Take Five #192: Read this before using AI to analyze sensitive company data, and more
Top five must-reads this week in the world of SMB acquisitions and operations
Subscribe to Take Five to get our top 5 quick weekly reads on the world of SMB, M&A, and EtA from the team at Kumo. Kumo aggregates hundreds of thousands of deals into one easy-to-use platform so that you can spend less time sourcing, and more time closing deals.
Take Five is created and sponsored by Kumo, a powerful deal aggregator to help supercharge your deal sourcing at withkumo.com.
What can you do with Kumo?
Browse 120,000+ deals from hundreds of brokers and every major marketplace, with 700+ unique deals added daily.
Save time and stop reviewing duplicate deals. Kumo matches identical deals across hundreds of sources so you can view unique business opportunities, even if they’re slightly different across different websites.
Get a daily email for deals that match your search criteria.
Take Five #192: Read this before using AI to analyze sensitive company data, and more
1. 👀 Read this before using AI to analyze sensitive company data (how to keep it discreet, reduce exposure)
Most companies face this binary choice with AI: Block it entirely for sensitive work and accept the productivity loss, or allow it freely and accept the data exposure risk.
There’s a third option, and it’s borrowed from cryptography, but requires zero crypto expertise.
Foundational models keep, track, and ‘may’ look at your llm calls to their models. So how do you use without giving them sensitive data (either for privacy or actual compliance)?
Your finance team can’t use ChatGPT to analyze P&L because it contains actual revenue numbers.
Your legal team can’t use Claude for contract review because client names are confidential.
Your HR team can’t leverage AI for comp analysis because... obvious reasons.
But... the AI doesn’t actually need to know you’re analyzing “data center cooling systems.” It could think you’re analyzing “Halloween costume sales” and the math still works. The relationships, patterns, and insights remain valid regardless of what you CALL the entities.
Entity Substitution Protocol
I call it “Homomorphic AI” (fancy term for a simple concept):
Replace sensitive data with fake placeholders
Send fake data to AI
Map AI’s output back to real data
The AI never sees your actual information.
A Concrete Example
Real contract you can’t show AI:
Liquid cooling systems: 45% discount
Edge computing hardware: 38% discount
Power distribution: 42% discount
What you send to AI:
Halloween costumes: 12% discount
Candy corn: 8% discount
Pumpkins: 15% discount
The AI analyzes “seasonal retail pricing.” You apply insights to your actual products.
The AI suggests: “Your Q4 pumpkin sales could benefit from bundling with costume purchases based on margin analysis...”
You translate: “Our Q4 power distribution could benefit from bundling with cooling systems based on margin analysis...”
Same insight. Zero exposure.
Read the rest of Josh Schultz’s article here.
2. Another reminder that seller notes can hide real risk in the fine print
3. “SBA 504 Loan Use Cases for SMB Buyers”
SBA 504 loans aren’t just for fixed asset purchases - they also support business acquisitions and expansions when substantial fixed assets are involved. These loans can finance projects that include owner-occupied real estate or long-term machinery, much like those for purchasing property or constructing facilities. However, the program doesn’t cover working capital, inventory, or speculative real estate investments. Occupancy requirements remain 51% for existing buildings and 60% for new construction.
Eligible Project Costs Financed
A 504 loan can cover a range of costs, including the purchase price, new construction, renovations, and site improvements. Even soft costs - like appraisal fees, legal expenses, and title insurance - can be included in the financing. Modernization efforts, such as upgrading facilities to meet health and safety standards or integrating robotics, are also eligible. For machinery or equipment to qualify, it must have at least 10 years of useful life remaining.
Down Payment Requirements
The down payment structure for acquisitions and expansions mirrors that of fixed asset financing. Established businesses operating for two or more years need a 10% down payment. Startups or businesses with new ownership must put down 15%. If the project involves a special-use property - like a hotel, cold storage facility, or winery - an additional 5% is required. For example, a startup acquiring a special-use property would need a 20% down payment. This is still more affordable compared to conventional commercial loans, where lenders often demand 20% to 30% upfront.
Job Creation Mandates
SBA 504 projects must create or retain one job per $90,000 of financing. Small manufacturers get a bit more leeway, requiring one job per $140,000. If your project doesn’t meet these job creation numbers, you might still qualify by meeting public policy goals. These could include promoting rural development, supporting veteran-owned businesses, or improving energy efficiency by at least 10%. In such cases, the Certified Development Company (CDC) overseeing the loan must maintain an average job creation rate across its portfolio to approve these exceptions.
Suitability for SMB Buyers
To qualify, a business must have a tangible net worth under $20 million and an average net income of less than $6.5 million (after taxes) over the last two years. The SBA portion of the loan is typically capped at $5 million but can go up to $5.5 million for manufacturing or energy-efficient projects. Fixed interest rates are tied to U.S. Treasury rates, and total fees are limited to 3% of the loan amount.
4. Business owners benefit from getting comfortable with being uncomfortable
5. “Boring” operational assessments can help SMB buyers understand and improve what they inherit post-close
Insight from an Operator’s Operator
We spoke with Meredith Grace about this challenge of finding a company with good operations. She started off in operations, not deals, and has personally operated around a dozen companies.
She now works in two main lanes:
Helping ETA buyers with operational diligence and post-close execution.
Helping burnt-out owners stabilize, turn around, and position their companies for a future sale.
Her work ranges from “keeping the company from dying” to “getting it ready to sell for a lot more than it is worth today.” When she talks about operational misses, Meredith can pull from a growing list of mistakes made by real companies that she has worked with to correct.
The Yin and Yang of Operations and Search
Operating a company is almost the opposite of how most searchers are wired. When you’re searching, you tend to ask high level questions such as:
What are we really buying and what type of person is selling it?
Why does this company exist and what is the core purpose?
What are the few goals that actually matter for the next couple of years?
A searcher will use those answers to drive everything else in a deal: offer, pricing, priorities, hiring, what to stop doing, etc.
But when you employ an operator’s mindset, you’ll ask much more practical questions. So practical that most searchers would prefer to call them “boring” questions. But Meredith knows from experience that there is a ton of value behind these questions. Here’s a few examples she shared with me:
Are the chart of accounts and expenses in the right buckets?
Is QuickBooks using the classes correctly so you can slice performance in a useful way?
How often does the team meet, and what’s on the agenda?
What is the company’s north star?
None of these questions sound exciting in a deal memo. However, the answers give you exactly the kind of structure you need if you want the financials to actually mean what you think they mean. Apologies if you had to read that sentence twice.
Read QOE Prep’s post here.
Loved what you read? Subscribe to Take Five to get our top quick reads every week from the team at Kumo. Kumo aggregates thousands of sources into one easy-to-use platform so that you can spend less time sourcing, and more time closing deals.