AI Won’t Replace Underwriters — But It Will Replace Bad Rules
AI Won’t Replace Underwriters — But It Will Replace Bad Rules
Artificial intelligence is everywhere in financial services right now. From fraud detection to customer support to document processing, AI is reshaping how institutions operate.
Naturally, mortgages are next.
Some headlines suggest AI will soon replace underwriters altogether. Others promise “instant approvals” and fully automated lending decisions with no human involvement.
But that’s not what’s actually happening — and it’s not what should happen.
AI won’t replace good underwriters. What it will replace are the outdated, rigid rules that prevent underwriters from doing their jobs well.
The Real Bottleneck in Mortgage Lending Isn’t People
For decades, underwriting has been governed less by professional judgment and more by policy constraints designed to satisfy secondary market investors.
Underwriters don’t just evaluate borrowers — they evaluate whether a loan fits into predefined boxes that can be easily sold, securitized, and priced.
That means:
- Income must look a certain way
- Credit profiles must follow predictable patterns
- Variability is treated as risk, even when it’s sustainable
- Exceptions are discouraged, even when they’re logical
When a borrower is denied, it’s rarely because an underwriter believes they can’t pay.
It’s because the system doesn’t allow flexibility.
AI doesn’t fix that on its own. But it can expose just how unnecessary many of those rigid constraints really are.
What AI Is Actually Good At in Underwriting
AI is extremely good at pattern recognition across large datasets. It can analyze:
- Income behavior over time
- Spending stability
- Payment consistency
- Asset flows
- Risk correlations that aren’t obvious in single snapshots
This is powerful — not because it removes humans, but because it gives them better tools.
Instead of relying on blunt proxies like:
- W-2 employment
- Fixed salary
- Perfect credit trajectories
AI can help assess what actually matters:
Can this borrower realistically and sustainably afford this home?
That’s the core of ability-to-repay, and it’s where underwriting should have always focused.
Why Rules Took Over Judgment in the First Place
Mortgage rules didn’t become rigid because lenders wanted to deny people.
They became rigid because large-scale lending depends on predictability.
Loans need to be:
- Easily priced
- Easily sold
- Easily modeled by investors
Human judgment introduces variability, and variability complicates securitization.
So over time, underwriting shifted from:
- Evaluating borrowers
to - Enforcing eligibility rules
Underwriters became compliance gatekeepers rather than financial evaluators.
AI doesn’t automatically change that — but it makes it much harder to justify continuing it.
AI Can Highlight Where Rules Don’t Match Reality
When AI models consistently show that certain borrowers:
- Have stable income despite variability
- Maintain long-term payment behavior
- Perform similarly to “qualified” borrowers
It becomes increasingly difficult to argue that they should be excluded purely because they don’t fit traditional categories.
In other words, AI can prove what many underwriters already know:
The rules are outdated.
And once that becomes visible at scale, pressure builds to change them.
Why AI Alone Still Isn’t Enough
Even with better risk analysis, traditional lenders still face structural limits.
They still must:
- Sell loans at par
- Meet investor eligibility criteria
- Fit within government-backed underwriting frameworks
So even if AI says:
This borrower is low risk
The loan can still be rejected because:
The capital markets won’t buy it
That’s why true innovation in lending can’t happen only inside existing mortgage structures.
You can upgrade the intelligence, but if the economic incentives don’t change, the outcomes won’t either.
Where Doorly Takes a Different Approach
At Doorly, we don’t treat underwriting as a box-checking exercise.
We treat it as what it should be: a real assessment of whether someone can afford a home responsibly.
That means evaluating:
- True income behavior, not just income format
- Cash flow consistency
- Financial sustainability
- Overall risk, not just isolated metrics
AI helps analyze that data efficiently and objectively.
But final decisions are grounded in:
- Ability-to-repay principles
- Long-term sustainability
- Real-world financial behavior
Not rigid eligibility templates.
AI as a Tool, Not the Decision-Maker
Good underwriting requires context.
Life isn’t linear. Income isn’t perfectly predictable. Financial stability doesn’t always show up in neat forms.
AI can:
- Surface insights
- Flag risk patterns
- Identify positive signals
But it shouldn’t be the final authority.
Underwriters still play a critical role in interpreting what the data actually means for a specific borrower in a specific situation.
The goal isn’t automation for its own sake. The goal is better decisions.
What Will Actually Change Because of AI
AI won’t eliminate underwriting jobs. What it will eliminate are excuses.
Excuses like:
- “We can’t model that risk.”
- “Our systems can’t evaluate that income.”
- “That borrower is too complex.”
As technology improves, the industry will no longer be able to hide behind technical limitations to justify outdated policies.
The conversation will shift from:
We can’t underwrite this
to:
We choose not to
And that’s a much harder position to defend.
The Future of Underwriting Is Human + Machine
The best lending decisions won’t come from:
- Humans alone
or - Machines alone
They’ll come from systems where:
- AI handles complexity and scale
- Humans apply judgment and accountability
- Rules evolve based on real performance data
That’s how underwriting becomes smarter without becoming careless.
And that’s how access expands without increasing systemic risk.
Replacing Bad Rules Is More Powerful Than Replacing People
AI doesn’t need to take underwriters out of the process to transform lending. It only needs to make it impossible to justify policies that were built for a workforce and economy that no longer exist.
When rules catch up to reality, underwriting becomes what it was always supposed to be: A thoughtful evaluation of real people, real income, and real financial capacity. Not a checkbox exercise.
That’s the future Doorly is building toward — where technology supports better decisions, not narrower ones.