AI for Finance — The Market Radar 150 companies across 22 segments, every row carrying its sourced traction and a citation. Each segment is a layer — open one for where the gaps are. Sources and what this is not
2026-08 vintage 150 of 150 sourced
101
gaps identified— show all
The shape of the market Estimated segment size (USD)
Market size Companies Gaps
What this shows: how big each segment is, per the best published estimate we could source. AML monitoring is the largest at $4400B and 14667x the smallest sized segment (AI-native carriers). Scopes differ, so treat each bar as its own estimate.
AML monitoring $4400B 25
Fraud & disputes $579B 25
Claims & broker ops $48B 25
Tax compliance $25B 25
AI accounting & audit $23B 26
Treasury & recon $22B 26
Execution & trading agents $20B 26
Core banking $19B 26
KYC/KYB & identity $7.8B 26
Spend & AP/AR $6.9B 26
Wealth & planning platforms $6.8B 26
Loan ops & collections $6.4B 25
Credit decisioning $4.2B 26
Orchestration & checkout $3.1B 26
Sanctions & entity resolution $2.9B 26
Crypto compliance $2.9B 25
Underwriting AI $2.9B 24
Compliance supervision $2.8B 25
AI-native carriers $300M 25 Bar colour = how crowded the segment is saturated contested open
Bar height is compressed (square-root) because the sourced range spans roughly 14667× — on a straight scale most segments would be invisible. Read the printed figure, not the height. 3 of 22 segments carry no bar: no credible published figure exists, and none was invented. Segment market sizes are third-party analyst estimates, each measuring a DIFFERENT scope on a different definition — point estimates for context, not a comparable series, and they must not be summed. 19 of 22 segments carry one; where no credible published figure exists the segment carries none rather than a substitute.
Who is building what Every company in the corpus, stacked by segment
150 of 150 companies in scope · 22 of 22 segments shown · layers ordered and sized by company count — the corpus carries no market caps, so a box earns attention by its evidence: open one for the sourced traction.
S1 Research copilots saturated 10 4 gaps identified — open for the gaps, incumbents and entry risks. 9fin UK · AI-native credit platform for leveraged finance, private credit and distressed debt — extracts covenants, terms and financials from debt docs and monitors deals AlphaSense US · AI market-intelligence search over filings, broker research, expert-call transcripts and news, with generative summarisation and agents for analysts Boosted.ai Other · Alfa — an agentic AI analyst for portfolio managers doing screening, risk monitoring, quant workflows and client-commentary generation Brightwave US · AI research agent that produces sourced investment analyses over filings, news and a proprietary knowledge graph for hedge funds and private-credit investors Daloopa US · AI extraction of historical fundamentals from filings and presentations, delivered as audit-linked, model-ready data into analysts' Excel models Finster AI UK · AI-native research and task-automation platform for investment banks and asset managers — earnings updates, peer comps, industry deep dives from firm templates Fintool US · AI equity-research copilot over SEC filings and earnings transcripts; V5 generates DCF models in Excel, decks and memos autonomously Hebbia US · Matrix — an agentic document-analysis workspace that runs one question across thousands of filings, CIMs and diligence docs in a spreadsheet-style grid Rogo US · Agentic AI platform for investment banking and investing workflows — builds comps, screens, and drafts pitch materials from firm data and market data YipitData US · Alternative-data research on transactions, web and app panels, mapped to named public companies and sold to funds and corporates as analyst-ready KPIs + Gap No vendor sells the supervisory record for an AI-drafted research note — a replayable, per-claim trace a compliance officer can test months later when the note is challenged. Every copilot outputs citations; none outputs evidence that it behaved. + Gap Non-filing corpora are thinly served the copilots are anchored on SEC filings, transcripts and news, while private-credit and private-market diligence documents are handled per-deal in a workspace rather than as a persistent, queryable corpus. + Gap Everything here is priced per seat into a headcount the product exists to shrink No one has built engagement- or output-priced research — per deal, per initiation, per model built — which is the pricing a buyer would accept if the seat count falls. + Gap Coverage below the large-cap US line Daloopa and YipitData both normalise into the mega-cap analyst's workflow, and a sub-$500M fund covering non-US small and mid caps has neither the data feed nor a price point aimed at it. S2 Execution & trading agents open 8 5 gaps identified — open for the gaps, incumbents and entry risks. Almanak Asia · No-code platform where a swarm of agents researches, backtests, deploys and risk-manages DeFi strategies inside multi-sig and TEE-secured vaults BestEx Research US · Algo Management System — execution algorithms for equities, futures and FX with built-in TCA and backtesting, sold as SaaS rather than broker-bundled Giza EU · Agentic DeFi application layer; its ARMA agent automated stablecoin yield routing on Base KelAI US · Autonomous research and signal-development engine for hedge funds — generates, backtests, validates and monitors investment signals in one system LTX (Broadridge) US · AI-powered corporate-bond e-trading venue; BondGPT answers bond questions and, since 2026, runs agents that monitor conditions and raise trade tickets on the platform Lumenai Investments US · Hedge fund built the other way round — autonomous AI agents originate, evaluate and risk-manage global equity long/short positions, with humans retained for governance and risk supervision Theoriq US · Onchain infrastructure for deploying and coordinating autonomous DeFi agents, plus AI-managed liquidity vaults Trumid US · Electronic corporate-bond venue whose AI layer now executes: Full Self Trading turns a trader's intent into coordinated execution across RFQ and Swarms protocols, and Smart Voice turns chat/voice negotiation into pre-filled tickets + Gap No independent, benchmark-referenced attribution exists for agent-run strategies TVL is the published metric and it counts deposits; nobody publishes a verified, third-party track record for an autonomous vault. + Gap Best-execution evidence for a non-deterministic router TCA was designed for rule-based algos with a reproducible decision path, and there is no equivalent evidence file for an order an agent routed on judgement. + Gap Pre-trade mandate enforcement a compliance officer can attest to — hard limits on venue, instrument, size and notional per window, sitting