Affirm AI Strategy: How the BNPL Leader Uses AI for Merchant Onboarding and Customer Discovery in 2026

TL;DR#

Affirm's AI strategy is a tale of two customers. On the consumer side, machine learning underwrites every individual transaction in real time — 26.8 million active shoppers, 6.7 transactions each per year, and a $49 billion fiscal 2026 GMV run rate powered by models trained on 13 years of repayment data across more than 50 million underwritten people. On the merchant side, Affirm serves 515,000 active merchant accounts (up 44% year-over-year), and the integration story is conversational, not transactional. Affirm's BoostAI promotional-financing tool lifts merchant GMV 5%–15% per its own disclosures, but the harder question is the one BoostAI doesn't answer: why a Shopify merchant turned on Affirm, what conversion lift they expected, and which financing terms map to their margins. The consumer side runs on probabilistic risk models. The merchant side runs on relationships — and the firms winning those relationships in 2026 are the ones replacing onboarding forms with AI-led conversations that capture intent the way Affirm's underwriting captures repayment risk.

What is Affirm's AI strategy?#

Affirm's AI strategy is a dual-sided stack that uses machine learning to underwrite consumer transactions in real time while expanding AI-driven merchant tools — most recently BoostAI for promotional-financing optimization and a Stripe/Google agentic-commerce integration — to grow merchant gross merchandise volume. The company underwrites every individual transaction rather than the consumer, generating a unique risk score per shopper, per merchant, per moment in time, which is the core technical differentiator behind a 35% year-over-year GMV growth rate in fiscal Q3 2026.

The strategy splits cleanly along Affirm's two-sided marketplace, and the AI investment looks very different on each side.

The Two-Sided Marketplace Problem#

Affirm is a marketplace where two distinct customers — consumers who want transparent installments, and merchants who want higher AOV and conversion lift — meet at checkout. That structural fact shapes how Affirm thinks about AI.

The two sides have very different economics:

This is the same structural pattern we mapped in Shopify's 4.6 million merchants research stack — the higher-volume side gets ML, the lower-volume side gets conversations. Affirm just happens to be the BNPL leader sitting on top of Shopify's distribution.

How Affirm's Consumer-Side AI Works: Transaction-Level Underwriting#

Affirm's consumer-side AI is built around one commitment: underwrite the transaction, not the customer. Most lenders run a credit check on the person and hand out a credit line. Affirm runs a fresh ML evaluation for every purchase, scoring the specific consumer, merchant, cart size, and moment in time.

The system is called AdaptAI. It has been trained on 13 years of data covering more than 50 million underwritten consumers and over $100 billion in originated loans, per Affirm's investor relations. The model adapts to macro conditions and proactively flags consumers eligible for higher limits.

What this enables:

• Merchant-context-aware approvals. A consumer denied for a $1,200 purchase at one retailer may be approved for the same amount at another, because the merchant's default profile factors into the score.
• Zero-fee installment splits at scale. The $13.8B fiscal Q2 2026 GMV (per Affirm's Q2 shareholder letter) leans heavily on Pay-in-4 and 0% APR financing — both only pencil out with precise underwriting.
• Continuous adaptation. The model isn't retrained in batch; it ingests outcomes continuously.

This is a classic ML use case — high volume, fast feedback loop, sharp success metric. It is not the AI problem Perspective AI exists to solve.

How Affirm's Merchant-Side AI Works: Promotional Optimization and the Conversation Gap#

In May 2026, Affirm disclosed BoostAI, an internal tool that lets merchants split-test promotional financing offers — different APR structures, terms, and placements — to optimize GMV. Merchants using BoostAI have seen 5%–15% GMV lift, according to American Banker's PaymentsSource coverage.

The same month, Affirm announced an expanded partnership with Google to embed real-time underwriting inside Google's Android wallet, browser, and AI-powered shopping surfaces, per Affirm's investor announcement. In March 2026, Affirm extended its Stripe partnership to support shared payment tokens for agentic commerce — letting AI shopping agents complete purchases with Affirm financing on a shopper's behalf.

