AI in Customer Experience: The AI You Never Noticed

You probably think AI became part of your life recently. You opened a chat app, typed a question, watched it answer like a person, and thought: so this is the AI everyone means. Here is the uncomfortable truth. You have been an AI customer for well over a decade — and AI in customer experience began long before any chatbot. The chat window is not where your relationship with artificial intelligence started; it is simply the first time anyone let you see it. Behind the scenes, the world’s biggest enterprise-technology companies were already working every day to quietly make your life better with AI, and you never once called it that.

Consider the engine under all of it. Long before any smart designed chatbot, Nvidia released a platform called CUDA in 2006 that let its gaming graphics chips do general-purpose math. In 2012, a neural network called AlexNet, trained on just two Nvidia gaming cards, shattered the record in a global image-recognition contest — the moment the modern AI era actually began. The researcher behind it, Geoffrey Hinton, later said flatly that it would not have happened without Nvidia. Every photo your phone auto-sorts, every translation, every voice assistant traces back to that quiet hardware breakthrough. You were benefiting from it years before the word “AI” entered your daily vocabulary.

That is the pattern worth understanding. AI was never a fashion that arrived with the chat boom. It was infrastructure — built patiently by business-to-business companies, hidden inside the products and services you already used, and improving your experience while asking nothing of your attention. Here is how a few of the giants did it, each in their own way, and each still pushing forward.

1. Samsung: the intelligence in the device you hold

In March 2017, Samsung put a voice assistant called Bixby into the Galaxy S8. What made Bixby distinctive was its ambition: unlike rivals that handled a few set commands, Samsung designed it to take near-complete control of supported apps through natural, context-aware speech — you could mix voice and touch, give incomplete instructions, and let it infer the rest. For years, millions of people carried a genuine AI assistant in their pocket and simply called it “my phone.”

But Samsung’s real AI advantage was never the assistant you noticed — it was vertical integration. Because Samsung makes the phone, the camera sensor, the memory, and the chips, it could embed machine learning at every layer: the computational photography that decides how to expose and sharpen each shot, the models that manage battery and memory, and the AI-driven visual inspection in its semiconductor fabs that catches microscopic defects no human could see. The intelligence spanned the entire stack, from the factory that made the chip to the photo in your hand.

2. SAP: the AI that quietly ran the businesses you rely on

You may never have logged into SAP software, but it has shaped your life as a customer — it runs the supply chains, payrolls, and logistics of a huge share of the world’s large companies. In January 2017, SAP launched a machine-learning layer called Leonardo across its enterprise applications. What was special about SAP’s approach was where it placed the intelligence: not in a flashy consumer-facing app, but deep inside the transactional core where the actual work of commerce happens.

The results were concrete. One early service, Service Ticket Intelligence, automatically read incoming customer-service tickets, categorized them, and proposed solutions. Another, SAP Cash Application, matched incoming payments to invoices automatically — dull, invisible, and enormously consequential for how fast businesses run. SAP’s insight was to treat AI not as a feature to show off, but as a quiet upgrade to the machinery of commerce itself: the reason a product was on the shelf when you wanted it, or a refund cleared before you had to chase it.

3. Oracle and Salesforce: the AI behind every interaction

The software that remembers you as a customer — that knows your history when you call, emails you the right offer, and routes your issue to the right person — has been AI-powered for years. In September 2016, Salesforce embedded an AI layer called Einstein across its customer-relationship products, scoring which leads to prioritize, predicting what a customer needed next, and classifying support cases automatically.

What made Einstein special was scale through automation. Most companies each have different data, workflows, and needs, so a single model cannot serve them all. Salesforce built what its engineers nicknamed a system to “build AI to build AI” — automatically generating a custom, self-tuning model for each of its many thousands of customers rather than hand-crafting one at a time. That is why Einstein could quietly personalize your experience across countless companies at once, and why Salesforce now reports it generates more than a trillion predictions a week. Oracle took a complementary path, weaving machine learning through its cloud applications and, notably, into the database itself — its “autonomous” database used AI to tune, patch, and secure itself with little human intervention, so the services built on it simply ran faster and more reliably. One company made the customer interaction smarter; the other made the very foundation self-managing — two different bets on where intelligence creates the most leverage.

