The investigation was the boring part. Now it's the product.
Socure just spent $156 million and one acquisition proving that the fraud investigation itself, not the alert, is where financial fraud detection software now competes. Here is what that means for how fintechs sell trust.
What Socure and Summit Partners actually announced#
Two years of fraud detection software marketing have chased the same headline, in folkfox's read: catch it faster, catch it earlier, catch more of it with one more model bolted on. Socure built a decade of that headline into a genuine business, and on 27 August 2026 it changed the sentence entirely. The company announced a $156 million strategic growth investment at a $5.2 billion valuation, and in the same breath said it had bought Fravity, an agentic operations platform that does the part of fraud detection software nobody ever puts on a landing page: the investigation itself, Socure, August 2026.
Summit Partners led the round, joined by Goldman Sachs Alternatives, Wells Fargo and Docusign among other backers, structured as primary capital plus a secondary tender offer that lets existing employees sell shares rather than wait years for an exit. Summit managing director Andy Collins framed the bet plainly: "We see identity as the first perimeter for trust in an AI-driven economy across nearly every use case," while colleague Matt Hamilton pointed to "the combination of durable growth and disciplined execution at this scale," Summit Partners, August 2026. The valuation has climbed briskly, from $4.5 billion at the 2021 Series E to $5.2 billion now, on more than $742 million raised since 2012, Crunchbase News, August 2026.
Fravity is the part fintech marketers should actually study. Founded in Austin in 2024 by Kedar Samant, who previously sold fraud-scoring firm Simility to PayPal, and Rushik Upadhyay, formerly a PayPal compliance architect, Fravity built more than seventy pre-trained AI agents that carry out the case-building work a human analyst used to do by hand: pulling documents, running sanctions and adverse-media screens, drafting the summary an investigator finally signs.
Socure is folding the whole platform into RiskOS, its orchestration and decisioning layer, under the name RiskOS_Agents, Biometric Update, August 2026. RiskOS already serves more than 3,000 customers across 190-plus countries, including nineteen of America's twenty largest banks, so the agents inherit a den built over a decade rather than a fresh, unproven trail.
The growth behind that price tag is real, not narrative. Socure closed the second quarter of 2026 with $364 million in annual recurring revenue, up 63 per cent on the year before, alongside 133 per cent net dollar retention and a churn rate of just 0.01 per cent, figures few fraud prevention software vendors at this scale can quote without flinching, Summit Partners, August 2026. Ninety-five new customers signed in the quarter alone, among them Circle, Cox Automotive, MoneyLion and Login.gov, Crunchbase News, August 2026.
None of that explains why an identity and fraud detection software company spent real money buying an investigation platform rather than another detection model. The answer sits one layer downstream of the alert, in the part of the workflow every fraud prevention software buyer complains about and almost nobody sells a fix for.
Why fraud detection software vendors are buying the investigation now#
Detecting a suspicious transaction and deciding what to do about it are two different jobs, and financial fraud detection software has spent most of its history solving only the first one. An alert fires, a rule or a model flags it, and a human analyst opens a dozen browser tabs, pulls a sanctions list, checks an adverse-media database, reads a KYC file and writes up what she found. Fravity's agents run KYC investigations, business due diligence, transaction-monitoring alert assessment, sanctions and politically-exposed-person screening and adverse-media research, the fraud detection tools stack's most expensive, least automatable stretch, PYMNTS, August 2026.
Those figures, and the 80 per cent cut in cost per case and up to 70 per cent fewer false positives Socure quotes alongside them, are the company's own reported deployment numbers, not an independent audit, Biometric Update, August 2026. Any financial fraud detection software vendor citing similar multiples should say so plainly rather than let one case study read as a guarantee for every customer.
The prize behind the shift is sized in the tens of billions, not millions. Crunchbase News puts the addressable financial crime investigation market at $71.1 billion, a bigger arena than the alert-scoring market fraud detection software has fought over for years, and one where fraud detection tools alone were never going to be the whole answer, Crunchbase News, August 2026.
The alert pile every fraud prevention software team already knows#
None of this happened in a vacuum. Fifty-three per cent of banks spend at least an hour reviewing a single fraud alert, and 37 per cent manually review at least 40 per cent of their compliance queue by hand, a Liminal Strategy finding SiliconANGLE folded into its own coverage of the deal, SiliconANGLE, August 2026. Every fraud prevention software team recognises that pile, because it is the reason the queue never actually empties.
