Everyone is talking about artificial intelligence these days, sometimes with excitement, sometimes with confusion or fear, and often with unrealistic expectations. Some imagine AI flipping burgers, flying spaceships, and managing every aspect of a business, all at once. But let’s take a step back. What is AI really capable of today, and what does that mean for private and hard money lenders?
At Liquid Logics, we believe in cutting through the noise. In this article, we’ll demystify AI by grounding it in how it actually works, what makes it powerful, and how private lenders can strategically apply it to achieve better, faster, and more secure lending decisions.
Understanding Human Decision-Making vs. AI: A Framework for Comparison
As humans, we constantly navigate the world using a series of conditional situational logic. In simpler terms, “if this, that or/and the other, then…”. Effectively, we build decision-tree making processes. However, our logic stems from our life exposure and experiences, our education, reading, wins, and losses. That’s what we call knowledge and intelligence. Obviously, that is limited to our experiences.
When adding consciousness and situational awareness, our ability to read between the lines, to feel the room, and to gauge intangible cues. That consciousness allows us to determine the level of complexity of the problem we are facing.
That’s something AI doesn’t have. However, it does have the capacity to aggregate large sums of data and experiences from countless sources.
Knowledge, Intelligence, and Consciousness
So, to summarize the overall picture for AI:
● Knowledge: The aggregation of data points AI has access to and considers as facts!
● Intelligence: Its ability to analyze, compute, and create permutations.
● Consciousness: Including self-awareness, situational awareness, is still 100% human.
AI has vast general knowledge and intelligence but lacks human consciousness. For example, in chess, AI might ignore that a knight can’t move diagonally if it’s desperate to win the next move and perceives a win is a must. If it runs out of options, it creates a case precedence in a law briefing and cites it to support its argument. So, when it reaches the end of the line and it exhausts its decision trees, and it is programmed/designed to find an answer, it creates the conditions or rules, then uses those new conditions as new facts, rules, because it lacks contextual awareness, and you may argue a moral compass.
That’s why conscious humans must define the rules and structure AI’s input and logic.
Where Does AI Get Its Intelligence?
AI is touted as intelligent, but where does its intelligence come from?
It starts with its ability to reason through “if-then” scenarios. The knowledge it uses comes from being trained on thousands (or millions) of use cases, business encounters, and decisions on what to do, and what not to do.
To function effectively, AI needs a baseline of general knowledge dataset. With that, it builds complex decision trees, combining every possible outcome of any general language statement to be able to understand what we are telling it and what we are trying to achieve in a given scenario. It then:
● Identifies valid decision branches
● Flags branches, it doesn’t understand (no data)
● Flags branches that it knows are incorrect
The rest of its knowledge is driven from building your own private lending AI knowledge base. More on that later.
Does AI know what’s right or wrong?
At its core, AI is not aware that this is right or wrong; it simply just knows correct and incorrect branches or answer sets. So, it doesn’t—at least not intuitively. It relies on feedback, rules, and context you provide. If it produces a result and we accept or reject it, that feedback becomes part of its learning. If it doesn’t have enough data to choose the right answer, it may guess; this is where the real risk lies.
There is growing evidence that AI over analyzes simple tasks to the level of overcomplicating them since it has intelligence but lacks consciousness, more alarming, it fails at the complex logic tasks OR tasks that require situational awareness. This presents a Dilemma of the level of trust you can give it for newly unverified arbitrary outcomes.
Interacting with AI: Why Language Matters
The most advanced AI tools today are built on LLMs (Large Language Models). These models don’t require traditional programming syntax. Instead, they understand natural language.
But there’s a common misconception:
Just because you say something in natural language doesn’t mean AI understands it the way you intended. Just because your command sounds clear to you doesn’t mean AI interpreted it the same way. The key challenge is aligning your intention with the language AI understands. Users must grasp this key point: You must give AI structured input — in the right sequence, with specific details.
A Practical Example:
Let’s say you upload five documents:
● A purchase contract
● A driver’s license
● An LLC operating agreement
● A bank statement
● An EIN letter
Then, you ask AI to match the signature on the purchase contract to the name on the driver’s license and ensure that the person is an authorized signer in the LLC agreement.
If you don’t clearly tell it how to process the documents or in what order, AI may only check two of the five, reaches what it thinks is a satisfactory decision tree branch, and give you an incomplete answer.