4. Beyond the software: AI in the physical world

The quiet benefit extends into things you can touch. When your car warns you that you are drifting out of your lane, you are relying on computer-vision AI pioneered by companies like Mobileye, founded in Jerusalem in 1999; by 2016 its chips were already in roughly 16 million vehicles. When Amazon gets your order to you at astonishing speed, AI is managing the robot fleets, routing, and inventory placement inside its fulfillment centers; at Siemens, the same kind of AI runs predictive maintenance so factories and infrastructure fail less often. None of this ever announced itself to you. It just showed up as safer roads, faster deliveries, and things that break down less.

5. Why B2B benefited you first — and quietly

There is a simple reason enterprise technology delivered AI to you long before consumer apps did: that is where the hard problems and the valuable data lived. A factory, a hospital, a global supply chain, a support center handling millions of cases — these gave AI both a reason to exist and the raw material to learn from. And crucially, B2B companies did not need you to notice. Their goal was not to impress you with a clever chat; it was to make the underlying product cheaper, faster, safer, and more reliable. The best infrastructure is invisible. You judge it only by whether your life works — and increasingly, it worked because of AI you never saw.

6. Where this is heading — and how fast

For most of the last two decades, this progress was a slow burn: a better forecast here, a smarter camera there. That pace has changed. According to McKinsey’s 2025 research, the vast majority of organizations now use AI in at least one part of their business, and the frontier has moved to “agentic” systems that can plan and carry out multi-step tasks on their own. It feels sudden only because the groundwork was laid invisibly since 2006 — the data pipelines, trained models, hardware, and enterprise plumbing. Generative and agentic AI are not starting from zero; they are accelerating on top of infrastructure that quietly matured for two decades, which is why a capability can leap from novelty to everyday tool in months. Notably, McKinsey’s own reading is that the winners pair this power with human judgment rather than remove people.

7. What this should change in your own organisation

The history is only useful if it sharpens how you act now. A few things worth weighing against your own organization:

Judge AI by the customer outcome, not the interface. The companies here won by making a product faster, safer, or more reliable — not by adding a visible bot. Before funding an AI project, ask which customer outcome it moves. If the honest answer is “it looks modern,” that is a warning sign.
Put intelligence where the leverage is, not where it shows. Samsung embedded it across the stack, SAP in the transactional core, Oracle in the database itself. The highest-value AI is often the least visible. Map where your real bottlenecks and richest data sit, and start there rather than at the shop window.
Your data is the moat, so treat it that way. Enterprise AI worked because these firms had proprietary data no one else held. Before chasing models, ask what unique data your organization owns and whether it is clean, connected, and usable — that asset compounds while models commoditize.
Design the human hand-off before you automate. The forward consensus is human-plus-AI, not human-out. Decide in advance which decisions AI may make alone and where a person must own the outcome — build that boundary in from the start rather than discovering it after something goes wrong.
Build on the escalator, not from the basement. Because the foundations already exist, you rarely need to build AI from scratch — you can stand on mature platforms and move fast. The competitive question is no longer “should we use AI” but “how quickly can we rewire a workflow around it before a rival does.”

The bottom line

AI did not suddenly appear; a technology that had been improving your customers’ lives for over a decade simply became visible enough to name. The quiet years built the engine — the current moment is that engine shifting into a higher gear, in the open. For any organization, the lesson is the same one these giants learned early: serve the customer so well, with intelligence placed where it truly helps, that the benefit is obvious even when the machinery is not. That is not a fashion that will fade. It is infrastructure — and the advantage goes to whoever builds on it with the most intent.

👉 Stay ahead of CX, AI, and innovation — Subscribe to my weekly LinkedIn Newsletter “CX Insights by Ricardo S. Gulko.

Share it if it’s useful — and let’s connect for more. Ricardo Saltz Gulko

My columns in several respected CX publications.