US organisations spend roughly $100 billion a year on fraud, compliance and risk operations combined, and 57 per cent of businesses report fraud attempts rising over the past twelve months, Summit Partners, August 2026; Biometric Update, August 2026. Socure itself claims an 8,000 per cent rise in AI-driven fraud attempts across its own network over the same window, a number worth reading as a vendor's own telemetry rather than an independently audited count, Crunchbase News, August 2026.
What the academic literature says about agentic case-building#
Academic work on the same problem agrees with the shape of the complaint even where it distrusts the marketing gloss. A study built through direct fieldwork with a fintech firm and its regulators found that "the cost and complexity of financial crime compliance continue to rise, often without measurable improvements in effectiveness," and argued that an agentic system earns trust only through explainability, traceability and clear role boundaries between what the software decides and what a person still must, Axelsen, Licht and Damsgaard, HICSS-59, 2026. That is the quiet, disciplined case for the investigation-as-feature; it is also the standard fraud detection tools marketing keeps failing to meet.
For a fox reading the wind before it crosses open ground, the lesson is the same one that study makes formally: measure before you market, and never claim more scent than the trail actually carries, a warning folkfox has sounded before in AI fraud detection is now a brand problem, not just a bank's, when regulators started treating AI-driven scam claims as a disclosure issue rather than a technical footnote.
The regulatory ground fintechs are marketing on#
Selling an agentic investigation feature means selling into a regulatory mood shifting under three separate agencies at once, and a fintech marketing team that ignores any one of them is building a claim on sand. FinCEN's April 2026 proposed rule is the biggest mover: it reframes anti-money-laundering and fraud-prevention programme requirements around measurable effectiveness rather than paperwork volume, distinguishing a programme's design from its actual implementation, with public comment closing 9 June 2026, FinCEN, April 2026. It is the same effectiveness standard already reshaping sponsor-bank compliance risk elsewhere in fintech.
That shift rewards exactly the kind of measurable outcome Fravity is selling: cost per case, false-positive rate, time to resolution, provided a fintech can actually show its working. The Office of the Comptroller of the Currency has said as much since 2018: private-sector innovation, including new technology, "can help identify suspicious activity and combat money laundering," but regulators "will not penalize banks that maintain effective anti-money laundering programs but choose not to pursue innovative approaches," OCC, BSA/AML Innovative Industry Approaches. Adoption is encouraged, not compelled, and a marketing claim implying otherwise is already overreaching.
The third agency is the one that punishes the overreach. In March 2024 the SEC fined two investment advisers a combined $400,000 for what it now calls "AI washing," false or exaggerated claims about AI capability the firms did not actually have. SEC chair Gary Gensler's warning reads like it was written for this exact moment: "Investment advisers should not mislead the public by saying they are using an AI model when they are not," SEC, March 2024. Fraud detection software marketing sits inside financial services precisely because it claims to protect financial services, which makes it a plausible next target for the same enforcement logic.
Vague, all-seeing, unliftable
Our AI agents catch every case of fraud before it happens, with zero false positives and no human involved.
Specific, sourced, human-anchored
Our agents draft the case file a licensed analyst already builds by hand, cite every source they touch, and a named human signs off before any account is closed.
The second version is duller to read and considerably harder for a regulator, or a rival's lawyer, to unpick. That is the whole point.
How to talk about agentic fraud detection tools without overclaiming#
None of this argues against marketing the capability. It argues for marketing it the way Socure itself did: specific numbers, a named deployment, a clear boundary between what the software finished and what a person still decided. A fintech selling fraud prevention software built on someone else's agentic layer, or its own, has a narrower set of safe claims than the vendor slide deck usually contains.
| What marketing wants to say | Where it breaks | A safer framing |
|---|---|---|
| "Our AI stops fraud" | Reads as a guarantee no regulator or opposing counsel will accept | "Our agents flag and document suspicious cases for a licensed analyst to decide" |
| "Fully automated compliance" | Implies no human review, the exact shape of an AI-washing claim | "Agentic case-building, human sign-off on every action" |
| "Zero false positives" | Unverifiable, and disproved by a single counter-example | "Up to 70% fewer false positives, per [named source and deployment]" |
Notice what survives: a number with a source attached, a named boundary between the agent and the analyst, and a claim scoped to one deployment rather than the whole species of fraud. Fraud detection software marketing that keeps those three things intact will still read as confident. It will just also be true, which is the only version of confident that ages well. If your brand strategy still leans on vague trust language, this is the moment to rewrite it into something a regulator could quote back at you; folkfox's content marketing and fintech marketing teams build exactly that kind of copy for regulated brands.
