Why Specialty Finance Is the AI Vertical Nobody Is Targeting
Harvey hit $11B with $190M ARR. Meanwhile $1.34T in equipment finance, 15,870 RIAs, and a wide-open vendor map sit ignored. Specialty finance is the bet.
TL;DR. AI capital has crowded into law firm AI (Harvey at $11B with $190M ARR) and call-center automation, but the real mid-market opportunity sits in specialty finance. Equipment finance is a $1.34T market with 71% of operators targeting AI for documentation. There are 15,870 SEC-registered RIAs, and only 15.84% of advisors use onboarding software. Factoring volume hit $139B in 2025. The vendor map is wide open and the unit economics are structurally better than law. If you run a $30-100M specialty finance firm, the math points one way.
There is a $1.34 trillion market sitting in plain view that the AI services industry has barely touched. Equipment finance. The 2025 ELFA Survey of Equipment Finance Activity reported 71% of operators targeting documentation as their primary AI use case, and 17.7% NBV growth in the independent segment while banks lost ground. The vertical AI vendor most operators in that market can name is "nobody yet." Meanwhile, Harvey raised $200M in March 2026 at an $11 billion valuation against $190M in ARR. That is roughly 58x ARR, in a sector where almost 60% of in-house counsel report no measurable savings from outside firms' AI deployments. The math is broken. This piece is about where it is not.
Why every AI vendor is fighting over the wrong vertical
If you spend any time on LinkedIn this quarter, you have seen three AI verticals get all the oxygen: law firms, call centers, and developer productivity. Each has structural problems that mid-market services firms are about to discover the hard way.
Law firm AI is overvalued against delivered productivity. Harvey is now $11B at $190M ARR, and yet the ACC/Everlaw productivity survey found 58% of in-house counsel saw no noticeable savings from outside counsel's AI tools. Clio's CEO Jack Newton has called the billable hour "structurally incompatible" with AI productivity. A $2,000 per hour partner using AI to do ten times the work captures the rent and does not pass it down. Mid-size firms grew demand about 5% in late 2025 while Am Law 100 firms grew about 2%. The cracks are there. The valuations are not yet correcting.
Call-center AI is real but narrow. Five9, NICE inContact, and Genesys are the buyers' shortlist. The deployments are measurable, the workflow is mature, the buyers are large, and the margins are commoditizing. A new AI services firm cannot land-and-expand here. The platform vendors will eat the application layer.
The interesting question is not whether these markets will keep getting funded. They will. The question is what other markets have the same operational-pain-times-buyer-budget product that nobody else is selling into. Specialty finance is the answer.
Where the actual money lives
Start with the size of the prize.
The Equipment Leasing and Finance Association reports the equipment finance market at $1.34 trillion in 2023, with 2024 holding above $1.3T. Independents grew net book value 17.7% while banks declined 1.3%. The independents are the buyers most starved for software leverage.
The Secured Finance Network 2025 Market Sizing Study puts asset-based lending commitments at $537B at year-end 2024, growing every year since 2018 and outpacing bank commercial-and-industrial growth. Factoring volume rose from $119.4B to $139.2B among long-term respondents in 2025. SFNet's CEO Rich Gumbrecht specifically called out shorter onboarding times "as firms invest in digital underwriting and faster verification systems."
The leveraged loan market sits at $1.55 trillion in outstandings by year-end 2025 per the Morningstar LSTA Index. The convergence of broadly syndicated lending with private credit means more BDCs are doing the work banks used to.
On the wealth side, Cerulli and the Investment Adviser Association count 15,870 SEC-registered RIAs at year-end 2024, up 3.1% year-over-year. Roughly 93% of them manage less than $1B in AUM. Too small for Goldman or Morgan Stanley to buy, too complex to run on spreadsheets and a single ops admin. This is the shape of the buyer almost every Granular discovery call describes when they talk about themselves.
