This Lender Used AI to Process 40% More Loans. The Playbook Works for Any Business

NewtekOne (NASDAQ: NEWT) just reported Q1 2026 results that changed the math on AI automation. They processed 961 loans in Q1, a 40% increase year-over-year, after launching their "Seven Day Business Loan" using AI. In March alone: 500 loan units. That's a 74% increase. CEO Barry Sloane didn't hide the reason: AI reduced the time to close, cut the cost to close, and attracted higher-quality credit applications.

The question every business owner should ask: if a publicly traded lender just proved AI automation works at scale, what's stopping you?

The Bottleneck Everyone Has

Loan processing isn't unique. Every business has a process that's killing efficiency: reviewing invoices, onboarding customers, vetting proposals, approving requests. The work isn't hard. It's repetitive. Someone has to check that the numbers add up, that the document is complete, that the application meets basic criteria. A human can do it in 5 minutes. But when you multiply by 100 applications a day, you need someone who costs $50K a year doing nothing but checking boxes.

NewtekOne faced the same math. Instead of hiring more processors, they automated the routine checks: document completeness, basic financial ratios, credit history flags, missing fields. The AI doesn't make the final decision. A loan officer still approves every loan. But the AI handles the triage, the document review, the preliminary scoring. What took 2 days now takes 4 hours.

That's where the 40% increase comes from. Not magic. Just math.

Practitioner Insight: Why This Actually Works

I've built 120+ automations across 15+ industries. The ones that stick follow NewtekOne's pattern: amplify human judgment, eliminate human drudgery.

Here's what doesn't work: trying to automate the decision. You'll fail because every business has edge cases. The lender with perfect credit but no established business history. The invoice that doesn't match the contract because the vendor made a typo. The proposal that's brilliant but doesn't fit the standard template. These need human judgment.

Here's what does work: automate everything up to the judgment call. Extract the data. Check it for completeness. Flag anomalies. Score risk. Queue it for review. A human spends 5 minutes reviewing a fully prepared decision package instead of 30 minutes hunting through documents.

NewtekOne's "Seven Day Business Loan" isn't fast because AI made the loans. It's fast because AI handled the paperwork.

The 3-Step Playbook That Works

Step 1: Map the Process
Document exactly what happens today. NewtekOne probably found something like: 1) Applicant submits documents, 2) Processor reviews for completeness, 3) Processor pulls financials and credit, 4) Processor scores risk, 5) Loan officer approves or denies. Steps 2-4 are your target. They're mechanical. They follow rules.

Step 2: Identify the Judgment Calls
Which decisions need a human? In lending, it's the final approval. In invoice processing, it's deciding whether an error is a typo or fraud. In customer onboarding, it's deciding whether to escalate a flag. These don't automate. But everything feeding into them does.

Step 3: Build the Automation Chain
Automate data extraction, validation, and scoring. Feed the results to your human decision-maker. The output: faster decisions, fewer errors, better decisions because the human isn't exhausted from paperwork.

From my 120+ projects: this playbook cuts decision time by 60-75%. The businesses that ship fast are the ones that started with 100 applications a month and didn't try to solve the edge cases first. They automated the happy path. Fixed the edge cases as they appeared. NewtekOne processed 500 loans in March. They didn't wait for perfection.

Where This Actually Works

Invoice Processing: Extract vendor name, amount, date, line items. Check against PO. Flag duplicates. Route for approval. ROI: 15 hours recovered per week per processor.

Customer Onboarding: Extract application data. Cross-check references. Flag missing documents. Score risk. Route to review. ROI: 12 hours recovered per week, faster time-to-value for customers.

Proposal Generation: Client briefs proposal engine. Engine pulls template, inserts client data, generates draft, flags customizations needed. Sales team reviews and sends. ROI: 6 hours per proposal, not 20.

Expense Approval: Employee submits receipt. System extracts amount and category. Checks policy. Routes to manager if needed. ROI: cuts approval time from 5 days to 2 hours.

The pattern is identical in all of these. The only thing that changes is what the routine checks are.

The Questions You Should Ask

How long does this take to build?

With a clear process map: 4-8 weeks. NewtekOne didn't launch the Seven Day Loan overnight. But they also didn't take a year. The businesses that move fastest are the ones that start with a specific bottleneck, not "let's AI our entire operation."

What if our process is completely custom?

The playbook doesn't care. You still have routine checks (automate) and judgment calls (humans). The custom part is identifying where those boundaries are. That's a 2-week discovery phase, not a blocker.

How much does this cost?

Build cost: $15K-$40K depending on system complexity and data sources. Operational cost: API fees, maybe $200-$500/month for 1,000+ transactions. ROI math from my 120+ projects: average payback in 8-12 weeks.

What if we only have 50 transactions a month?

You don't have an automation problem. You have a hiring problem. Automation pencils out when the repetition is so high that the time savings exceed the build cost. NewtekOne processes 500 loans a month. That's volume. If you process 50 invoices a month, a single part-time person is cheaper. But if you're growing and you can see 500 invoices a month in 18 months, build it now.

How do we know if our team will accept it?

They will, because you're not replacing them. You're eliminating the worst part of their job. Invoice processors don't love reviewing documents. Loan officers don't love triage work. They love making decisions. Show them that the automation handles the prep work and suddenly they're focused on judgment, not paperwork. Adoption happens fast.

Stop Building Theories, Start Building Results

NewtekOne's 40% increase in loan volume isn't theoretical. It's not a pilot. It's a 961-loan quarter from a public company that had to report this to shareholders. The automation works because it's built on a principle that never changes: amplify human judgment, eliminate human drudgery.

The question isn't whether this playbook works. It works. The question is whether you'll deploy it while your competitor is still researching.