outside the agent — now that agents raise live tickets rather than suggestions. + Gap The agentic step landed in corporate credit only Municipals, syndicated loans and repo share the dealer-quote, request-for-quote shape that made LTX work and have no equivalent. + Gap Orderly wind-down and recourse for a retired agent Giza returned funds by choice, and no vendor sells the unwind machinery, user notice and dispute path that a mandated retirement would require. S3 AI accounting & audit saturated 7 4 gaps identified — open for the gaps, incumbents and entry risks. Basis US · AI agents that run client accounting, tax and audit workflows inside accounting firms' existing systems Campfire US · AI-first ERP for mid-market finance teams with a proprietary accounting model for reconciliation and variance detection DataSnipper EU · Intelligent-automation layer inside Excel for audit tie-out, document extraction and testing, now shipped as AI agents Fieldguide US · Agentic AI platform for audit and advisory engagements — workpapers, testing, request management Numeric US · AI accounting-close platform whose cash-matching product reconciles bank transactions against general-ledger entries automatically Rillet US · AI-native ERP with a general ledger that ingests source systems directly and drafts the close Truewind US · AI bookkeeping and close automation that turns source documents into GL-ready journal entries, sold to accounting firms and startups + Gap Nobody sells the audit committee an assurance layer over the agents deterministic replay of why an entry was posted, drift monitoring on journal-entry classification, and change control on prompts and models at the standard a SOX control owner would accept. + Gap Financial data remediation has no product category Chart-of-accounts, vendor-master and subledger cleanup is the reason pilots stall, and it is currently bought by the hour from staff-augmentation firms. + Gap No outcome-priced audit model exists — per engagement, per opinion, per control tested — that would align a vendor with a partnership's economics instead of asking it to shrink its own revenue. Nor is anyone selling compression of the audit fee to the corporate buyer who actually pays it. + Gap There is no parallel-run bridge for an ERP cutover the AI-native ledger and the incumbent system producing the same close side by side for two quarters, with a daily reconciliation report the controller can hand the external auditor. S4 Tax compliance contested 6 4 gaps identified — open for the gaps, incumbents and entry risks. Black Ore US · Tax Autopilot — autonomous AI preparation and review of US returns (1040, 1041, 1065, K-1, K-3) for CPA firms, with extraction, review workflow and audit trail Filed US · AI tax preparer that plugs into existing practice software (CCH Axcess, Drake, Lacerte, UltraTax) to populate returns from source documents and run automated review Fonoa EU · Global indirect-tax platform — validation, determination, e-invoicing and returns on one data model across 190+ jurisdictions Juno US · CPA-founded AI layer between client source documents and tax-filing software, extracting data from IRS forms and unstructured financial documents for small and mid-size accounting firms Kintsugi US · AI sales-tax engine that reads billing/revenue systems and automates nexus tracking, registration, calculation and filing for e-commerce and SaaS sellers Numeral US · Automated US sales-tax compliance — nexus monitoring, registration, filing, remittance and exemption certificates for e-commerce and SaaS + Gap Mandate readiness is sold as a consulting engagement, not a product no vendor tells a group CFO 'these six entities are ninety days from a mandate they will miss, here is the per-jurisdiction gap list', which is the diagnostic that precedes every purchase. + Gap The sub-scale multinational is unserved — the company with three or four foreign registrations that cannot justify a 190-jurisdiction suite but is exposed to the same penalties as one that can. + Gap Audit defence has no artefact When an authority challenges a position, the evidence needed is the certificate chain plus the determination rationale for each line, assembled and preserved; nobody packages that as a durable record. + Gap An agent that determines a treatment or files a return produces no machine-checkable reason for its answer — the same assurance hole as the rest of the finance stack, but here it carries a statutory penalty and a named taxpayer. S5 Treasury & recon open 7 5 gaps identified — open for the gaps, incumbents and entry risks. Embat EU · Agentic treasury management for mid-market companies — bank reconciliation, payments, cash forecasting and an AI analyst ('TellMe') Ledge Asia · Payment-operations and close platform — connects banks, PSPs, ERPs and billing systems and runs continuous reconciliation plus AI agents that prepare workpapers and journal entries Modern Treasury US · Payment operations platform automating money movement end to end - initiation, approvals, bank connectivity and continuous reconciliation - via API and ledger Nilus US · Treasury and cash-management platform — multi-bank visibility, AI forecasting and payment ops Panax Asia · AI cash-flow and liquidity management for mid-market and large corporates — pulls bank and ERP data into a live cash position with forecasting and recommended actions Round UK · AI finance-automation platform spanning treasury, AP, payments, FX and payroll for European growth companies, with an agentic workflow builder Trovata US · Bank-API-native cash management and treasury platform with a generative-AI assistant over corporate cash data; now a full TMS after acquiring ATOM + Gap No vendor prices against the residual 1–10% of reconciliation exceptions, which is where the overnight staffing and the error-correction cost actually sit — everyone sells the 90–99% that is already automated. + Gap Corporate actions and settlement-fail exceptions are long-tail work with verifiable ground truth in the settlement record — ideal agent shape — but the fully loaded cost is spread across several cost centres, so no executive owns the saving and no vendor can name the budget it displaces. + Gap The mid-market treasurer sits between a spreadsheet and an enterprise suite with nothing designed for them, which is precisely the band Embat and Round are funded at single-digit-to-thirty-million scale to serve. + Gap Nobody sells exception clearing as an outcome — per exception resolved, with an evidence trail — even though the buyer's real alternative is a night shift rather than another dashboard. + Gap Multi-bank statement normalisation is a permanent undifferentiated tax every vendor pays privately; there is no neutral, testable connectivity layer the application vendors could build on instead of each rebuilding it. S6 