The Math That Doesn't Work
If you've ever tried to sell a small business, you know the first problem: finding the right buyer. Not a buyer — the right buyer. Someone with operator experience, capital in the right range, and a thesis that matches what the company actually is.
Historically, M&A firms solved this by throwing research teams at it. OffDeal, a small-business M&A platform, used to spend about $12,000 and a full week on each new search — junior analysts combing databases, checking LinkedIn, cross-referencing funding histories, and filtering.
Most of that work was repeatable. None of it was what the bankers were hired for.
What Changed in 2026
OffDeal rebuilt the workflow on the Claude Agent SDK. Instead of a team, one agent now does the entire search pass: identify candidate buyers, score them against the listing's thesis, pull recent signals (funding, hiring, M&A activity), and hand a ranked shortlist to the banker.
The numbers that came out of that rebuild:
- Cost per search: $12,000 → $200. A 60× reduction. Most of that is API + data cost; the human in the loop reviews the final ranked list instead of building it.
- Eval accuracy: 25% → 85%. OffDeal runs every candidate list through a held-out eval of "buyers a banker would actually call." The agent's list is 3.4× more likely to include those than the old human pipeline was.
- Throughput: 2 bankers closed 8 deals totaling $91M. Classical M&A shops need 4-6 bankers for that kind of volume because of the research overhead. OffDeal does it with 2 plus the agent.
The full case study is on Anthropic's customer page.
Why This Matters to an SMB Owner (Who Isn't Selling Their Business)
The OffDeal pattern isn't about M&A. It's about a specific shape of work:
- You have a repeatable research workflow.
- The inputs are messy (LinkedIn, databases, spreadsheets).
- The output is a ranked list of candidates with reasoning.
- The cost of getting it wrong is high enough that humans still need to review the final list.
That shape shows up everywhere in SMB operations:
- A real-estate agent finding buyers for a new listing.
- A recruiter filtering 400 applicants down to the 15 worth calling.
- A bookkeeper deciding which past-due invoices to chase first.
- A construction firm evaluating which subcontractors to re-use on a bid.
In every one of those cases, the research-to-ranking pipeline can move from people-hours to agent-minutes, with a human doing only the final judgment call. That's the economic story of 2026 AI — not replacing the judgment, replacing the grunt work that feeds it.
How Ascero AI Applies the Pattern
Our Inbox Hunter Agent Pack ships this exact workflow for SMB sales teams. Upload your cold lead list, the agent enriches each record with real-time data, scores against your ICP, pulls contact details, and hands you a weekly top-50 to actually call. Same mechanism OffDeal built — compressed for an SMB budget.