What is Affirm's AI strategy?
Affirm builds its AI in-house and points it at three jobs: deciding each loan, choosing which financing offer to show, and writing its own software. Its fiscal 2026 10-K says its solutions use machine learning, artificial intelligence and cloud-based technologies and that Affirm prioritizes building its own technology. On offers, the FQ2'26 shareholder letter of February 5, 2026 says AdaptAI was deployed on nearly all Affirm surfaces and contributed to an approximately 10% year-over-year improvement in end-to-end checkout conversion, and that BoostAI, a newer model that runs automatic A/B tests of financing programs, showed merchants an average 5% to 15% increase in GMV. In engineering, Affirm paused delivery for a week in February 2026 and made Claude Code its default agentic coding tool; by April 2026 over 60% of pull requests were agent-assisted. CEO Max Levchin told analysts on May 7, 2026 that using AI this way is "unequivocally accretive to the bottom line", and CFO Rob O'Hare put the spend at "very low single-digit millions per quarter". Levchin said on the same call that Affirm is "not planning AI-related layoffs, full stop." Technology and data analytics expense rose $157.4 million, or 27%, in fiscal 2026. Affirm does not disclose a separate AI budget, AI headcount or revenue from AI products.[1][2][3][4][5]
Data as of Oct 22, 2026 · Affirm profile updated Oct 19, 2026 · 17 sources cited on this page · 52 fintechs profiled on this site · Markdown
What have Affirm's leaders said about AI, and when?
Dated statements beyond those on the profile (the FQ3'26 letter's "truly AI-first" line, the CNBC interview and the Investor Forum). Quotes as published.
| Statement or move | Who | When | Source |
|---|---|---|---|
| AdaptAI and BoostAI, "a pair of our internally-developed yield optimization technologies", played "an outsized role" in the holiday quarter; AdaptAI contributed to an approximately 10% year-over-year improvement in end-to-end checkout conversion | Max Levchin, CEO | February 5, 2026 | FQ2'26 shareholder letter (Form 8-K)[2] |
| Discussing BoostAI as a pipeline for merchant revenue: "It looks more and more like an advertising model versus a cost of acceptance model"; "It's definitely early days for BoostAI" | Max Levchin | February 5, 2026 | FQ2'26 earnings call, reported by American Banker[6] |
| "It's long overdue to get rid of bad software"; companies that make software without proprietary data or value-add will be replaced by AI coding | Max Levchin, on the Sourcery podcast | April 14, 2026 | Business Insider[7] |
| Using AI in development is "unequivocally accretive to the bottom line"; "we think we can 10x this productivity further"; "we don't anticipate any sort of a decimation of the engineering team" | Max Levchin | May 7, 2026 | FQ3'26 earnings call transcript[4] |
| AI developer-tool costs are not material to the P&L, "very low single-digit millions per quarter in terms of spend" | Rob O'Hare, CFO | May 7, 2026 | FQ3'26 earnings call transcript[4] |
| "We are not planning AI-related layoffs, full stop"; AI tools are "giving us rocket boosters, wings, whatever metaphor you want" | Max Levchin | May 7, 2026 | FQ3'26 earnings call, reported by American Banker[5] |
| Technology and data analytics expense up $157.4 million, or 27%, partly from amortization of internally developed software; solutions use "machine learning, artificial intelligence, cloud-based technologies" | Fiscal 2026 annual report | August 27, 2026 | Form 10-K[1] |
Which AI products has Affirm built, and in what order?
All built in-house except the coding agent. Results are Affirm's own measurements.
| Product | What it does | Built on | Since |
|---|---|---|---|
| AdaptAI | Selects the financing offer for each transaction to maximize conversion and minimize risk; deployed on nearly all Affirm surfaces, including digital wallets, by February 2026 | In-house yield optimization model | Opened to merchants April 30, 2025[8][2] |
| BoostAI | Runs automatic A/B-test programs at merchants, introducing new financing programs and merchant-funded incentives; merchants saw an average 5% to 15% GMV increase; no integration work required | In-house model | Disclosed February 5, 2026[2] |
| BoostAI users | 47 enterprise customers and hundreds of small and medium-sized businesses | Same | February 5, 2026[6] |
| Real-time bank-balance signals in underwriting | Uses account balances and cash-flow trends for Affirm Card users who link a bank account or hold an Affirm Money Account | In-house underwriting models | January 14, 2026[9] |
| Agentic coding workflow | One task, one agent session, one pull request; over 60% of pull requests agent-assisted by April 2026 | Claude Code (Anthropic) as default tool | February 2026[3] |
| Generative AI customer-service chatbot | Disclosed as a use of generative AI in the risk factors | Vendor not named | Disclosed in the fiscal 2026 10-K[1] |
| Transformer underwriting model | Reads the order and timing of events in a credit history; 3.4% more completed purchases than a control group in first deployment | In-house transformer model | September 17, 2026[10] |
Who leads AI at Affirm, what does the 10-K say, and what is not disclosed?