Spend & AP/AR contested 6 5 gaps identified — open for the gaps, incumbents and entry risks. AppZen US · AI agents for T&E, accounts payable and corporate-card programs — 100% expense audit, invoice capture, GL coding and PO matching on its own ZenLM domain models Brex US · Corporate cards and expense/AP automation for startups and enterprises, repositioned around AI agents for receipts, approvals and vendor management HighRadius US · Order-to-cash, record-to-report, treasury and AP automation for large enterprises, now shipping a fleet of AI agents across those suites Payhawk UK · Spend-management and 'office of the CFO' platform with four production AI agents — financial controller, procurement, travel and payments — that execute finance workflows inside existing policy and approval rules Ramp US · Corporate cards, bill pay, procurement and expense with agents that code invoices, flag duplicate SaaS subscriptions and block out-of-policy spend at the swipe Tabs US · AI accounts-receivable platform that reads B2B contracts to drive invoicing, revenue recognition, cash application and collections + Gap Receivables is the unsubsidised half of the segment and Tabs is nearly alone in it — cash application, collections and invoice-dispute work remain heavily manual, and no interchange stream is available to undercut a software price there. + Gap Non-standard billing is where invoicing actually fails usage, milestone and revenue-share contracts whose terms have to be read before an invoice can be correct. That is document work, not payments work, and payment-funded platforms have no reason to specialise in it. + Gap Agents now block out-of-policy spend at the swipe but nobody sells the exception record — who overrode the block, on what basis, replayable at audit — which is the artefact a controller needs precisely because the enforcement is automatic. + Gap Buyers whose card programmes are constrained — regulated entities, public sector, businesses on procurement mandates — cannot be reached by an interchange-funded model at all, and are served by nobody AI-native. + Gap The Brex acquisition created a displaced-preference pool companies that chose an independent platform specifically to avoid a bank now sit inside one, and no vendor is selling to that motivation. S7 Credit decisioning contested 6 4 gaps identified — open for the gaps, incumbents and entry risks. Ocrolus US · Document and cash-flow analytics infrastructure that turns bank statements and paystubs into structured underwriting inputs Oscilar US · Agentic risk platform unifying fraud, credit underwriting, onboarding and AML decisioning on one data layer, with 30+ specialised AI agents RDC.AI (Rich Data Co) AU · Explainable AI decisioning for business and commercial lending — borrower-behaviour models banks can defend to regulators, sold into bank credit and risk teams Taktile US · No-code decision engine for onboarding, credit, fraud and compliance decision flows at lenders and banks Upstart US · AI credit underwriting marketplace — models price consumer, auto and home loans for bank and credit-union partners on far more variables than a FICO score Zest AI US · Machine-learning credit underwriting and fraud models for banks and credit unions, deployed into their existing loan origination systems + Gap The mainstream community-bank commercial loan — a $250k-2M owner-occupied CRE or equipment note underwritten off tax returns and a rent roll — has neither the standardisation that makes SBA 7(a) automatable nor the volume to justify a bespoke model, and no vendor here covers it + Gap Nobody sells the bank's second line an independent validation service for a purchased AI credit model — the model-risk review, challenger model, and disparate-impact testing that a supervisor will ask for, produced by someone who does not also sell the model + Gap Cash-flow underwriting exists for small-business funders (Ocrolus's density is with MCA and online lenders); the equivalent for thin-file consumer credit at a bank, with adverse-action reason codes derived from cash flow rather than bureau attributes, is unserved + Gap No vendor prices against post-deployment model drift a lender gets a scorecard at go-live and finds out it decayed from its own vintage curves months later, with no continuous monitoring product sold separately from the model vendor S8 Loan ops & collections contested 7 5 gaps identified — open for the gaps, incumbents and entry risks. Blend Labs US · Digital lending origination platform for banks and mortgage lenders; Autopilot is its AI agent that reviews borrower documents against guidelines, calculates GSE-qualified income and issues follow-ups mid-application Casca (Cascading AI) US · AI-native loan origination system for business lending — application intake, document collection, KYB and credit analysis InDebted AU · Digital-first collections agency using ML to choose contact channel, timing and treatment across the delinquency lifecycle interface.ai US · Agentic AI platform for credit unions and community banks — voice/chat service agents, extended in 2026 to early-stage delinquency outreach Prodigal US · AI agents and intelligence layer for loan servicing and collections — voice agent, call analytics, automated call documentation and agent assist for lenders and collection agencies Salient US · AI voice/text agents that handle consumer-loan servicing calls — payments, extensions, payoffs, insurance, disputes — with real-time compliance enforcement against FDCPA/TCPA/UDAAP rules TrueAccord US · Digital-first debt collection — the Heartbeat ML engine picks channel, timing and offer per consumer instead of call-centre dialling + Gap The pre-delinquency cure moment — identifying the borrower about to miss and offering a restructure before a 30-day mark reaches their bureau file — is worth more than recovery after the fact and is thinly served; InDebted bought Receeve to reach toward it rather than building the category + Gap The long tail of small creditors (medical practices utilities, SMB lenders, property managers) has no compliant agentic option, because FDCPA and TCPA exposure at their volume is uninsurable and no vendor packages the liability with the software + Gap Nobody sells the examiner-facing evidence layer as a product a replayable, per-call attestation that an autonomous agent stayed inside disclosure, consent, frequency and dispute-handling rules, produced independently of the vendor whose agent made the call + Gap Servicing transfer and boarding — reconciling loan tapes, payment histories and escrow between servicers — is high-volume, error-prone and entirely unautomated by anyone in this segment + Gap Commercial and equipment loan origination at community banks Casca's intake automation assumes a standard document set, and no vendor has built the messier tax-return-and-rent-roll intake most US small-business credit actually runs on S9 Core banking open 6 4 gaps identified — open for the gaps, incumbents and entry risks. 