Mercury's AI Customer Research Strategy: How the $3B Startup Bank Onboards 200,000 Founders With Conversation

TL;DR#

Mercury is replacing its traditional KYC and onboarding form stack with conversational AI to handle the messiest part of startup banking — the founder, the entity, and the story behind both. Founded in 2017 by Immad Akhund and valued near $3 billion, Mercury serves more than 200,000 startups and small businesses and processes over $156 billion in annual transaction volume, per the company's 2024 disclosures. Its newer consumer arm, Mercury Personal, extends the same conversational pattern to the personal accounts founders, operators, and remote workers actually use. The core insight: when your customer is a Delaware C-corp three weeks old, a 50-state LLC with five co-founders, or a YC-backed entity whose product changed last week, a static onboarding form is a worse compliance instrument than a structured conversation. Mercury's bet is that AI onboarding tools — applied to KYC narrative capture, edge-case escalation, and post-activation discovery — are the only way to onboard founders at startup velocity without breaking BSA/AML triage. For other fintechs and operators, Mercury's playbook is a preview of what AI customer research will look like across the $2 trillion global business banking market.

What Is Mercury Doing With AI in Customer Onboarding?#

Mercury is using AI onboarding tools to convert its KYC, entity-structure, and use-of-funds capture from a static form sequence into a conversational interview that adapts to whatever the founder actually tells it. Instead of asking a Delaware C-corp founder the same 27 questions that a multi-member LLC sees, the application uses prior answers — entity type, formation state, ownership structure, expected transaction profile — to dynamically branch what it asks next, escalate edge cases to human reviewers, and capture the unstructured "why" that compliance teams need but legacy forms throw away. This is Mercury's response to a known industry problem: business bank onboarding is the single highest-friction moment in fintech, and the form has been the bottleneck for two decades.

Two product surfaces tell the story. Mercury IO, the core business banking product, handles the formal application — entity verification, beneficial-owner KYC, expected activity profile. Mercury Personal, launched in 2024 and broadly rolled out in 2025, extends the conversational model into consumer accounts for founders and operators who'd otherwise be stranded on Chase. The same conversational onboarding pattern — adaptive questions, narrative capture, AI-first triage with human escalation — anchors both. It is the same model we cover in our breakdown of Stripe's onboarding philosophy and the Ramp customer onboarding strategy, and it shares structural DNA with how Brex listens to founders at scale.

Why Startup KYC Breaks Traditional Banking Onboarding Forms#

Startup KYC breaks traditional banking onboarding forms because the entity, the founder, and the use of funds are all simultaneously novel, ambiguous, and high-velocity — three properties that static schemas are explicitly designed not to handle. Legacy bank onboarding forms assume a stable counterparty: a registered business, a clear NAICS code, a predictable transaction pattern, an owner with five years of history. Startups violate every one of those assumptions, often within the same application.

Three structural failures dominate:

Entity complexity. A "startup" might be a one-week-old Delaware C-corp with one founder, no revenue, and a SAFE in flight. It might be a Cayman holdco above a U.S. opco. It might be five co-founders across three countries on a single cap table. Static forms collapse all of that into the same six fields. The result is either an under-collected file (which a BSA officer rejects on review) or an over-collected file (which a founder abandons before submitting). Both are failures.

Regulatory ambiguity. Banks are governed by the Bank Secrecy Act and the FinCEN Customer Due Diligence rule, which require institutions to understand the "nature and purpose" of every customer relationship and to file Suspicious Activity Reports when transactions diverge from that purpose. A form captures categories; it does not capture nature and purpose. AI onboarding tools that can ask "tell me how your customers pay you, and what your three biggest months last year looked like" capture exactly the narrative compliance teams need — and the answer can be re-queried against transaction reality later. Our conversational intake guide walks through this same triage pattern in adjacent industries.

Founder velocity. Founders are the world's most form-fatigued demographic. The SaaS-onboarding research is clear: each additional form field reduces completion, and account opening sits at the worst possible spot in the funnel — high-stakes, high-friction, pre-value. The 2026 AI Customer Onboarding benchmark and the broader AI customer onboarding adoption report both show conversational onboarding flows outperforming static forms on activation by double-digit percentages. For Mercury, where the application IS the activation, that gap is existential. It's the same dynamic Notion solved by abandoning forms for its 100M users.

Inside Mercury's Conversational Founder Onboarding#

Mercury's conversational founder onboarding works by replacing the linear application form with an AI interviewer that captures narrative entity descriptions, runs AI-led compliance triage in the background, and escalates only the genuinely ambiguous cases to human review. The structure has three layers that operate simultaneously, not sequentially.

Layer 1 — Narrative entity description. Rather than asking a founder to map their business to a NAICS code (NAICS has 1,012 industry codes; founders know maybe three), the conversation asks them to describe what the company does in their own words, then proposes a structured classification back to them for confirmation. This is the same pattern Perspective AI customers use for running a customer discovery interview at scale — the AI does the schema-mapping work, the human just confirms. The captured narrative is then re-used downstream: it informs the expected-activity profile, it flags risk-sensitive categories (MSB, crypto, gambling, cannabis, etc.) for human review, and it becomes a reference document if the account is ever escalated to enhanced due diligence.

