AI Is Now Processing SBA Loans. What That Means for You.

What does the Celtic Bank and Casca partnership actually mean for small business lending?

If you've applied for an SBA loan recently, you know the process is slow. Weeks of document gathering, underwriter queues, back-and-forth on financials. The Celtic Bank and Casca partnership announced April 29 is a direct attempt to fix that. Celtic Bank is using Casca's AI platform to automate and accelerate parts of the SBA loan origination process, which historically has been one of the most paper-heavy lending workflows in the industry.

This matters because Celtic Bank is not a small player. It ranked third on the Business Journal's 2025 list of SBA lenders in the Central Valley by dollar amount, with $8.77 million in loans to businesses in a region where it has no physical branches. That's a lender doing real volume, remotely, now layering in AI underwriting support.

How does AI actually speed up SBA loan approvals?

Casca's platform sits inside the loan origination workflow and handles tasks that currently require human time: document parsing, data extraction from financial statements, preliminary eligibility checks, and application completeness reviews. Instead of an underwriter manually pulling figures from a borrower's tax returns and P&Ls, the AI does that first pass in minutes.

The practical effect is that humans focus on decisions, not data entry. Underwriters get a cleaner package faster, which compresses the time between application submission and a credit decision. For a borrower, that can mean days off the process, not just hours.

SBA 7(a) loans, the most common type, took an average of 30 to 90 days to close depending on lender, loan size, and document readiness before these kinds of tools became common. AI-assisted origination is being cited by lenders as a path to getting that closer to two to three weeks for well-prepared applicants.

Is this actually good for small business borrowers, or just good for the bank?

Honestly, both. Banks benefit because they process more applications with the same headcount. But borrowers benefit in a specific and underappreciated way: the AI catches incomplete or inconsistent applications earlier. Instead of submitting a package and waiting three weeks to hear that page 4 of your tax return is missing, the system flags it at intake. That kind of early feedback loop is genuinely useful.

There's a catch worth naming. AI underwriting systems trained on historical loan data can reflect historical biases. If certain borrower profiles were underserved before, the model may perpetuate that unless the lender actively audits for it. That's a governance question for Celtic Bank and Casca to answer publicly. Borrowers should know it's a fair thing to ask about.

The speed improvement is real. The fairness question is still open.

What types of SBA loans does this affect?

The announcement doesn't specify which SBA loan programs are in scope, but given Celtic Bank's volume and focus, SBA 7(a) loans are the most likely starting point. Here's a quick reference on the main programs:

| SBA Loan Type | Typical Use | Max Amount | Known For | |---|---|---|---| | 7(a) Standard | Working capital, equipment, acquisition | $5 million | Most common, most flexible | | 7(a) Small | Smaller capital needs | $500,000 | Faster processing path already | | 504 | Real estate, major equipment | $5.5 million | Fixed-rate, two-lender structure | | SBA Express | Fast turnaround needs | $500,000 | 36-hour SBA response commitment |

AI tooling will have the biggest impact on 7(a) Standard applications because those are the most document-intensive. If you're applying for an SBA Express loan, the timeline is already compressed by design.

Does this change how you should prepare a loan application?

Yes, in one important way. AI-assisted origination systems reward clean, complete, consistent documentation. A human underwriter can call you and ask a clarifying question. An AI parsing system will either flag the inconsistency for review or, in some cases, score it negatively without explaining why.

That means the preparation discipline that good borrowers always needed is now more important. Specifically:

• Match your numbers across documents. Your tax returns, bank statements, and P&L should tell the same story. Discrepancies that a human might overlook get caught by document parsing tools.
• Have your business financials current. An AI system checking application completeness will often require trailing 12 months of financials. If yours are six months old, that creates a delay.
• Explain anomalies in writing upfront. A COVID revenue dip, a one-time expense, a partner buyout. Write a brief narrative addendum. Some AI platforms surface these to underwriters; others just flag the anomaly. Give the human reviewer what they need to make a good call.

Is AI-assisted lending spreading beyond Celtic Bank?

Yes. Celtic Bank and Casca are one data point in a broader pattern. Lenders including Funding Circle, Biz2Credit, and Bluevine have used AI-driven underwriting for years on non-SBA small business loans. What's newer is AI tooling entering the SBA-guaranteed lending space, which has traditionally been more conservative and process-bound because of SBA compliance requirements.

Casca specifically is positioned as a workflow AI for banks, not a direct lender itself. That means other SBA lenders could adopt similar tools, and some likely will. The competitive pressure to reduce turnaround time is real. If one lender can approve a well-prepared borrower in 15 days and another takes 45, the market responds.

For SMB owners watching this: the lending landscape for SBA capital is changing faster than most people realize. That's worth paying attention to when you're planning your next capital raise.

What we'd actually do

• Get your financials in shape before you need capital. Reconciled books, current P&L, at least two years of tax returns ready to go. AI underwriting rewards preparation. Don't start this when you're already in a cash crunch.
• Apply with multiple lenders simultaneously. Celtic Bank operates fully remotely with no Central Valley branches and still ranked third by volume. Geographic barriers to SBA lending are eroding. You're not limited to local banks.
• If you want to understand how AI is affecting your financing options (and your operations more broadly), join the community at skool.com/aiforbusiness. We track what's actually changing in AI-driven business tools and help operators use that to their advantage.