10x Banking UK · Cloud-native core banking platform with an AI suite (AI Migrate for legacy data mapping, AI Build for generative product configuration) aimed at replacing bank core systems Engine by Starling UK · Starling Bank's spun-out SaaS core banking platform — the bank's own stack sold to other banks, including its AI/automation tooling Maximum US · AI-native core banking operating system — real-time ledger plus a runtime for banks to build and deploy custom operational agents Posh AI US · Conversational and generative AI voice/digital assistants purpose-built for banks and credit unions — call deflection, self-service banking, and an internal knowledge assistant Thought Machine UK · Cloud-native core banking (Vault Core) and payments platform where product logic is written as code-level 'smart contracts' Tuum EU · Cloud-native, API-first core banking platform in four modules (core, payments, lending, cards) aimed at migrating mid-tier European banks off legacy cores in months + Gap Nobody sells the migration itself — a parallel-run harness that operates the new core alongside the old, reconciles both ledgers daily and produces the cutover evidence a regulator and an audit committee will accept. Every core vendor sells the destination; no one sells the bridge + Gap An agent-safe write layer over a legacy batch core idempotency, reversal, per-agent limits, dual control and a full action log, so a bank can let agents act on the core it already has without a replacement programme + Gap Real-time state for banks stuck on batch a core-agnostic balance, hold and posting API that gives agents and downstream apps intraday truth without touching the underlying system + Gap The examiner-ready evidence pack for agent actions in the system of record — what the agent saw, what it decided, under whose authority, replayable months later — sold independently of whoever supplied the agent S10 Orchestration & checkout contested 7 5 gaps identified — open for the gaps, incumbents and entry risks. Pagos US · Independent payments-intelligence layer — ingests transaction-level data across processors, normalises and enriches it (BIN data, network context) so merchants can lift authorisation rates and cut cost Payrails EU · Payment operating system for large enterprises — orchestration across acquirers, payouts, tokenisation, unified analytics and automated reconciliation Primer UK · Payment orchestration and workflow infrastructure unifying PSPs, fraud tools and payment logic behind one no-code layer, now pushing AI agents into payment operations Silverflow EU · Cloud-native card processing platform connecting acquirers directly to the card networks, with data-rich responses used to tune authorisation performance Solidgate UK · Payment processing and orchestration for subscription and digital businesses, bundling smart routing, retry logic, reconciliation and chargeback prevention Stripe US · Payment processor whose Payments Foundation Model - a transformer trained on billions of transactions - drives acceptance optimisation, card-testing defence and adaptive checkout Yuno Other · Payment orchestration layer letting merchants connect to many PSPs, acquirers and local payment methods through one integration, with routing and failover logic + Gap No vendor prices against measured authorisation lift — merchants buy routing intelligence on a per-transaction or platform fee while the claimed benefit is a counterfactual nobody holds them to; a held-out-split, success-fee model is constructible and absent + Gap There is no neutral, auditable authorisation-rate benchmark by geography, method and issuer that a merchant can use to compare vendor claims — every number in this segment comes from the party selling the lift + Gap Failed-subscription-payment recovery for the mid-market dunning, retry timing and card-updater coverage are bundled inside processors for large merchants and unserved as a standalone, verifiable product below that + Gap Nobody sells the issuer side of the same problem — a decline-reason feedback product that tells an issuer which of its declines were good customers, which is where the auth-rate ceiling actually sits + Gap Agent-initiated payments need their own authorisation heuristics and none of the orchestration layers price or route for them yet, even as merchant-side fraud vendors race to authenticate the same traffic S11 Fraud & disputes saturated 8 4 gaps identified — open for the gaps, incumbents and entry risks. Chargeflow US · Generative-AI chargeback automation that assembles representment evidence and files disputes across 100+ payment and commerce platforms, priced on success Featurespace (Visa) UK · Adaptive-behavioural-analytics transaction monitoring for card and account payments, sold to banks and processors; now part of Visa's Risk and Identity Solutions unit Forter US · Identity-graph fraud decisioning for enterprise e-commerce, now extended to authenticating AI shopping agents and orchestrating agentic-commerce protocols Justt Asia · Chargeback representment platform that tailors evidence per case, A/B tests dispute arguments, and scores whether a dispute is economically worth fighting Resistant AI Other · Document-forgery detection and transaction-monitoring models that harden other firms' automated financial decisioning against manipulated inputs Riskified Asia · Chargeback-guarantee fraud decisioning for large e-commerce merchants, plus policy-abuse, account-takeover and dispute-resolution products Sardine US · Behaviour-biometrics and device-intelligence risk platform issuing real-time decisions for fraud, AML compliance and credit underwriting Signifyd US · Guaranteed-outcome commerce protection: approves or declines orders and absorbs the chargeback liability on approved fraud, plus returns and policy-abuse cover + Gap Fraud defence from the buyer's side of an agentic purchase — the consumer or business whose agent is phished, over-charged or manipulated into an unwanted order — has no vendor at all, and card liability rules written for humans clicking buttons do not cleanly assign fault; whoever moves first also defines the dispute-reason taxonomy + Gap Issuer-side pre-dispute deflection a cardholder can file a 'did not recognise' claim in two taps with no evidence, and no one sells issuers a deflection product with the sophistication merchants now buy for representment, even though the merchant-side success-fee model shows a symmetric one is constructible + Gap Deepfake and injection-resistant identity proofing is claimed by every vendor on its own traffic and independently benchmarked by nobody; the open slot is a continuously refreshed