Layer 2 — AI-led compliance triage. Behind the conversation, an AI layer scores the application against Mercury's risk policy in real time. Low-risk applications — clean entity, U.S. founders, low-volume expected, software/services category — auto-approve. High-risk applications get routed to specialist reviewers with the relevant narrative passages pre-flagged. This is the conversational analogue of the underwriting pattern we documented in our AIG commercial insurance underwriting case study and the Zurich commercial lines customer discovery breakdown: AI handles the routine 80%, humans handle the consequential 20%, and the conversation creates an evidence trail that survives audit.

Layer 3 — Edge-case escalation. When a founder gives an answer that doesn't fit any pre-baked branch — "we're a U.S. C-corp but our customers pay us in stablecoins from a Singapore counterparty" — the conversation surfaces the ambiguity rather than forcing it into a wrong field. The case is flagged, a human reviewer is paged, and the founder gets a transparent "we need to ask a few follow-ups" message rather than a generic rejection or a silent queue. This single behavior is what most banks fail at — and it's why founders rate Mercury's onboarding so consistently above Chase, BofA, and SVB-successor offerings in independent reviews. Our founder customer research playbook covers the same listening pattern from the other side.

The post-activation layer matters just as much as the pre-activation one. Once the account is open, Mercury runs conversational customer research at cadence — segmented prompts to specific founder cohorts, in-product feedback that goes deeper than NPS, win-loss interviews when accounts close. This is the playbook we describe in How Top Founders Are Rethinking Customer Research and the continuous discovery report. Onboarding is just the front door; the conversation is the building.

Mercury Personal — Extending the Conversational Pattern to Consumer Banking#

Mercury Personal extends the same conversational onboarding pattern from business banking into consumer accounts, on the bet that the operators and founders who already trust Mercury IO will pay for a personal account that doesn't treat them like a 2008-era bank customer. Mercury Personal launched in 2024 with a waitlist and reached broader availability in 2025. The product targets a specific demographic: founders, operators, software engineers, remote workers, and other "complex consumer" segments who Chase and Bank of America underserve and who don't trust the dozens of neobanks that have flamed out since 2021.

The conversational pattern travels intact. Identity verification in consumer KYC is, on paper, simpler than business KYC — fewer entities, fewer beneficial-owner trees — but the narrative requirements are just as real. A high-net-worth founder who just had a liquidity event has a transaction profile that looks nothing like a steady W-2 employee, and a static form will either rate-limit the account into uselessness or miss enough context that the BSA officer can't defend the relationship later. Conversational onboarding handles this naturally: it asks about expected inflows, sources of wealth, and primary use cases the same way a wealth manager would in a first meeting. That same narrative pattern is what we cover in our voice-of-customer at scale teardown, and it generalizes well beyond banking — see the Maven Clinic women's health onboarding case study and the Webflow onboarding strategy for adjacent applications.

The strategic logic for Mercury IO → Mercury Personal is the same one Square ran from merchant to consumer with Cash App: own the operator's business banking, then own their personal banking, then own the wallet share. Conversational onboarding makes that cross-sell mechanically cheaper because the AI already has signal — the founder it's onboarding into Mercury Personal is, in many cases, a customer whose business it already understands. Personalization at that level is impossible with forms.

What This Signals for $2 Trillion Global Business Banking#

What Mercury signals for global business banking is that the $2 trillion industry's onboarding moat — the regulatory complexity that used to keep new entrants out — has flipped into a moat for whoever automates the narrative capture first. McKinsey's 2024 Global Banking Annual Review put global banking revenues at roughly $7 trillion, with business and commercial banking representing the largest single profit pool. Within that, account opening, KYC remediation, and ongoing CDD are the highest-cost-per-customer activities — and the slowest. A 2023 Boston Consulting Group analysis of corporate onboarding put commercial onboarding cycle times at 90+ days at incumbent banks, with up to 30% of applications abandoned before activation.

Mercury is showing that the right AI onboarding tool collapses that cycle from quarters to hours for the routine cases and from quarters to weeks for the genuinely complex ones. Three downstream effects worth watching:

1. Incumbents will buy or build conversational onboarding by 2027. The competitive pressure from Mercury, Brex, Ramp, and the EMEA equivalents (Qonto, Revolut Business) is now visible at the largest commercial banks. Citi, JPMorgan, and HSBC have all publicly announced "AI-first" onboarding initiatives in 2025–2026. The Allianz customer research strategy and Prudential's policyholder research approach show the same pattern in insurance: incumbents are racing to add conversational layers on top of existing core systems.

2. The unit economics of fintech will hinge on activation, not acquisition. When customer acquisition cost is fixed and onboarding completion is variable, the activation funnel becomes the lever. The 2026 form replacement report and the State of AI Onboarding 2026 document the same lift across SaaS — banking is later but bigger.

