AI Loan Origination, Tokenised Bricks and BNPL Growing Up: Inside the New Credit Stack
How a September packed with AI underwriting scale-ups, VARA-licensed property tokens and BNPL profit-hunts is quietly rebuilding the plumbing of lending.
Credit is having one of those weeks where every corner of the market seems to be moving at once. Small business lenders are pushing agentic AI from pilot to production, buy now pay later heavyweights are trading growth for margin, and PropTech platforms are stamping ownership onto blockchains from Dubai to Wroclaw. It is the sort of quarter that reminds you lending is not one industry, but a stack of very different bets on how money should meet risk.
Here is what actually changed in the last seven days across Lending, Credit and PropTech, and why any CFO, credit committee or founder should care.
AI Loan Origination Moves from Slide Deck to Balance Sheet
For most of 2024 and 2025, AI in lending was a story of pilots. As of this week, it has become an earnings story. According to PYMNTS reporting on the latest quarter, fintech lenders using automated underwriting are processing dramatically larger application volumes, with one major AI-driven platform seeing small business lending grow 48 per cent year on year and now making up the majority of its portfolio.
Vendors are quantifying the operational lift with the sort of numbers that make traditional loan officers nervous. Pennant Technologies notes that early agentic AI deployments in 2026 are delivering roughly 90 per cent faster processing, 67 per cent higher conversion and end-to-end automation of borrower communications. Timvero, another loan origination specialist, is telling customers to expect origination costs to fall by up to 60 per cent, with cycle times cut by a factor of ten.
The implications for SME finance are not trivial. Global Trade Review reported this month that specialist SecureLend has extended its AI stack into factoring, one of the trickiest, most manual corners of trade finance. Meanwhile, Biz2X, the technology arm behind Biz2Credit, is pitching its next generation SME lending suite around AI-native document intake, cash-flow scoring and dynamic pricing.
Why this matters for banks
The uncomfortable truth for incumbents is that agentic underwriting is not a feature, it is a cost curve. As CIO&Leader argued in a September 2026 analysis, AI, compliance automation and embedded finance are combining to create a new growth cycle in SME lending in which the winners are those who can price risk in minutes rather than days. Traditional lenders that treat AI as a productivity tool for existing underwriters will find themselves competing with rivals whose entire origination flow has been redesigned around it.
BNPL Grows Up (and Grows a Margin)
If SME lending is the growth story, buy now pay later is the maturity story. Klarna, freshly floated as KLAR, delivered Q2 revenue of 1.04 billion dollars, a 27 per cent year on year rise that beat consensus of 996.78 million dollars, per Yahoo Finance. Earnings per share came in at one cent, defying expectations of a six cent loss.
The market reaction was, however, less enthusiastic. Klarna cut its full year 2026 revenue guidance to a range of 4.08 billion dollars to 4.16 billion dollars, blaming roughly 600 million dollars of foreign exchange headwinds and softer German volumes. The shares fell 21 per cent on the print, and President's Capital Management moved the stock to Neutral from Buy with a 16.30 dollar target as of 9 September 2026, according to MarketBeat data cited by Yahoo Finance.
Affirm, by contrast, is currently the market's favourite child. It posted quarterly revenue of 993 million dollars, up 34 per cent and comfortably ahead of the 882 million dollar consensus. Adjusted EPS came in at 23 cents, more than double the 11 cent expectation, as Forbes columnist Zennon Kapron highlighted this summer in a comparison piece that has aged rather well.
The strategic read
Juniper Research still pegs BNPL at more than 50 per cent of the embedded finance market in 2026, so the category is not shrinking. What has changed is the scorecard. Investors are no longer paying up for GMV growth alone. They want unit economics, disciplined credit and diversified revenue lines such as merchant services and card-linked lending. Klarna and Affirm are both profitable now, but the market is telling us that scale without margin discipline is a story it has heard before.
PropTech Gets Real About Tokenisation
While lenders retool their credit engines, PropTech is quietly building the pipes for a very different kind of ownership. Tokenised real estate has been on the horizon for years. September 2026 is the month it started to feel operational.
Analysts covering the sector this week noted a rise in tokenisation flows through VARA-licensed platforms in the UAE, with one described model letting an owner mint 10,000 tokens at 100 euros apiece, sell 8,000 to retail investors and retain control of the underlying asset. A separate case study cited in September PropTech coverage described a 42-unit multifamily building in Wroclaw whose tokenisation cut closing costs by 28 per cent and shortened settlement to a matter of hours.
On the operating side, the plumbing is upgrading too. Qiddiya Investment Company selected Yardi's cloud-based real estate platform this month to run its commercial property operations, a marquee deal in Saudi Arabia's giga-project pipeline. VTS, meanwhile, has launched VTS AI, a generative and agentic layer aimed at leasing teams and asset managers who are drowning in tenant data.
Two forces converging
Two long-running trends are colliding here. The first is the maturation of asset servicing software, where names such as Yardi, MRI and VTS have gone from spreadsheets-in-the-cloud to full data platforms. The second is the legal and regulatory scaffolding for tokenised ownership, from the UAE's Virtual Assets Regulatory Authority to Europe's MiCA framework and the UK Government's own tokenisation taskforce launched earlier this year.
For lenders, this is a slower-burn story than AI origination, but potentially more transformative. Once property, receivables and inventory can be tokenised with credible legal wrappers, secured lending starts to look very different. Collateral becomes programmable, fractional and, in principle, tradable in secondary markets that never really existed before.
Embedded Finance Keeps Compounding
Underneath all of this sits the wider embedded finance flywheel. Bain & Company still projects embedded finance transaction value in the US alone doubling to 7 trillion dollars by 2026, and PitchBook's Q2 2025 Embedded Finance Tracker showed VC funding into embedded start-ups growing 22 per cent year on year even as broader fintech funding cooled.
The most concrete recent example came out of Vietnam, where Visa and Pismo announced a strategic collaboration with Circle Asia Technologies to unlock credit for millions of consumers, including the launch of the country's first genuinely AI-powered PayLater card. India, for its part, saw its markets regulator approve the IPO of Pine Labs, one of the largest names in point-of-sale financing and merchant technology in South Asia.
The pattern beneath the noise
Zoom out, and one pattern keeps repeating. Whether it is Affirm at the checkout, Circle in Ho Chi Minh City, or a Wroclaw landlord issuing tokens, the direction of travel is the same: credit and ownership are being unbundled from the balance sheets that used to hold them and rewired into software. That does not mean banks disappear. It does mean their role becomes less about being the interface and more about being the risk-taker, liquidity provider and licence-holder underneath somebody else's user experience.
What to Watch Next
Three things are worth keeping on the radar over the next fortnight. First, the next batch of BNPL numbers, which should tell us whether Affirm's beat is a one-off or the start of a durable divergence from Klarna. Second, further tokenisation pilots from the UK taskforce and DTCC, both of which are moving faster than most incumbents seem to appreciate.
Third, any move by regulators on AI underwriting explainability, especially in the EU where the AI Act's high-risk classifications increasingly bite in consumer credit.
For now, the message from this week is clear. Lending is not being disrupted so much as recomposed, one AI agent, one token and one embedded checkout at a time.