BoostAI, the Google integration, and Stripe agentic-commerce tokens are all optimization layers on top of an existing merchant relationship. They make merchants who already integrated Affirm convert more shoppers. None answer the question driving Affirm's 515,000-merchant install base: why does a merchant decide to integrate Affirm in the first place, and what do they expect after they do?

That's the conversation gap in Affirm's stack — and it's where conversational research belongs.

The Merchant Onboarding Problem That Forms Can't Solve#

Walk Affirm's actual merchant onboarding flow and the gap becomes obvious. A Shopify merchant clicks "Add Affirm," authenticates, pastes in API keys, enables test mode, runs a few test transactions, and goes live. The integration is well-engineered — the Adaptive Checkout API presents biweekly and monthly options side-by-side, and the dashboard shows GMV impact in near-real time.

But ask the harder questions and form-based onboarding falls apart:

• Why did this merchant choose Affirm over Klarna, Afterpay, or Sezzle?
• What conversion lift are they targeting? What's their threshold for "Affirm is working"?
• Which product categories and AOV bands will they push through?
• Are they defending AOV in a tightening consumer environment, or expanding into lower-credit-band shoppers?

These don't have dropdown answers. A merchant expanding into lower-credit-band shoppers needs a completely different account-management touch than one defending $400+ AOV against cart abandonment. But Affirm's onboarding form, like every BNPL competitor's, captures structured fields — EIN, monthly volume, AOV, vertical — and stops there. The qualitative "why now" is lost the moment the merchant clicks Submit.

The same pattern shows up across the fintech cluster: Stripe's 4 million businesses research operation, Plaid serving 8,000 fintechs, Chime's challenger-bank onboarding, SoFi's member-first conversational discovery, and Robinhood's customer-conversation strategy all share the same shape — consumer side runs on ML; partner side runs on conversation.

Why the Klarna Comparison Matters Here#

Affirm's largest BNPL competitor, Klarna, made the AI customer-experience case loudly when it announced its AI assistant was doing the work of 700 human customer-service agents — a story we covered in the Klarna AI customer service case study. The deflection-first framing left product, growth, and merchant-success teams without clear AI infrastructure of their own.

Affirm's strategy reads as the inverse: heavy investment in proprietary ML for consumer underwriting (where the data volume and feedback loop justify it), with optimization AI now layered on the merchant side. What's still missing across both Klarna and Affirm is the layer between "merchant signs up via form" and "merchant gets optimized via promotional AI" — the layer where merchants tell you, in their own words, what they're trying to accomplish.

That's also where 73% of top SaaS companies are dropping activation forms in favor of conversational onboarding. B2B SaaS has crossed this bridge. Fintech marketplaces — Affirm, Klarna, Shop Pay Installments — are next.

The Two-Sided AI Stack: Where Each Layer Belongs#

Here's how the layers fit together on a mature BNPL marketplace:

Affirm has layers 1–3 in production. Layers 4 and 5 are the conversational-research layer — and the data captured there determines whether 515,000 merchants stick, and whether 22% YoY consumer growth becomes repeat-transaction frequency or churn. For a broader map, see the modern customer research tools stack for 2026.

Where Perspective AI Fits in the Affirm Pattern#

Perspective AI is the AI-interview platform built for the conversation layer underneath every form-based onboarding and survey-based churn motion. For a BNPL marketplace like Affirm, the use cases are well-defined:

• Merchant onboarding interviews — replace the static post-signup form with a 5–7 minute conversational interview that captures intent ("why did you pick Affirm?"), expectations, and category mix. Route answers into the merchant CRM so account management segments by motivation, not GMV band.
• Merchant churn-risk interviews — when a merchant's GMV-through-Affirm drops 30% in a quarter, an automated interview asks what changed before the renewal conversation, instead of relying on a relationship manager's hunch.
• Consumer post-decline interviews — when a consumer is declined for a transaction, a follow-up captures whether they completed the purchase elsewhere and what alternative they'd have preferred.
• Consumer repayment-friction interviews — for consumers who flag a hardship event, conversational research surfaces why repayment broke down in a way ML can never see from transaction telemetry alone.