3. Compliance teams become product partners. When the onboarding conversation IS the compliance record, the compliance function moves upstream into product design. This mirrors the shift we cover in the forward-deployed engineering survey — domain experts embedded in product, not gating product. Compliance was historically a late-stage reviewer; conversational onboarding makes them a first-class user of the AI pipeline.

The Mercury playbook is portable. Any vertical where onboarding involves narrative-heavy compliance — wealth management, insurance, healthcare, B2B SaaS with regulatory tiers — can adopt the same pattern. The Carta equity platform research approach and our Glean enterprise discovery breakdown both show the conversational-onboarding pattern at work outside banking. Perspective AI is built for exactly this surface — the Intelligent Intake product, the Concierge agent, and the Interviewer agent together replace the form-driven intake stack with a conversational layer that compliance, product, and ops can all trust. For teams ready to map this against their stack, the use cases library and the pricing page are the next stops.

Frequently Asked Questions#

Who founded Mercury and what is its current valuation?#

Mercury was founded in 2017 by Immad Akhund (CEO), Max Tagher, and Jason Zhang, and is most commonly cited at a roughly $3 billion valuation following its 2024 fundraising round, per the company's public statements and reporting in TechCrunch and the Financial Times. Mercury reports serving more than 200,000 startups and small businesses and processing over $156 billion in annual transaction volume. The company is headquartered in San Francisco and partners with FDIC-insured Choice Financial Group and Column N.A. as its underlying banks.

What is Mercury IO versus Mercury Personal?#

Mercury IO is Mercury's core business banking product for startups and small businesses, covering checking, savings, treasury, and corporate cards. Mercury Personal is Mercury's consumer banking product, launched in 2024 and broadly available in 2025, targeting founders, operators, and other complex consumer segments. Both products share the same conversational onboarding architecture and the same partner-bank infrastructure; the difference is the regulatory regime — business KYC for IO, consumer KYC for Personal — and the product surface area.

How is conversational onboarding different from a smart form?#

Conversational onboarding is different from a smart form because it captures narrative answers in the customer's own words, branches dynamically based on context, and escalates ambiguity to human review rather than forcing it into the wrong field. A smart form just hides or shows fields conditionally — it still requires the user to translate themselves into the schema. Conversational onboarding inverts that: it captures the natural-language answer first and structures it afterward, which is what compliance teams actually need for narrative-heavy categories like nature-and-purpose, source of funds, and expected activity.

Why is KYC harder for startups than for established businesses?#

KYC is harder for startups because the entity is new, the beneficial-owner structure is often complex, the expected transaction pattern doesn't exist yet, and the regulatory category may shift as the product evolves. Established businesses give banks years of operating history, audited financials, and a stable industry classification — startups give banks an idea, a cap table, and a SAFE. AI onboarding tools handle this gap by capturing the founder's narrative description of the business and re-validating it against actual transaction behavior over the first 90 days, rather than trying to lock in a static classification upfront.

Can other fintechs copy Mercury's conversational onboarding model?#

Other fintechs can copy Mercury's conversational onboarding model, and many already are — but the harder part is not the conversation, it is the compliance orchestration behind it. The conversational layer is increasingly commodity (any team can build it on top of modern AI models or buy it from a platform). The differentiator is the policy logic that decides which answers auto-approve, which escalate, and how the narrative gets re-queried over time. That is institutional knowledge, not software, and it is the same lesson surfaced in Sierra AI's enterprise listening playbook.

How does conversational onboarding affect compliance and BSA/AML risk?#

Conversational onboarding generally improves compliance and BSA/AML risk posture because it captures richer, more auditable nature-and-purpose narrative than form fields can, and because the AI triage layer routes ambiguous cases to human reviewers earlier. The risk to manage is hallucination and inconsistent capture — the conversation must be deterministic enough that the same input always produces the same compliance outcome. Mercury, and other fintechs adopting the pattern, address this by using structured AI workflows with policy-coded branches rather than free-form LLM chat, and by versioning the conversation logic the same way they version any other regulated artifact.

Conclusion#

Mercury's bet is that the future of startup banking — and, eventually, the $2 trillion business banking market — runs on AI onboarding tools that treat the founder as a narrative, not a form submission. The 200,000+ startups Mercury already onboards are the proof-of-concept; Mercury Personal is the wedge into a much larger consumer footprint; and the conversational pattern itself is portable across every vertical where compliance, complexity, and customer velocity collide. Forms are not just a UX problem in banking — they are a compliance instrument that captures the wrong data at the wrong moment. Conversation captures the right data because that is how founders actually describe their businesses.

Perspective AI is the platform behind this pattern outside of banking. The Intelligent Intake product replaces form-driven intake with a conversational layer that captures the narrative compliance and ops teams need. The Interviewer agent handles structured customer research at scale, and the Concierge agent handles the activation moment itself. Whether you are onboarding founders, policyholders, patients, or enterprise buyers, the lesson from Mercury is the same: stop asking customers to translate themselves into your schema, and start a real conversation. Start a research project or explore the use cases to see what the pattern looks like for your team.