adversarial evaluation run by a party that does not sell verification, since a static certification would decay with each new generative model + Gap Guarantee pricing is opaque to the merchant nothing lets a merchant compare a vendor's chargeback guarantee against its own realised fraud and false-decline cost on a like-for-like basis, so the highest-margin product in the segment is bought without a benchmark S12 AML monitoring contested 6 5 gaps identified — open for the gaps, incumbents and entry risks. Bretton AI (formerly Greenlite AI) US · AI agents that run KYC/KYB reviews, AML and sanctions investigations and transaction analysis inside banks' existing compliance workflows Flagright EU · AI operating system for financial-crime compliance — transaction monitoring, watchlist screening, risk scoring, case management and AI forensics with explainable, audit-ready outputs Hawk EU · Explainable-ML AML platform covering transaction monitoring, payment and customer screening, customer risk rating and fraud detection for banks and payment firms ThetaRay US · Cloud AI transaction-monitoring and screening for cross-border payments — unsupervised anomaly detection over correspondent-banking and fintech payment flows, plus entity-resolution screening Tookitaki Asia · FinCense AML and fraud platform plus the AFC Ecosystem — a federated-learning network where member institutions share typologies so detection models update across the community Unit21 US · No-code fraud and AML risk infrastructure for fintechs and their partner banks, rebuilt around agents that tune rules, investigate alerts and file regulatory reports + Gap No vendor can show detection lift on a held-out set of confirmed launderers, because SAR outcomes are rarely returned; a consortium or regulator-brokered feedback loop that makes that measurable is unbuilt and would reset how the whole segment is bought + Gap Alert volume is attacked at the wrong end — everyone processes alerts faster and nobody sells rules-layer tuning that stops generating them, which is where both the analyst cost and the supervisory criticism originate + Gap Entity resolution across transliterations, corporate ownership chains and shell layers is the actual failure mode behind screening false positives, and nobody sells it as a component that sits underneath an incumbent screening engine rather than as a platform replacement + Gap There is no examiner-facing evidence product for agentic decisions why an agent closed an alert, under what model version and what data, replayable a year later during an examination — the model-risk artefact that credit underwriting was forced to build and compliance has not + Gap The long tail — payment institutions, e-money firms and mid-size crypto businesses too small for a tier-1 platform but supervised as if they were not — has no product priced for it S13 KYC/KYB & identity contested 6 5 gaps identified — open for the gaps, incumbents and entry risks. Alloy US · Identity and risk decisioning orchestration for banks and fintechs - routes each applicant across many data vendors under configurable policy, plus perpetual KYC/KYB monitoring Baselayer US · AI KYB — real-time business verification, risk scoring and ongoing portfolio monitoring for sanctions, fraud and entity changes, sold to banks, fintechs and government agencies Incode Technologies US · AI biometric identity verification and authentication — document + face matching, liveness and deepfake/synthetic-identity defence for banks, telcos and government Persona US · Configurable identity-verification and KYB platform combining document, biometric and database signals, with deepfake and injection-attack detection Socure US · ML identity-verification and fraud platform predicting identity risk at account opening across consumer, business and government use cases Sumsub UK · Full-cycle verification platform — document/biometric IDV, KYB, deepfake detection, Travel Rule and transaction monitoring in one onboarding orchestration layer + Gap No neutral, continuously re-adversarialised benchmark for deepfake and injection resistance that a buyer can compare vendors on — every claim in this segment is self-reported on a non-comparable population, and a static certification would rot with each new generative model release + Gap Perpetual KYB on ownership chains sold as an unbundled subscription to firms that already bought onboarding elsewhere — re-verifying beneficial ownership as it changes, rather than re-running the onboarding snapshot + Gap Nobody prices against the manual review queue Verification is sold per check, while the cost sits in the applicants automation kicks out to a human — a per-cleared-exception price is unoccupied + Gap Identity proofing for agent-initiated onboarding an autonomous agent opening or operating an account on a person's behalf has no established proofing path in any of these products + Gap Public-sector identity proofing at procurement scale — the five-year $163M Login.gov identity-proofing order is the only government-shaped item on this map, and it went to Xcelerate Solutions as prime contractor with Socure subcontracting, so the seat in front of the federal buyer belongs to a systems integrator rather than to any vendor here. That buyer's evaluation criteria are also published rather than vendor-defined S14 Sanctions & entity resolution contested 7 4 gaps identified — open for the gaps, incumbents and entry risks. Castellum.AI US · Financial-crime compliance platform pairing in-house sanctions/PEP/adverse-media risk data (refreshed every 5 minutes) with AI agents that resolve AML/KYC screening alerts ComplyAdvantage UK · Financial-crime data and screening platform: a proprietary sanctions, PEP and adverse-media risk database plus customer and payment screening and monitoring Kharon US · Sanctions and geopolitical risk-intelligence data — resolves ownership networks behind designated parties to expose derivative sanctions, export-control, forced-labour and supply-chain exposure Napier AI UK · Cloud-native AML compliance platform (Napier Continuum) covering screening, transaction monitoring and risk assessment, with a sandbox for testing rules before deployment Quantexa UK · Entity-resolution and network-analytics platform that stitches fragmented records into connected entity graphs for financial-crime investigation, KYC and fraud Sayari US · Counterparty and supply-chain risk intelligence — Sayari Graph resolves corporate ownership, trade and shipment data across 250+ jurisdictions to expose hidden sanctions, forced-labour and export-control exposure Silent Eight Asia · AI that adjudicates name-screening and transaction-monitoring alerts, writing the resolution narrative so an analyst reviews a decision rather than raw hits + Gap Cross-script entity resolution as an unbundled layer that