ML and conversational AI are complementary primitives, not substitutes. AdaptAI tells you what happened. Perspective AI tells you why — why the merchant integrated, why the consumer churned, why a decline happened where a longer-term option would have closed the sale.

Every two-sided fintech marketplace — payments processors, BNPL providers, neobanks with merchant arms — faces the same split. The high-volume side gets ML. The partner side has historically been built on forms, calls, and QBRs — none of which capture intent at the moment it's freshest. For Affirm, the prize is large: a 5–10% lift in merchant retention at 515,000 merchants compounds, and the data feeds back into product decisions about which financing terms to launch and which verticals to push into next.

Frequently Asked Questions#

What is Affirm's AI strategy in 2026?#

Affirm's AI strategy in 2026 is a dual-sided stack: real-time machine learning underwriting for consumers and promotional-optimization AI for merchants. The consumer side runs on AdaptAI, which underwrites every individual transaction (not the consumer) using 13+ years of repayment data across 50 million+ borrowers. The merchant side recently added BoostAI for split-testing promotional financing offers (5–15% GMV lift) and is expanding into agentic-commerce via partnerships with Stripe and Google.

How does Affirm's BNPL AI differ from Klarna's?#

Affirm's BNPL AI differs from Klarna's in where the investment is concentrated. Affirm has invested heavily in proprietary real-time underwriting ML for consumers and is layering merchant-optimization AI on top. Klarna has publicly emphasized AI customer-service deflection — most famously, an AI assistant doing the work of 700 agents. Both leave the merchant-onboarding conversation layer unfilled, which is where platforms like Perspective AI fit.

Why does Affirm underwrite each transaction instead of the consumer?#

Affirm underwrites each transaction because risk varies by merchant, cart size, and moment in time — not just consumer credit profile. A shopper with a strong repayment history may still be high-risk at a specific high-default-rate merchant. Per-transaction underwriting lets Affirm approve more consumers safely, support zero-fee installment splits without losing margin to defaults, and adapt continuously to macro shifts.

What is Affirm's BoostAI tool?#

BoostAI is Affirm's ML tool that helps merchants split-test promotional financing offers — different APR structures, term lengths, and placements — to optimize GMV. Merchants using BoostAI have seen 5%–15% lift. It sits on top of the merchant relationship rather than replacing it, which is why merchant discovery and onboarding still depend on a separate conversational layer.

How big is Affirm's merchant network?#

Affirm reported 515,000 active merchant accounts as of fiscal Q3 2026, up 44% year-over-year, with $11.6 billion in quarterly GMV (35% YoY growth). Distribution scales primarily through Shopify's Shop Pay Installments partnership — every eligible U.S. Shopify merchant gets access to Affirm's Adaptive Checkout — plus the expanding Stripe and Google integrations announced in 2026.

Where does conversational AI fit alongside Affirm's underwriting ML?#

Conversational AI fits in the layers Affirm's ML can't reach: merchant onboarding intent, churn diagnosis, consumer post-decline reasoning, and repayment-friction context. AdaptAI tells you what happened. Conversational AI tells you why the merchant integrated, why the consumer churned, and why repayment broke down — questions a form can't capture and an underwriting model has no signal for.

Conclusion#

Affirm's AI strategy is the clearest 2026 example of a two-sided marketplace using two-sided AI. The consumer side is a textbook ML problem, and transaction-level underwriting (AdaptAI, trained on $100B+ in loan history) is genuinely category-defining. The merchant side is structurally different, and BoostAI plus the Stripe and Google integrations are smart optimization layers on top of an existing merchant relationship.

But the relationship itself — the layer where a merchant tells you why they integrated Affirm and what they'd need to renew — still runs on forms. That's the gap in every BNPL provider's stack, not just Affirm's. The companies closing it treat conversational research as the partner-side counterpart to real-time underwriting, not an afterthought.

If you're building the AI customer-research layer for a fintech marketplace — onboarding interviews for merchants, churn-risk interviews for partners, post-decline interviews for consumers — Perspective AI is the platform built for exactly this work. Start a research study or explore the AI interviewer agent that runs the conversations.