These are not future markets. They are operating today, growing today, and they have document-heavy workflows that AI agents are unusually well-suited to compress.
The three paths into specialty finance AI
Three workflows that map to actual P&L lines and have benchmarked ROI today.
Equipment finance underwriting and residual value
A bank-affiliated equipment finance underwriter spends roughly 70% of their time on data extraction from scanned PDFs of dealer invoices, customer financial statements, and tax returns. The SEFA 2025 survey found 71% of equipment finance respondents are explicitly targeting documentation as their primary AI focus. Production deployments cut sub-$250K credit decisions from five days to same-day in 80% of cases. Residual value modeling, which most independents still do by gut and three Excel tabs, is open.
RIA back-office and onboarding
The 2025 T3/Inside Information Software Survey found only 15.84% of advisors use onboarding software at all. Fenergo's data shows the average financial institution spends $72.9M per year on AML and KYC alone, and 70% of firms reported losing clients in 2024 because onboarding took too long. Best-case AI deployments cut onboarding time 87%. For a $50M RIA, the founder is also the chief compliance officer, the head of operations, and the rainmaker. Onboarding compression is the difference between scaling to $150M AUM with the team they have and capping out at $80M.
Factoring and continuous credit monitoring
A single mid-market factor onboards 80 or more clients per year, each with their own PDF layouts and invoice formats. The FactorEvo + Decipher Credit hard-gate continuous-decisioning integration announced in 2025 is the first vertical-AI-meets-specialty-finance integration of note in trade press. BDC portfolio monitoring is the same problem at a different scale: Ares, Blue Owl, and KKR all dropped 8 to 12 percent in February and March 2026 on UBS's analysis that 25 to 35 percent of the private credit market has direct AI-disruption exposure to its software-borrower base. Blackstone built an internal AI risk scoring framework for BCRED. Most mid-market BDCs have not. The quarterly investor expectation is now AI-exposure scans every reporting cycle. The workflow to produce them is missing.
The compliance flywheel
The reason these workflows will get funded in 2026 is not the productivity story. It is the compliance one.
FinCEN imposed over $1.5B in BSA penalties in 2024. Fenergo logged $4.6B in global AML fines for the year and tracked a 417% YoY increase in the first half of 2025 alone. OCC Bulletin 2025-26 mandates explainability for AI used in community banks up to $30B in assets, which creates a moat for vendors who build explainability in from day one.
This is why specialty finance moves faster than law firms. There is no equivalent regulator standing over an Am Law partner saying you have eighteen months to demonstrate auditable AI-assisted compliance. There is one standing over the chief compliance officer at a $60M RIA. That asymmetry compounds.
Why specialty finance pays better than it looks
The TAM slide for specialty finance reads as fragmented. Eighteen thousand RIAs. A hundred top-tier equipment finance firms. A few hundred factoring shops. Compare that to "every Fortune 500 GC" and you can see why VCs pattern-match law first.
The unit economics tell the opposite story.
Specialty finance margins are spread and fee based. Efficiency gains drop to the bottom line because the buyer is not negotiating against a $1,000 per hour billing rate. They are negotiating against a 200 basis point spread on a $50M book. Every analyst hour saved is real margin. A services firm building reference deployments in these verticals captures higher-margin services wrap revenue because the buyer has no in-house AI team and is not running a vendor RFP this quarter. They are running an SEC exam prep. They want a partner, not a platform.
Compare to law: a 30% productivity gain at an Am Law 100 firm gets debated in client billing meetings and discounted in alternative fee arrangements. The capture goes to the firm, the client, and the partner's bonus pool, in roughly that order. The vendor sees the smallest slice.
Compare to call centers: a 30% productivity gain at a 5,000-seat BPO gets passed straight through to the enterprise buyer in renewal negotiations. The vendor sees zero net margin lift over two contract cycles.
Specialty finance is the inverse. The operator captures, the operator pays, and the relationship gets stickier each quarter because the regulator does not let them rip and replace.