installs underneath any incumbent screening engine — Cyrillic, Arabic and Chinese transliteration plus corporate ownership chains — sold on precision rather than list coverage. No vendor in this set sells resolution on its own + Gap Nobody prices against the alert-clearing pool Screening is sold per name screened while the cost is per alert a human closes; a per-cleared-alert contract with a precision floor is unoccupied + Gap Designation-event re-screening as a product when a list changes, re-testing the existing book against the delta and producing the evidence file, rather than a periodic batch re-run + Gap The audit artefact for AI-closed alerts — what a model-risk function or examiner accepts as proof that an automated disposition was correct, which is the actual blocker on letting adjudication run unsupervised S15 Adviser copilots saturated 6 4 gaps identified — open for the gaps, incumbents and entry risks. FINNY US · AI prospecting and marketing platform for financial advisers — scores and matches high-intent prospects, then runs the outreach Jump US · AI operating system for financial advisers — meeting prep, notetaking, recaps, follow-ups, CRM updates and compliance records Powder US · Parses complex financial and estate documents (statements, trusts, tax returns) into structured, adviser-ready extracts, with hallucination detection on the extraction TIFIN AG US · AI asset-growth platform for wealth enterprises — prioritises prospects, spots referral and consolidation opportunities in an adviser's existing book, and scores retention risk Zeplyn US · AI meeting assistant for financial advisers — records client meetings, produces compliant notes, and pushes structured data into CRM and planning workflows Zocks US · Privacy-first AI assistant and notetaker for financial advisers that extracts structured client data from conversations and pushes it into CRM and planning tools + Gap Pricing against realised adviser outcomes — assets gained, clients retained, meetings converted — instead of per-seat time saved, which is precisely the value a custodian or CRM can zero out + Gap Executing what the meeting agreed rather than recording it the account changes, forms, transfers and beneficiary updates that still fall to an operations person after the recap is written + Gap Copilots built for the non-RIA distribution channels — insurance-licensed producers and bank-based advisers — whose product set, supervision regime and required records differ from the RIA shape every row here assumes + Gap Data portability advisers are accumulating years of structured client detail inside one vendor's CRM writes with no export standard, and nothing exists to move a book between copilots S16 Wealth & planning platforms contested 6 5 gaps identified — open for the gaps, incumbents and entry risks. Altruist US · Vertically integrated custodian and wealth platform for independent RIAs — custody, clearing, model portfolios, tax management and fixed-income trading in one stack, positioned as the AI-forward custodian Conquest Planning Other · Financial planning software with an AI Strategic Advice Manager that generates and ranks planning strategies for advisers and institutions Farther US · AI-native RIA — an all-in-one adviser platform (asset location, execution, risk, planning insights) operated as the wealth manager rather than sold to one Robinhood (Cortex) US · AI research assistant inside a retail brokerage — market analysis, custom indicators and real-time news synthesis, sold as part of the Gold subscription Savvy Wealth US · AI-augmented RIA — Savvy Intelligence unifies client data, meeting capture, planning and back-office work so advisers who join the firm run on one AI-native stack Wealth.com US · Estate and tax planning platform for advisers; Ester Intelligence reads estate documents and runs deterministic distribution calculations advisers can check + Gap Plan generation sold as an API rather than a seat the AI-native RIAs and adviser copilots now need strategy modelling underneath them, and the engines here are packaged for enterprise seat licences + Gap No independent evaluation of machine-generated advice quality — a strategy-ranking engine's output is assessed only by vendors that sell one, so a buyer has no way to compare two engines' recommendations on the same household + Gap Decumulation is the hardest planning job — tax-aware withdrawal sequencing, longevity and long-term-care risk — and no row here leads with it + Gap Households below the threshold that justifies a recruited human adviser Farther's model scales by recruiting advisers with existing books, which structurally skips the mass-affluent tier + Gap Jurisdictional variants of a planning engine Conquest's density is Canadian, and re-expressing tax and product rules per market is unbuilt work rather than a translation S17 Compliance supervision open 6 4 gaps identified — open for the gaps, incumbents and entry risks. Behavox UK · AI conduct surveillance — Quantum for communications monitoring and Polaris for trade surveillance across nine asset classes, pulling chats, email, voice and trade records into one case with evidence lineage Greenboard US · AI-native compliance operating system for SEC/FINRA-regulated firms — comms archiving with AI supervision, employee compliance monitoring, marketing review, plus GreenboardGo, a conversational layer that answers policy questions Hadrius US · AI-native compliance system of record for RIAs, broker-dealers, private funds and compliance consultants — six agentic modules covering marketing review, comms surveillance, trade and personnel monitoring Luthor US · Marketing-compliance automation for RIAs and broker-dealers — pre-publication review of marketing content, plus surveillance, disclosures and filings against SEC/FINRA rules Norm Ai US · Turns regulations into machine-executable AI agents that review content and workflows in real time, and supervise other AI agents deployed inside regulated firms Saifr US · Fidelity Labs-incubated regtech applying AI to marketing/advertising compliance review, adverse-media monitoring and electronic-communications surveillance for advisers, broker-dealers and insurers + Gap No supervision product priced for the small and mid-sized RIA that just bought a copilot — the gap is not capability but a price point between an institutional platform and a seed-stage tool + Gap Review at the point of generation rather than in a separate console an evidence trail produced inside the adviser copilot as the client email or recap is drafted, which is where the content actually originates + Gap The examiner-facing attestation for the AI tools themselves — how each model was tested, versioned, constrained and monitored — which firms now need and no vendor in this set sells as its primary product + Gap Marketing-rule