Levchin describes himself as "personally invested and involved in" software development. President Libor Michalek oversees engineering, risk, operations, product and design, and SVP of Product Vishal Kapoor fronts Affirm's agentic commerce announcements; no chief AI officer is named in the 10-K.[4][11][12][1]
The fiscal 2026 10-K says Affirm uses generative AI in "certain aspects of our business, including our customer service chatbot", and that generative AI may create inaccurate or misleading outputs or reflect unintended biases. Its stated controls include "human involvement in the training and monitoring of our AI tools" and aligning AI development with secure development guidelines. It names the EU AI Act among the laws its business is subject to.[1]
Levchin told analysts that AI-written code still passes "unique-to-Affirm checks and balances" because "if we make an underwriting mistake, if our engine somehow treats some consumer unfairly or if we're off by a penny here and there, like none of that is okay."[4]
What is not disclosed: a dollar AI budget beyond the CFO's range for developer tools, model providers other than Anthropic's Claude Code, revenue attributable to AdaptAI or BoostAI, and the share of customer contacts the chatbot resolves.[4][3][2][1]
What does the Affirm profile record on this topic?
The rows below are the entries of the Affirm AI strategy profile that bear on this page, as of Oct 19, 2026; the profile holds the full record of 13 dated moves, 8 use cases and 15 sources.
| Use case | What the record shows | Status | Source |
|---|---|---|---|
| Transformer-based real-time underwriting | Every purchase is underwritten individually in real time; the fiscal 2026 10-K says the risk model turns key user identifiers and transaction details into over 1,000 data points (over 500 in the fiscal 2025 10-K). The transformer model, live at US checkout since September 2026, added 3.4% completed purchases in initial deployment. | In production | [10][1][13] |
| ML fraud detection | Affirm's fiscal 2026 10-K says its machine-learning risk models, calibrated on data from over 553 million loans, are custom built to detect fraud and price risk; after a basic identity check, its fraud models use approximately 350 data points (approximately 200 in the fiscal 2025 10-K) to decide near-instantly whether to block a transaction, alongside rules and investigator review. | In production | [1][13] |
| Generative-AI customer-service chatbot | The fiscal 2026 10-K says Affirm currently uses generative AI in its customer-service chatbot, with human involvement in training and monitoring. | In production | [1] |
| Agentic software development | After the February 2026 AI Retooling Week, over 60% of Affirm's pull requests are agent-assisted, with Claude Code as the default agentic coding tool and human checkpoints at planning, review and merge. | In production | [3][14] |
| AdaptAI personalized credit offers | AI-powered platform matching promotions such as 0% APR or special repayment terms to a consumer's purchase and profile, in use on Affirm's own products and offered to merchants since April 2025. | In production | [8][1] |
| Date | Move | Detail | Source |
|---|---|---|---|
| Sep 18, 2026 | Levchin: no staff cuts to fund AI | On CNBC's Closing Bell Overtime, Levchin discussed the new AI underwriting system and said Affirm has not had to cut staff to implement AI. | [15] |
| Sep 17, 2026 | Transformer underwriting model goes live | Affirm launched a transformer-based model that reads the order and timing of events in a credit history; in its first deployment it produced 3.4% more completed purchases than a control group, with a proprietary algorithm providing explainability at checkout speed. | [10] |
| Aug 27, 2026 | Fiscal 2026 10-K and a second President | The 10-K reports $50.2 billion of GMV, 27.8 million active consumers and 2,358 employees, and discloses a generative-AI customer-service chatbot; the same day Michael Linford became a second President alongside Libor Michalek, who keeps engineering, risk and product. | [1][11] |
| May 12, 2026 | Investor Forum: transformer models outperforming | At its 2026 Investor Forum, President Libor Michalek said Affirm had processed $150 billion of volume and tracked 2.3 billion repayments over 14 years, and that its newer transformer-based underwriting architecture was already outperforming previous model generations. | [16] |
