Education · A brief history
AI in business didn't start in 2022.
The LLM revolution feels sudden, but it's the seventh wave of intelligent technology to hit business — and every wave has followed the same pattern: skepticism, early movers, quiet advantage, and then everyone else scrambling to catch up. Here's how we got from punch cards to systems that answer your phone.
Seven decades in seven stops
The road to the LLM revolution.
The foundations
Alan Turing asks whether machines can think (1950); the Dartmouth workshop coins the term "artificial intelligence" (1956). Business computing, meanwhile, is gloriously mundane — payroll, inventory, punch cards. The two worlds haven't met yet, but the important thing has happened: businesses have started trusting machines with work.
Business milestoneThe world's first business computer wasn't built by a tech company — it was LEO, built in 1951 by J. Lyons & Co., a British chain of tea shops, to run bakery valuations and payroll. Early-mover advantage, chapter one.The expert systems era
The first commercial AI: expert systems — thousands of hand-written if-then rules encoding what a human specialist knows. They worked, in narrow lanes, and corporations paid millions for them. Then their brittleness caught up: the rules couldn't learn, maintenance costs ballooned, and the field slid into the "AI winter" of frozen budgets.
Business milestoneDigital Equipment Corporation's XCON system (1980) configured complex computer orders automatically and saved the company an estimated $25M a year — proof that encoded expertise had real ROI, even before machines could learn.Machines that spot patterns
AI stops trying to encode human rules and starts learning from data. Neural networks quietly go to work on credit card fraud, banks score loans statistically, airlines invent yield management. Deep Blue beats Kasparov in 1997 and makes headlines — but the money is in the invisible stuff running behind every swipe of a card.
Business milestoneFraud-detection neural networks were screening a huge share of U.S. card transactions by the mid-90s — for most consumers, the first AI that ever touched their lives worked at a bank, not a lab.The web learns what you want
The internet turns machine learning into a growth engine. Google ranks the web and auctions ads with algorithms; Amazon's "customers also bought" recommendations drive a meaningful slice of its revenue; Netflix offers $1M to anyone who can improve its recommender. Personalization becomes the quiet superpower of e-commerce — and the first taste of AI deciding who gets recommended.
Business milestoneThe Netflix Prize (2006) made recommendation algorithms a public spectator sport — and signaled that predicting customer preference was now worth serious money.The deep learning decade
In 2012, a deep neural network crushes the ImageNet image-recognition benchmark, and everything accelerates. IBM's Watson wins Jeopardy! (2011); Siri (2011) and Alexa (2014) put AI's voice in pockets and kitchens; predictive analytics spreads through sales, logistics, and marketing. And in 2017, Google researchers publish "Attention Is All You Need" — the transformer architecture. Nobody outside the field notices. It's the most important business paper of the decade.
Business milestoneVoice assistants trained an entire generation of customers to expect answers by asking — the exact behavior that now routes "who should I call?" through AI.The large language revolution begins
Transformers plus internet-scale text produce something new: models that write, reason, and converse. GPT-3 (2020) makes it available through an API; then on November 30, 2022, ChatGPT gives it a chat box — and reaches an estimated 100 million users in about two months, the fastest consumer adoption in history to that point. For the first time, AI needs no data science team. It needs a sentence.
Business milestoneThe interface collapse: seventy years of AI required experts to operate. ChatGPT made the operator anyone who can type — including your customers.AI goes to work
Copilots land in office software, Google ships AI Overviews above its own search results (2024), assistants gain live web access and start taking actions — booking, calling, buying. In local business, the change is concrete: AI now answers phones, qualifies leads, documents losses, scores calls — and, most consequentially, recommends companies by name when customers ask who to call. The seventh wave isn't a tool your business might use. It's a layer your customers already do.
Business milestoneEvery prior wave automated work inside the business. This one also stands between you and your customer — which is why visibility in AI answers became a discipline of its own.The pattern
Same story, seven times.
A tea company bought the first business computer. A hardware maker banked millions on encoded expertise while rivals called it a fad. Banks ran neural networks a decade before "AI" was a headline. Amazon built recommendations while competitors built bigger catalogs. In every wave, the businesses that moved while the technology still felt strange bought years of advantage at a discount — and the ones that waited paid full price to catch up.
The LLM wave is following the script exactly. The difference this time is speed: the gap between "early" and "late" is being measured in months, not years. That's not a reason to panic. It's a reason to start with a diagnosis of where AI already touches your business — what it says about you, what it could answer for you, and what it should be doing by next quarter.