review (testimonials, performance presentation, disclosures) for firms with no compliance department, priced below the consultant currently doing it by hand S18 Underwriting AI contested 6 5 gaps identified — open for the gaps, incumbents and entry risks. Akur8 EU · Machine-learning actuarial platform for non-life insurance pricing and reserving — transparent GLM/ML rate models actuaries can justify to regulators, plus optimisation and rating-engine deployment modules Cytora UK · Risk-digitisation platform that turns unstructured broker submissions (PDFs, emails, ACORD forms, images) into structured, decision-ready risk data for commercial underwriting Federato US · RiskOps underwriting platform — submission triage, appetite matching and portfolio-aware risk selection for carriers and MGAs FurtherAI US · Insurance-specific AI workspace with pre-built agents for submission intake, loss runs, SOV intake, policy checking, claims intake and underwriting audits hyperexponential UK · hx Renew — pricing decision-intelligence platform for commercial and specialty insurers, letting actuaries build and ship rating models over large and alternative datasets Sixfold US · AI underwriting platform — agents that read submissions against carrier appetite and guidelines inside existing workbenches and policy admin systems + Gap Independent measurement of whether AI-selected risks bind at a better or worse loss ratio than human-selected ones — the only metric that settles the category, and one no vendor can credibly grade itself on + Gap The regulator-facing evidence file for a declined or repriced risk which features drove the decision, how the model was tested for bias, what changed between versions. Explainability is named as an adoption barrier by carriers while decisioning ships anyway + Gap Life and health underwriting — accelerated underwriting against medical evidence — has no dedicated AI-native player in this set; every funded row here is commercial or P&C shaped + Gap Reinsurer-side tooling the party carrying the tail of straight-through binding has nothing that audits a cedent's automated selection before the treaty is written + Gap Contracts priced on the quality of bound premium rather than per submission cleared, which would align the vendor with the carrier's actual objective instead of its queue S19 Claims & broker ops contested 6 5 gaps identified — open for the gaps, incumbents and entry risks. Avallon US · AI agents for claims operations — intake, document summarisation, status tracking and exposure analysis, wired into claims management platforms, IVR and data warehouses Bevaya (formerly Roots Automation) US · Insurance 'digital coworkers' built on InsurGPT — reads submissions, legal demands and medical records and drives claims, underwriting and policy-servicing workflows EvolutionIQ US · AI guidance for disability and workers'-compensation claims — reads medical records and claim history to tell adjusters the next best action on each claim Five Sigma US · AI-native claims management platform and Clive, a multi-agent claims expert that runs intake, triage, coverage, liability, fraud checks and settlement on top of existing claims systems Fulcrum US · AI agents for insurance brokerages — coverage and claim analysis, proposal generation, policy checking and certificate issuance Liberate US · Voice AI agents for insurance back-office work — inbound and outbound calls for servicing, FNOL and sales, integrated with carrier systems + Gap No vendor prices against the residual ambiguous-coverage files — the minority of claims where adjuster hours and leakage actually concentrate. Every funded entrant sells the automatable majority and hands the hard tail back untouched. + Gap Nobody sells the loss consequence of automation there is no independent measurement telling a carrier whether AI-triaged claims paid out more or less than a matched human-handled cohort. The entrants all report throughput (automations per month, hours saved, days to delivery) because that is what a pilot can measure. + Gap Broker-side commission and premium reconciliation — matching carrier statements against agency management records — is a monthly manual grind at every brokerage and sits outside the policy-checking and certificate-issuance wedge the funded entrant took. + Gap Subrogation and recovery identification is an outbound job with a directly measurable dollar return, yet the funded companies are all on the intake and servicing side of the file. + Gap No product exists on the claimant's side of the same conversation Status tracking is built for the carrier's convenience; the policyholder chasing that status has no vendor at all. S20 AI-native carriers open 7 5 gaps identified — open for the gaps, incumbents and entry risks. Armilla AI Other · MGA and Lloyd's of London coverholder writing standalone AI liability insurance and AI performance warranties — cover for model error, harmful outputs, agent failures, data leakage and AI-regulation defence costs Corgi Insurance US · AI-native full-stack insurance carrier underwriting startups, with AI running underwriting, claims and policy operations Counterpart US · AI-native MGA and specialty carrier for small-business management-liability and AI-era exposures — agentic broker service, underwriting, risk mitigation and claims in one system Florin US · Full-stack insurance carrier with no human underwriters — a model reads ACORD forms, enriches risk from satellite imagery, permits and public records, simulates policy-year outcomes and returns a bindable quote, aimed at robotics, space, advanced energy and defence risks Klaimee UK · Insurance warranties for AI agents — cover attached to agent performance and failure Ominimo Other · AI-native motor insurer — prices individual driver risk with gradient-boosted models (XGBoost) rather than traditional rating tables, distributed with partner carriers across CEE Vouch US · Insurance built for startups and high-growth technology companies — brokerage plus its own MGA (Corix), with instant quoting across 80+ carriers and coverage extended to AI-specific risks + Gap No affirmative cover for autonomous financial execution — the loss an agent causes through its own transacting decisions, as opposed to third-party harm from a model's output. The existing standalone policy is shaped around model failure and harmful outputs; a bank or fintech deploying financial agents still holds that tail uninsured, which is a named reason deployments stall at pilot. + Gap There is no shared AI incident and loss record which is the input every underwriter needs before capacity widens The whole segment is selling policies; nobody is selling the data layer that makes policies underwritable. + Gap Reinsurers and fronting carriers backing an AI-native book have no independent way to assess the model