AI does not know what it doesn’t know. You must:
● Be verbose
● Structure your prompts
● Specify the type of outcome you want
So yes, while AI speaks your language, but only if you’re extremely clear, sequential, and intentional. A new vendor’s market is now emerging, selling specific “AI prompts.”
Building Your Own Private AI Knowledge Base
Popular models like ChatGPT, Gemini, and DeepSeek each interpret languages differently. Their ability to understand your intent depends on how their language models are trained; each has its own quirks. So while you need to know the nuance of each one to instruct it or “speak to it,” it also needs to know what you know about your specific industry and your specific logic or secret sauce.
To succeed in using AI for your business, you need a custom knowledge base not only for the private lending practices and rules, but also your own processes and secret sauce built around your operational and go-to-market advantages:
● Your business logic
● Risk mitigation standards
● Internal guidelines
● Specific credit and underwriting box(es)
Concerned about your data being shared with competitors? Don’t be. Any serious organization building AI workflows with tools like ChatGPT will develop a private, segregated knowledge base. This base is not part of the public model. It’s specific to your needs, rules, and priorities.
Garbage In, Garbage Out: Training Your AI Knowledge Base
A loose or vague description of your lending rules will lead to garbage results. AI is only as smart as the clarity of your input.
Feedback Loops: Teaching AI What Success Looks Like
● AI doesn’t know what it doesn’t know. That’s why feedback is vital.
● It learns through feedback. You must:
● Clearly define what counts as a failure
● Verbally affirm what counts as success
Also, make sure in your instructions and prompts to the AI engine if it is okay to make assumptions, OR tell it explicitly not to guess if it doesn’t have the data to support a valid outcome. Otherwise, it may invent an answer just to meet your expectations.
If a result is incorrect, tell it so. If it’s spot-on, mark it as a win. Over time, AI uses that feedback to refine its outputs.
But be cautious, if AI runs out of known outcomes, it may default to guessing. That’s why clear boundaries, validations, and rules are necessary. If a rule doesn’t exist, the AI can’t follow it.
Today, we call the people who structure these interactions AI engineers. They know how to craft prompts that guide AI in your language. Think of them as translators between your business and the engine powering your decisions.
Areas Where You Can Use AI in Private Lending
There are multiple opportunities for private lenders and funds to use AI agents
o LO Agents and assistants
o Voice agents and interactive chat agents, including borrower follow-up and engagement.
o Credit analysis and risk agents, including eligibility and pricing
o Processing and underwriting agents to help with document collection and evaluation.
o Servicing and payment management agents
Direct to AI engines, or out-of-the-box stitching tools together
Like every new technology, it is very tempting to try to hire “two men and a truck” to stitch concepts together! AI, as an evolving technology in particular, is being touted as the end game where you can get efficiency and full of point solutions and tools needing to be stitched together for an outcome promising speed and accuracy.
Private Lending, in particular, is riddled with entrepreneurial operators who admittedly acknowledge lack of tech knowledge, yet are eager to explore low-cost, quick fixes and solutions. While that may work in a few specific cases, when it comes to scaling, efficiency, and long-term growth will be greatly challenged if the tool set is not part of a cohesive system and well thought out approach.
Beware of outfits trying to sell you partial point solutions. The truth is that every outfit trying to do so is doing it because of their limitation and inability to deliver a comprehensive system that includes the AI tools you need, where you need it. Otherwise, you would be taken back a few steps, stitching your lending operation with spreadsheets and manual tracking, defeating the objectives and diluting the benefits you would achieve from the gains you get from AI to begin with.
Final Thoughts: What to Say and What Not to Say
Artificial Intelligence is powerful, but it’s not magical. It’s only as good as the structure, feedback, and guidance you provide. As leaders in private lending, it’s time to demystify AI, use it with confidence, and build smarter, more adaptive systems grounded in the knowledge and consciousness only your team can provide.
When working with AI:
● Be precise in your instructions
● Don’t assume it understands context
● Break down your process step-by-step
● Train it on what to do, and what not to do
To ensure that AI is not making assumptions, you still need to tell it what to do, how to do it, and what result you expect.
At Liquid Logics, we’re building more innovative solutions by blending powerful AI with decades of industry expertise. If you’re curious about how AI can be safely and effectively applied to your loan origination workflow, let’s talk.
AI doesn’t have consciousness or situational awareness, but you do. Use it to build your use cases something powerful.