What buyers of financial fraud detection software will ask next#
The sharpest reaction to the deal did not come from a bank. Amit Spitzer, chief technology officer at Glilot Capital, posted the read that mattered within hours of the announcement.
The Fravity pickup is the interesting line item. Socure built its name on fraud and identity signal, buying an agentic GRC startup means they now want to own the compliance evidence trail too. That's a different buyer inside the customer, not just a bigger product.
"That's a different buyer inside the customer, not just a bigger product," @amitspofficial, 1 September 2026, and it reframes the marketing problem more than the product one. A fraud detection tools buyer inside the fraud team wants a lower false-positive rate. A compliance buyer inside the same building wants an audit trail she can hand to an examiner without editing it first.
Q2 2026 ARR
Up 63% year on year.
Net dollar retention
Existing customers are spending more, not just staying.
Logo churn
Close to the floor a subscription business can reach.
Socure's own retention numbers, 133 per cent net dollar retention against 0.01 per cent logo churn, suggest existing customers are already buying both the detection story and the compliance one. A fintech pitching a similar feature to a market that has not signed yet needs proof for each buyer separately, not one slide trying to serve both.
There are two types of companies that matter in the AI-driven global economy: those that are AI-native, and those that fight the consequences of AI acceleration.
The honest read here, in folkfox's book, is simple: the fox that survives the thicket is not the one that claims to see every rabbit, it is the one that can show exactly which trail it followed to the one it actually caught. Fraud detection software that can do that, and say so plainly, is the version fintech marketing should be building toward for the rest of 2026.
Frequently asked questions#
What tools are commonly used for fraud detection?
Most fraud detection tools fall into a few families: identity verification and device fingerprinting, transaction-monitoring rules and machine-learning models, sanctions and politically-exposed-person screening, and increasingly agentic AI layers like Fravity that carry out the investigation an analyst used to do by hand. Rules-based fraud detection tools rarely disappear; they usually sit underneath the newer agentic layer rather than being replaced by it.
What is fraud detection software?
Fraud detection software is any system that flags a transaction, application or account as suspicious using rules, statistical models or machine learning. It answers one narrow question, is this worth a closer look, and hands that closer look to a human analyst or, increasingly, to an agentic AI layer built specifically for investigation rather than detection.
How is financial fraud detection software different from fraud investigation software?
Financial fraud detection software flags the alert; fraud investigation software, the newer category Fravity built, gathers the evidence, runs the sanctions and adverse-media checks and drafts the case file a human analyst finally signs. Socure's bet is that buyers increasingly want both bought together under one platform, rather than stitched from separate vendors.
Is agentic AI fraud investigation software regulated the same way as fraud detection tools?
Not yet, specifically. FinCEN's April 2026 proposed rule and the OCC's long-standing guidance both judge a compliance programme by its measurable effectiveness rather than the specific technology behind it, and neither mandates AI. The SEC has already fined firms for overstating AI capability, so any vendor's effectiveness claims need a named, sourced deployment behind them.
What is fraud prevention software, and how does it differ from fraud detection tools?
Fraud prevention software tries to stop fraud before it happens, through identity verification, authentication and access controls. Fraud detection tools catch it after an attempt is already under way, by scoring transactions or behaviour as suspicious. Most fintechs run both, and the newest layer, agentic investigation, sits between the two, turning a detection alert into a documented decision.
How much did Socure raise, and what does the Fravity deal actually change?
Socure raised $156 million in a strategic growth round led by Summit Partners at a $5.2 billion valuation, and used the same announcement to reveal it had acquired Fravity, an agentic AI investigation platform, folding it into its RiskOS decisioning system as RiskOS_Agents. The practical change is that the investigation step, not just the alert, is now something Socure sells as a feature.
Read more on this topic#
AI fraud detection is now a brand problem, not just a bank's
ASIC removed 19,400 AI-driven scam sites in FY26, up 182 per cent. What brand and compliance teams need to know.
Read the pieceEmbedded Finance and the Sponsor Bank Left Holding the Risk
Sponsor banks now face formal enforcement risk that fintech marketers cannot afford to ignore.
Read the pieceFintech AML requirements just got a €2.6 million reminder
Five honest checks before your neobank's growth story outruns its compliance file.
Read the pieceTwenty per cent of the world's money. Nought point two of its stablecoins
Revolut's euro stablecoin bet is about distribution, not issuance, and the same lesson applies here.
Read the pieceReady to market the investigation, not just the alert?
Regulated fintechs need the brand strategy and content marketing folkfox builds to talk about agentic fraud detection software honestly, with claims that survive an examiner's second look.