doing the selecting. That is a third-party assurance job with a well-capitalised, clearly named buyer and no vendor attached to it. + Gap Life and health is untouched by the AI-native risk-takers here — the named books are commercial and specialty P&C (startups, robotics, space, advanced energy, defence). Accelerated life underwriting has a well-defined job, long cycle times and medical evidence to reason over, and no AI-native carrier in this scan. + Gap Agent performance warranties are being sold as insurance, but nobody sells the measurement that would trigger one an independent conformance test that determines whether the agent actually breached its mandate. Without it, every claim is adjudicated on argument. S21 Crypto compliance saturated 7 5 gaps identified — open for the gaps, incumbents and entry risks. Blockaid US · On-chain security and threat-detection layer — simulates and scores transactions and dApp connections in real time inside wallets and exchanges to block scams, drainers and malicious contracts Chainalysis US · Blockchain data platform for compliance, investigations and market intelligence, sold to exchanges, banks and government agencies Crystal Intelligence EU · Blockchain analytics for AML compliance and investigations — transaction risk monitoring, fraud detection and law-enforcement tracing across chains, plus the co-developed Scam Alert address database Elliptic UK · Onchain analytics and digital-asset risk decisioning for banks, payment firms and exchanges, with an agentic AI copilot that triages compliance alerts Merkle Science Asia · Predictive crypto risk and threat-detection platform for exchanges, banks and regulators, positioned on behavioural prediction rather than post-hoc tracing Solidus Labs US · Crypto-native market-integrity monitoring — HALO runs trade surveillance, transaction monitoring and threat intelligence across centralised and DeFi venues using market-abuse typologies TRM Labs US · Blockchain intelligence platform tracing transactions across chains for AML, sanctions screening and law-enforcement investigations + Gap Risk scoring for the controlling agent rather than the wallet Per-session agent wallets created on demand destroy address-as-actor clustering, and no incumbent can rebuild the unit of analysis without devaluing the entity label graph its law-enforcement credibility rests on. + Gap Nobody sells recourse to the wrongly de-risked customer The person or business frozen, exited or refused after a screening hit is not the buyer, and a vendor paid by banks cannot credibly argue against a bank's decision — leaving evidence assembly, a source-of-funds narrative and a structured appeal as an unserved job with a different payer. + Gap Detection lift is asserted and never independently measured Every vendor now ships an investigation copilot; none publishes a benchmarked false-negative rate that a bank's model-risk function could accept as effective challenge, which is exactly what the second line will start demanding. + Gap Entity resolution across scripts, transliterations and aliases is the actual sanctions failure mode, while the category still sells list and chain coverage. The buyer's pain is a match that should have fired and did not, not a data source that was missing. + Gap There is no examiner-ready evidence pack for the AI triage itself — banks are now running non-deterministic vendor agents over regulated alert queues with no supervisory method to validate them, and the compliance vendors supplying the agents are not the ones supplying the validation. S22 Agent payment rails contested 9 5 gaps identified — open for the gaps, incumbents and entry risks. Catena Labs US · AI-native bank for agent transactions, pairing stablecoins with an onchain identity system so an agent can be an authorised counterparty Coinbase (Agentic Wallets / x402) US · MPC agent wallets with session caps and per-transaction limits, plus x402, the HTTP-402 pay-per-call payment protocol for agents Crossmint US · Wallet, stablecoin and payment API infrastructure that enterprises and AI agents use to transact onchain without handling keys Halliday US · Agentic Workflow Protocol putting programmable, immutable guardrails around what an AI agent is permitted to execute onchain Mastercard US · Agent Pay for Machines — network-level authorisation and settlement for machine-initiated payments, extending card-network rules to agents Natural US · Payments orchestration for agent-to-agent commerce, including authorisation and dispute handling, built by ex-Stripe and Square staff Nava US · Trust and guardrail layer keeping AI financial agents inside their mandate — the control side of agent payments rather than the rail Payman AI US · Developer-first payment API that gives an AI agent a spending account with human-approval controls Skyfire US · Payment and identity network letting AI agents pay for APIs and services autonomously with a verified agent identity + Gap Conditional-release escrow as the structural substitute for a chargeback — funds locked at authorisation and released only against a verified delivery attestation. Stablecoin settlement is irreversible and the stack reached real transaction counts with no remedy behind it; exactly one funded company here names disputes as a product, and nobody sells escrow as a standalone layer any rail can call. + Gap A portable agent counterparty record — transaction count, dispute rate, mandate breaches, revocations — that a merchant, lender or insurer can query before extending trust. The identity work in this segment answers who an agent is; nothing answers whether it has behaved, and no rail operator can credibly run that register for its competitors' agents. + Gap The buyer's side of agentic fraud is unbuilt Every product here defends the merchant, the enterprise or the network; the consumer whose agent was manipulated into a purchase has neither a defence nor a counterparty to appeal to. + Gap Mandate conformance evidence for the deploying enterprise's second line — a per-decision reconstruction of why an agent spent what it spent, in a form a supervisor or auditor accepts. The guardrail vendors enforce at runtime and stop there; the artefact that gets a pilot approved for production is a separate product with a separate buyer. + Gap Cross-rail reconciliation of machine spend An enterprise running agents across x402, card rails and stablecoin wallets has no single ledger, and treasury cannot close a book it cannot see — the same exception-handling shape that goes unowned everywhere else in payment operations. Point-in-time snapshots, not market censuses. Absence from a radar is not evidence a company does not exist, and funding and traction age fast — re-verify a row against its source before it informs a decision. Not investment, commercial or legal advice.