| May 7, 2026 | Shareholder letter: engineering 'truly AI-first' | Levchin's FQ3'26 letter said the retooling week made agentically written code dominate non-agentic 'essentially overnight', that an order-of-magnitude productivity gain is plausible, and that Affirm still expects to modestly grow its development team. | [14] |
| Apr 23, 2026 | Affirm details its AI Retooling Week | Affirm's engineering blog said that in February 2026 it paused normal delivery for a week and asked over 800 engineers to take real tasks through a fully agentic workflow, with Claude Code as the default tool; over 60% of pull requests are now agent-assisted. | [3] |
| Aug 29, 2025 | Levchin cites a 5% merchant-volume lift from AdaptAI | After fiscal Q4 2025 results (revenue $876 million, up 33%), Levchin highlighted AI, noting early AdaptAI deployments had delivered an average 5% lift in merchant volume. | [17] |
| Apr 30, 2025 | AdaptAI opened to merchants | Affirm made AdaptAI, its AI-powered promotions platform, available to merchant partners for personalized, real-time promotions and credit offers at checkout, after it drove nearly 10% incremental improvements in conversion rates on the Affirm App and Affirm Card. | [8] |
Is Affirm cutting jobs because of AI?
No, according to its CEO. Max Levchin told analysts on May 7, 2026 that Affirm is "not planning AI-related layoffs, full stop", and said Affirm does not anticipate "any sort of a decimation of the engineering team". His FQ3'26 letter said Affirm expects to modestly grow its development team.
What are AdaptAI and BoostAI?
Two in-house Affirm models that choose financing offers. AdaptAI picks the offer for each transaction and, per the FQ2'26 letter, contributed to an approximately 10% improvement in end-to-end checkout conversion. BoostAI runs automatic A/B tests of financing programs; merchants saw an average 5% to 15% increase in GMV.
Which AI coding tool does Affirm use?
Claude Code. Affirm's engineering blog says it chose Claude Code as its default agentic coding tool for its February 2026 retooling week, when more than 800 engineers moved real work to agents, and that over 60% of pull requests were agent-assisted by April 2026.
How much does Affirm spend on AI?
Affirm gives only a range. CFO Rob O'Hare said in May 2026 that AI developer-tool spend was "very low single-digit millions per quarter" and not material to the P&L. Total technology and data analytics expense rose $157.4 million, or 27%, in fiscal 2026, but AI is not broken out.
What AI risks does Affirm disclose?
Its fiscal 2026 10-K says its generative AI uses, including its customer-service chatbot, may produce inaccurate, incomplete or biased results that are hard to detect, which could harm its reputation or bring damages or fines. It also cites the EU AI Act among the laws that apply to it.
- Affirm Holdings Annual Report (Form 10-K) for the fiscal year ended June 30, 2026
- Affirm FQ2'26 Shareholder Letter (Exhibit 99.1 to Form 8-K)
- How Affirm Retooled its Engineering Organization for Agentic Software Development in One Week
- Affirm (AFRM) Q3 2026 Earnings Call Transcript
- Affirm's Levchin rejects AI layoffs
- Affirm details new AI tool for merchants
- Max Levchin says vibe coding will replace a certain kind of company
- Affirm launches "AdaptAI" - its AI-powered promotions platform - to merchant partners
- Affirm updates underwriting with enhanced signals to better reflect consumers' real-time finances
- Affirm launches transformer-based machine learning model for real-time underwriting
- Affirm Holdings Form 8-K (appointment of Michael Linford as President)
- Affirm expands Stripe partnership to support Shared Payment Tokens for agentic commerce
- Affirm Holdings Annual Report (Form 10-K) for the fiscal year ended June 30, 2025
- Affirm FQ3'26 Shareholder Letter (Exhibit 99.1 to Form 8-K)
- Never had to cut staff to buy more AI tokens, says Affirm CEO as it rolls out AI underwriting
- Affirm Targets Card Networks With Broader Payments Push
- Affirm stock rises 10% as CEO Levchin notes continued consumer strength
When Affirm moves on AI, you'll read it here first.
when a bank, fintech or regulator moves on AI, it is in the next morning's brief · six sourced stories · 7 am PT · free
plus every regulation, bank and agent page update, the morning after · leave any morning