SHIFT Community Update #53
From experimentation to operational deployment across fintech
Hey SHIFT Community,
Across the market, momentum is increasingly coming from firms turning new infrastructure into scalable products, operational workflows and measurable customer outcomes. Open banking is maturing, AI is moving closer to core operations, and partnerships between incumbents and fintechs continue to signal where the next phase of value creation is likely to come from.
At SHIFT, that is exactly the kind of progress we care about.
In this month’s newsletter, we look at the market shifts shaping fintech, firms building real traction across the ecosystem, standout progress from SHIFT community members, Woodhurst’s latest report release, and an upcoming SHIFT breakfast session focused on innovation funding and scaling.
🔍 Across the Market
Execution continues to define the fintech market this month, with activity increasingly concentrated around firms turning infrastructure, AI and partnerships into practical commercial outcomes
Some of the developments that stood out this month:
NatWest announced the firms selected for its 2026 Fintech Programme, reinforcing continued institutional appetite for specialist fintech partnerships and AI-led innovation.
Barclays released its AI 100 Watchlist, highlighting the UK’s fastest-rising AI businesses across sectors. Fintech was the largest category, with 14 companies included across the list. Among the fintech firms featured were MQube, Aveni, Quantexa, Payhawk and Signal AI.
Santander-backed Ebury secured fresh funding to accelerate expansion, product development and AI capability across its international payments business.
Corpay’s partnerships with JP Morgan and BVNK highlighted continued convergence between traditional payment infrastructure and blockchain-enabled settlement models.
Open banking is moving beyond experimentation, while AI increasingly dominates the conversation: a theme echoed at this year’s London Venture Capital Summit and across the main agenda items for the upcoming London Tech Week.
🚀 SHIFT Community shoutouts
A few strong signals from across the SHIFT community this month:
Doshi was named in Tech Nation’s Breakout 50, following a standout year that also saw the company win MoneyLive Startup of the Year and surpass 200,000 active users.
Adfin announced an $18 million Series A round, bringing total funding to $30 million as the company continues scaling its embedded finance and agentic finance offering across modern payment operations and workflows.
BankiFi partnered with Lloyds Banking Group on the launch of a free, fully embedded Making Tax Digital solution for businesses, continuing the firm’s expansion across SME banking infrastructure and embedded financial tooling.
📊 Woodhurst Updates
This month, Woodhurst published their latest report:
Powering the Lenders: The Data Modernisation Playbook
Built from conversations with leaders across regulated lending and financial services, the report explores a challenge we continue to see across the market: growing ambition around AI and analytics, but data foundations that are often not yet ready to support it.
The whitepaper looks at how organisations can move beyond fragmented reporting towards trusted, scalable data infrastructure, and the practical steps required to make that transition successfully. Take a read here.
📅 Events
SHIFT x ForrestBrown: Non-Dilutive Funding Breakfast
We’re hosting our next SHIFT breakfast session with ForrestBrown, focused on how fintechs and technology-led businesses can better access non-dilutive funding, innovation incentives and grants as they scale.
The session will bring together a small group of fintech founders for a practical discussion on how funding markets are evolving and where opportunities are emerging across the ecosystem.
📍 Kensington, London
📅 9th June, 8.30am - 10.00am
This is an invite-only session. If you’re interested in attending, please reach out to sami.jabbar@woodhurst.com.
🏗️ The Build Series
For the sixth instalment of the Build Series, we sat down with Crediflow, one of our SHIFT members building FlowSpread, an AI-native platform transforming commercial lending workflows and financial spreading.
1. Has there been difficulty digitising financial spreading, if so, why do you think this workflow has been a difficult problem to solve?
The difficulty isn’t reading the text; it’s understanding the context and accounting for human subjectivity.
Financial spreading has been the bottleneck in commercial banking for decades, and standard digitization tools like OCR (Optical Character Recognition) have historically failed to solve it.
The problem isn’t text extraction, it’s contextual understanding and structural variance of accounting standards and formats. Commercial lenders deal with messy, unstandardized corporate data. Every company’s accountant formats a balance sheet or P&L slightly differently. Line items are named uniquely, debt schedules are buried deep within multi-page PDF footnotes, and regional accounting frameworks dictate entirely different reporting standards.
Traditional automation relies on rigid templates: if a line item moves, the system breaks. Furthermore, single LLMs alone often ‘hallucinate’ or lack cross-document consistency checking, which is crucial in credit risk. To truly automate spreading, software cannot just parse text; it must understand accounting frameworks and principles, read between the lines of footnotes, categorise and dynamically map disparate data points into a unified, audit-ready structure output. That requires a cognitive, reasoning and deterministic architecture, not just a data extractor.
2. FlowSpread is designed to process unstructured financial documents across formats, languages and reporting standards. What were the biggest technical or operational challenges in building something flexible enough for real-world commercial lending environments?
Building for enterprise banking requires moving beyond simple prompt engineering to a multi-agent system of action and deterministic architecture that guarantees data lineage and zero hallucinations.
The biggest technical challenge was achieving absolute deterministic accuracy and complete auditability within a highly flexible system. In enterprise credit underwriting, a 95% accuracy rate is actually a failure. If an AI miscategorises a single line item or flips a negative sign on a line item, this could lead to a loan being mispriced.
Operationalising this for a production ready lending environment required two major breakthroughs:
Moving from OCR and Single LLMs to Agentic deterministic Workflows: We had to move away from standard OCR and prompt engineering and instead build a specialised document intelligence pipeline with network of AI agents and deterministic logic. One agent acts as the expert accountant analysing context, another validates language and regional reporting differences, and a third performs strict mathematical validation. They cross-check one another’s work in real-time.
Deterministic Data Lineage (Source-to-Cell mapping): Credit teams are naturally sceptical of ‘black box’ AI. The breakthrough with FlowSpread was engineering exact coordinate mapping. Every single cell in a generated output is permanently linked back to its precise pixel location in the original unstructured source PDF document. If a credit analyst or loan officer wants to audit a figure, they click the cell and see the exact highlight in the original PDF. Solving that technical lineage loop was critical to earning clients’ trust.
3. What do you think commercial lending operations will look like in 3-5 years’ time with the support of building an AI infrastructure for commercial credit?
Moving from a reactive, manual workflow to a proactive, continuous, autonomous credit model where relationship managers actually have time to manage relationships.
In 3 to 5 years, the role of the commercial credit analyst will completely transform from data entry to strategic risk management. We will see the rise of “continuous underwriting”.
Right now, lenders operate reactively: a business applies for a loan, the manual process runs for weeks, a decision is made, and the financial health of that business isn’t deeply reviewed again until the annual review.
With a robust AI infrastructure underpinning the industry, the entire lifecycle becomes continuous and real-time. More importantly, post-decision portfolio monitoring will move from annual snapshots to continuous, real-time risk evaluation. AI infrastructure will dynamically track borrowers’ data, market adjustments and macroeconomic data, alerting risk teams to potential vulnerabilities before a default occurs.
Ultimately, it will democratize access to capital. By slashing the operational cost of processing complex deals, financial institutions will be able to efficiently serve mid-market and small businesses that were historically locked out of traditional debt markets due to manual underwriting overhead and increased cost of lending.
If you want to reach out to Crediflow to learn more, message us on Slack!