The vendor map is wide open
There is no Harvey-equivalent for specialty finance. Hebbia is the closest analog, but its center of gravity is large capital markets and PE diligence, not the $30-100M operator running a factoring desk in Charlotte or an equipment finance shop outside Chicago. Hebbia raised $130M at $700M from Andreessen, Index, Thiel, and Google Ventures in July 2024 and acquired FlashDocs in June 2025. They are competing for Apollo's seat, not yours.
The capital is starting to assemble around the gap. Carlyle's AlpInvest AI Co-Investment IV-1 closed in October 2025. Keystone National Group put $44M into an AI/tech-focused specialty finance company in 2025 per ABF Journal. Apollo announced a $25B private credit JV with Citigroup in September 2024 that creates exactly the mid-market book that will need portfolio-monitoring AI in 18 months. HSBC named its first Chief AI Officer in 2026. No publicly named specialty finance lender or mid-tier BDC has appointed a comparable role yet.
That gap is the opportunity.
What a $40M specialty finance firm should actually do this year
Three concrete moves.
Map your document burden in dollars, not hours. For a $40M factoring shop, the question is not "how much time does our team spend on credit decisioning." The question is "what is the basis-point cost of our decisioning latency on our book." Most operators have never run that math. Once you do, the build-vs-buy question becomes obvious.
Pick one workflow with regulatory urgency. Onboarding compression for an RIA. Sub-$250K credit decisions for an equipment finance shop. Continuous credit monitoring for a BDC. Do not try to AI-enable everything. Pick the workflow where the SEC examiner or the OCC bulletin is already pushing you, and where you can show the regulator an audit trail at the end of year one. If you need help framing the vendor evaluation, our guide for operators without a CTO walks through the questions to ask.
Do not wait for a Harvey-for-finance to arrive. It will. By the time it raises a Series B at $700M, your competitors will already have shipped two production agents on top of it and you will be priced into year two of a multi-year platform contract you cannot unwind. The vendor layer is unconsolidated right now precisely because it is the right time to ship.
If you want help thinking through which workflow to start with, that is the conversation we have on most Granular discovery calls these days. We build fixed-price, four-week vertical AI deployments for mid-market specialty finance operators. Book 30 minutes and we can sketch the first one against your actual book. The first wave of $1.5B in compliance penalties and 71% of operators chasing the same documentation use case is not a market that stays empty for long.
FAQ
Does this apply to a $25M RIA or a $30M factor? Yes. The benchmark deployments below $100M in revenue are where the unit economics work best. Smaller firms have less in-house IT, less internal politics around AI, and a more direct line from "compress one workflow" to "see the margin in the next quarter."
What about banks? Banks are subject to OCC Bulletin 2025-26 and have larger compliance budgets, but they also have slower procurement, in-house data science teams that want to build, and regulatory baggage that adds 9 to 12 months to any deployment. The non-bank specialty finance segment is faster.
Is this a build or a buy? For a $30 to $100M operator, almost always neither, exactly. The right answer is a services-led deployment of vertical workflows on top of a thin platform layer. You do not want to own the model infrastructure, and you cannot afford to wait for a horizontal vendor to bend to your workflow. Our take on the build-vs-buy question is here.
How fast does an actual deployment ship? A focused workflow agent (onboarding compression, credit decisioning, document extraction) ships in four to eight weeks in our experience. The longer pole is the integration into your existing core system, not the AI work. We wrote about actual AI deployment timelines at $50M operators last week.
Keep Reading
- What Hyperscaler Capex Means for Mid-Market MEP: The parallel Money Flows piece on where mid-market AI capital is actually heading, and what hyperscaler buildout means for the contractors building the buildings.
- AI for Mid-Market Insurance: What Actually Works: The adjacent specialty vertical where the same operational logic applies, with three real-world AI use cases that already ship.