efficiency
For many organisations in the 1980s, this was the first time AI moved from research labs into real business environments.
Why 1985 mattered to AI

After a period of reduced funding and limited progress in the 1970s [often referred to as the “AI winter”], the success of expert systems helped spark a renewed wave of investment and interest. This era became known as the AI boom of the 1980s, driven largely by the belief that human knowledge could be captured, structured and scaled through technology.
For the first time, businesses began to see AI not just as an academic concept, but as a practical tool – seems pretty weird to think about because most people think that AI has only just started to be used!
The limitations [and the lesson]
Despite their success, expert systems weren’t perfect as they relied heavily on manually created rules, which meant:
- They were difficult to maintain
- They struggled with unexpected scenarios
- They lacked the ability to truly “learn” from data
Over time, these limitations contributed to another slowdown in AI progress. But importantly, the lessons learned in the 1980s shaped what came next.
From rules to learning
While expert systems dominated 1985, the seeds of modern AI were already being planted.
Researchers were beginning to explore neural networks and machine learning, approaches that would eventually allow computers to learn from data rather than rely solely on predefined rules. Fast forward to today, and this shift underpins everything from:
- Autonomous systems
- AI assistants
- Predictive analytics
A moment worth remembering
Looking back at 1985, it’s clear that expert systems represented something bigger than just a technology trend. They were one of the first real demonstrations that AI could deliver value in the real world. And while the technology has evolved dramatically since then, the core ambition remains the same:
Using intelligent systems to support better, faster decision-making.
As we continue celebrating 40 years of CSG, it’s fascinating to reflect on how far technology has come and how ideas from decades ago continue to shape the innovations we see today.