Why doesn't daily AI use make you better at it?
Most loan officers use AI every day. Ask around your office and you'll find people running it constantly — for emails, for social captions, for quick research. But daily use and skill growth are not the same thing, and most people are only tracking the first one.
Here's the pattern worth noticing: type a question, read the answer, close the tab. Repeat tomorrow. That's usage. It is not the kind of deliberate practice that actually compounds into a skill.
Think of it like a driver's license. Someone can drive eight hours a day, every day, for a year, and still only be rated for a standard car. The hours logged don't upgrade the license — you have to go learn the next thing on purpose. AI works the same way. Asking it questions all day doesn't make you better at prompting it. Only deliberate practice does.
What is the Context / Goal / Constraint framework?
There are three parts to a prompt that actually moves the needle, instead of just producing an answer you have to fix:
● Context — Who do you want the AI to act as? What's the situation? Who is actually going to read or use what it creates? Skipping this is why generic AI output reads generic.
● Goal — What do you actually want it to produce? Not "help me with this" — the actual artifact. A follow-up email. A social caption. A comparison sheet. Vague goals get vague output.
● Constraint — What can it not do? This is the part almost everyone skips, and it's usually the difference between a usable draft and one you have to rewrite from scratch. Length. Tone. Whether jargon is allowed. In lending specifically, this is also where you build in guardrails: don't let it quote a rate, and don't let it guess at anything you haven't confirmed. A bad AI output is often just a missing constraint, not a bad tool.
How should loan officers apply constraints when using AI?
The constraint step matters more in this business than most, because the cost of a wrong or unverified output is higher than in most industries. Two constraints worth building into your default prompts:
● Never let it quote a rate. Rates move, are borrower- and scenario-specific, and are exactly the kind of number AI will confidently produce without knowing it's wrong.
● Never let it guess. If it doesn't have the information, the instruction should be to say so — not fill the gap with something plausible-sounding.
Beyond those two, the everyday constraints — length, tone, and how much jargon to use — are what separate a draft you can send as-is from one you have to heavily edit. Setting them up front, every time, is faster than fixing the output after the fact.
What's one AI rep a loan officer can start this week?
Pick one thing you already do every week or every day — a borrower follow-up, a referral partner check-in, a social post. Write the prompt with all three parts: who it's acting as and who's reading it, exactly what you want it to produce, and what it's not allowed to do. Then run that same structured prompt ten times over the next couple of weeks on that same recurring task.
That repetition — not just daily use — is the rep that actually compounds. Do it enough times and you're not just someone who uses AI. You're someone who's good at it.