What is AI? A Primer
Society has long been fascinated with the idea of machines that think, act independently, and maybe even have emotions. In our earliest history, the “Golem” represented the concept of a being created from inanimate components that could be imbued with motion, action, understanding, and feeling. More recently, the concept has been expressed as well-known characters such as C-3PO from the Star Wars franchise, or Data from Star Trek.
Often, the fiction of artificial intelligence is far more fantastic than the current reality; for one, human emotion is something our creations can only mimic for now. They also currently lack the ability to grow beyond their programming in ways that could make them truly independent. And actual thought? While we might be close to something like a thinking machine, their abilities in this area remain far inferior to those of humans.
It’s unlikely to remain that way for very long.
In the meantime, society has taken an abiding interest in all things AI, and marketers have taken note. Everything shy of breakfast cereal is touting AI as a foundational component—but it can be very difficult to understand what AI truly is, and whether it is contributing meaningfully to those things claiming to use it.
AI doesn't mean one single thing—it’s a collection of technologies that simulate some form of human thinking or behavior. But in general, Artificial Intelligence refers to computer-based or computer-driven systems that are capable of doing things that typically require human effort. Some of the most common of these are understanding language, identifying images and patterns, making decisions, and learning from data.
What is the modern history of AI?
The idea behind practical AI has been evolving for decades—well before LLMs and chatbots entered the scene.
The First True AI: Logic and Reasoning
The development concepts of AI date back to the mid-20th century, when scientists and mathematicians began exploring whether machines could replicate human reasoning.
-
In 1956, the term “artificial intelligence” was coined at the Dartmouth Conference, a gathering now considered AI’s symbolic birth.
-
Early efforts focused on symbolic AI—models that mimicked human logic using rules and structured knowledge.
One of the first landmark systems was MYCIN (1970s), an expert system used for medical diagnosis. It could recommend treatments based on symptoms—but only by following predefined rules, not by learning.
This era showed that machines could simulate thought, but only in narrow, hardcoded ways.
The Rise of Machine Learning
In the 1980s and 1990s, a new idea took hold: rather than telling machines how to reason, what if we let them learn patterns from data?
This shift to machine learning meant:
-
Models could be trained on examples instead of rules
-
Accuracy improved with more data
-
AI could start adapting to new situations, not just repeating logic
This set the foundation for modern AI—including the deep learning systems we use today.
Deep Learning and the AI Boom
The 2010s brought a tipping point:
-
Increased computing power necessary for what researchers were trying to accomplish
-
Massive datasets made available for helping systems learn
-
New model designs, like convolutional and recurrent neural networks
AI began beating humans at tasks like:
-
Image recognition
-
Speech transcription
-
Games like Go and chess
Then came transformers—a groundbreaking architecture introduced in 2017 (not cars that turn into robots, though admittedly that would be kind of awesome). Transformers enabled AI to understand sequential data like language at scale and speed.
That’s what powers modern tools like GPT, Claude, and Botkeeper’s next-gen transaction classifiers.
Today: The Age of Applied AI
Today, AI is embedded into tools we use every day:
-
Predicting spending habits
-
Automating bookkeeping
-
Chatting with customers
-
Drafting content, code, and reports
But there’s a critical difference between:
-
General-purpose models that aim to do everything
-
And domain-specific AI designed to do one thing exceptionally well
Understanding that difference—and using the right AI for the right task—is what separates a flashy feature from a truly transformative solution.
What Are the Different Types of AI?
AI isn’t all one thing. There are different model types suited to varying tasks. This matrix helps break them down for you:
What AI Is Not
-
It’s not magic or human-level general thinking (yet).
-
It doesn’t have emotions or true understanding—it’s statistical and mathematical, not conscious.
-
Many tools marketed as “AI” are just automation or pre-programmed logic. While true AI can encompass both, neither on its own defines AI.
What Can’t AI Do (Right Now)?
Understand Context Like a Human
AI can mimic understanding, but it doesn’t really know:
-
Why a client’s transaction is unusual
-
That a company just went through a merger
-
How personal relationships influence financial decisions
Even the most advanced models struggle with:
-
Ambiguity
-
Sarcasm or humor
-
Unspoken cultural context
AI reads what is written—not what is meant, implied, or felt.
Exercise Judgment or Ethics
AI doesn’t know:
-
What’s “right” or “fair” in a moral sense
-
When to break a rule because the rule doesn’t fit
-
How to act with compassion or common sense
For accounting, that means AI can’t:
-
Decide whether a gray-area deduction is acceptable
-
Recognize when a client’s emotional state matters more than efficiency
-
Choose between technically correct and ethically sound
Adapt to Rapidly Changing Real-World Scenarios
AI models are trained on data from the past. When something changes fast—new tax law, pandemic rules, or client industries shifting overnight—AI doesn’t adapt unless it’s explicitly designed to adopt new information on the fly, or is purposely retrained.
It doesn’t “know” what just happened unless you tell it.
Apply Unrelated Data Effectively
A unique feature of human thought and logic is the ability to effectively adapt and apply information, techniques, or hypotheses learned elsewhere to a completely new situation. AI can certainly get creative, but it doesn’t excel as humans do at linking two totally different things, like grass clippings and packaging.
Be a Person
AI is a great imitator. While in the future, these words could look comedically out-of-touch, for now it’s impossible to transmit the human experience to a computer system. People are a mixture of nature and nurture, individual and shared experiences, trial and error, and most of all: context. Computers have none of these with which to address requests.
The lesson here is to take care in both expectations of and response to AI—it might seem human, but in truth it isn’t even close.
What is Licensed AI?
Commercially licensed AI is AI software/models you’re allowed to use for business purposes under a paid (or negotiated) license.
These models are frequently designed for general tasks, and are often used by wide-ranging businesses with products, clients, and purposes very different from one another.
What Are the Risks of Licensed AI?
“AI-powered” is everywhere. It’s on websites, in pitch decks, and attached to features that, just a few years ago, were called automation. But beneath the buzzword lies a fundamental truth:
Not all AI is created equal—and not all AI is created for your task.
That’s especially true when companies rely on licensed general-purpose AI, like GPT-4 or Claude, as the brains behind their product. These models are incredibly powerful—but they’re also generalists, trained to be good at everything and expert at nothing.
Let’s unpack the risks of this approach—and why purpose-built, domain-trained AI is often not just better, but essential.
What comes next for AI?
Technology development has been accelerating for some time, and there’s no reason to assume it might slow down anytime soon. The AI field is under intense research—in many ways, it’s spurring a race similar to the moon landing, with major countries competing to churn out the most advanced and capable AI possible.
Because of the educational and commercial benefits AI brings, rapid progress is all but assured. If AI is something you make use of in any of its forms, you’d be well-served to keep current on major developments so you can know what to expect down the road.
AI is and will continue to be a major factor in both life and business. Being educated about how it works and what it’s capable of will put you ahead of the crowd.