The cherry, the cupcake, and the ledger - Money 20/20, Day 1
Day One at Money 20/20 Europe 2025 opened with the usual constellation of logos, but the conversation landed on ground every finance leader knows by heart: Can we trust it? Can we audit it? And does it move the P&L?
Across many headline sessions, chats, and interviews, three themes kept surfacing: infrastructure before innovation, AI as a measurable co-pilot (not a magic wand), and an urgent push to de-risk cross-border cashflow.
1. Stripe’s Reliability Math: Why Uptime Still Rules
The morning opened with Stripe’s newly appointed CTO Rahul Patil, who came armed with figures designed to make any treasurer’s pulse slow.
-
“In 2024 we facilitated $1.4 trillion in global commerce—roughly 1.3% of world GDP—and we did it at 99.99989% uptime,” he told a packed Na.i.ture Stage. “That’s about 35 seconds of blip in an entire year.”
Patil’s subtext was clear: scale matters only if the rails never wobble. For finance officers mapping liquidity across dozens of entities, the allure is obvious.
Even the session’s lighter “over-hyped/under-hyped” rapid-fire segment boiled down to reliability. Asked whether AI will reduce the need for developers, Patil replied, “Over-hyped. We’ll write more software than ever, so you’ll need more builders, not fewer.”
Translation: budget for AI talent, but keep the SRE team fully funded.
The other Stripe headline—stablecoins—landed squarely on treasury desks. Patil claimed merchants that accept stablecoins grow “two to three times faster globally” and post “average order values three times larger.”
Far more provocative was his assertion that stablecoin-based treasury management is already live at SpaceX’s Starlink. Patil framed the tokenized dollar as simpler overseas working capital, not a crypto side gig.
Takeaway: Stablecoins have crossed from edge case to credible liquidity tool, but only when paired with industrial-grade uptime. The board will ask whether your payment infrastructure can match Stripe’s five-nines; be ready.
2. Klarna’s Cost Story: AI as a P&L Lever
Conversational commerce darling Klarna provided the day’s most practical AI case study. CMO David Sandstrom recapped an internal rollout that slashed campaign production costs by double-digit percentages and, more importantly, shrank time-to-market from days to hours.
“We’re not doing anything new,” he said. “We’re doing everything faster, cheaper and better thanks to AI.” The company may be best known for consumer-side chatbots, but Sandstrom insisted the bigger win is operational efficiency:
“AI on the creative side means the cost of failure drops to near-zero. We can prototype ten ads, kill eight and post the winners before lunch.”
For CFOs, the hero metric is Klarna’s head-count trend: flat to slightly down even as content output keeps rising.
Sandstrom added that AI hasn’t cooled hiring for strategic roles—just redirected budget from repetitive production to insight work.
Takeaway: Treat GenAI as a throughput accelerator with a measurable payback window. The fastest path to CFO approval starts with hard savings in marketing ops, not nebulous “personalization” promises.
3. Mastercard’s Fraud Math: 20% Lift, 28 Hours Saved
If Patil set the reliability bar, Mastercard’s Chief AI & Data Officer Greg Ulrich raised expectations on fraud and internal productivity.
Speaking with Fortune’s Ryan Hogg, Ulrich detailed how multimodal GenAI improved Mastercard’s transaction-risk score by 20% across 159 billion annual card swipes.
“We can map a region’s merchant landscape and predict the sixth shop on a cardholder’s route, even if they’ve never been there,” he said. The kicker: fewer false positives and faster approvals, a win for both revenue and customer-service spend.
Internally, Mastercard’s AI “consulting agent” serves 3 000 employees and has cut portfolio-insight prep from three weeks to “two to three minutes.”
Another agent removed 28 hours from client onboarding. Ulrich labeled the philosophy “co-intelligence,” a balance where machines handle rote tasks while analysts focus on “insight work, creativity and strategy.”
Headcount? Still rising. AI hasn’t replaced jobs but has “shifted the bell curve right,” making low-performers visible for coaching.
Takeaway: AI investments should carry discrete KPIs—fraud-rate delta, hours removed, satisfaction scores—and a quarterly review cadence. Anything less is R&D theater.
4. Experian and Koodoo: Consistency Beats Flash
Mid-morning, Experian COO Shail Deep and Koodoo CEO Andrei Lebed demoed two AI products that push advice, not just data, to end users.
Koodoo’s mortgage agent has already passed the UK’s regulated mortgage exam with 94% accuracy. Its first-time-buyer journey flips the old rate-table paradigm into WhatsApp-based conversation and, crucially, hands off to a human only for final sign-off.
The Financial Conduct Authority’s reaction? Surprisingly positive. “They liked the transparency,” Lebed said. Consistency across advisors could shrink compliance tail risk as much as it improves customer experience.
Experian’s Eva personal-finance coach heads the same direction: actionable AI that freezes a credit file on request, suggests subscription cancellations and logs future follow-ups. Deep admitted the back-end still requires “QA on steroids,” but early tests show tangible NPS gains.
Takeaway: AI advice engines may not sit in the finance stack today, but their regulatory-grade audit trails could soon inform credit-risk models and customer due-diligence checkpoints.
5. IBM Cloud: The “Cupcake” Parable of Infrastructure
Amid the AI euphoria, IBM’s Kamini Belday reminded CFOs that no innovation matters if the plumbing fails.
Only 15% of core banking and payments workloads currently reside in the cloud, she said. Her metaphor landed: “AI is the cherry on top. You need a perfectly baked cupcake first—and that cupcake is infrastructure.”
IBM’s sell is a zero-trust, compliance-by-design cloud mapped to FFIEC, PCI and forthcoming DORA mandates.
Whether or not you buy IBM, the larger point stands: regulators are moving from principles-based to framework-based oversight. Any transformation roadmap that punts infrastructure risk to “phase two” will stall at the approval gate.
Takeaway: Budget cycles must separate “cupcake” spend (core modernization) from “cherry” pilots (AI, stablecoins). Mix them and neither gets funded.
6. Trust & Transparency: Agentic AI
The afternoon panel—Beyond the Hype: Building Trust and Transparency with AI Agents—added a missing layer: explainability and liability in an agent-driven world. Moderator Georgina Bulkeley (Google Cloud) pressed three executives for practical governance.
Liberis: Auditable, Bias-tested Underwriting
CEO Rob Straathof unveiled Ada, an AI underwriter built with Google. Three controls anchor the design:
- Audit trail—every recommendation can be traced back to source data.
- Bias review—personal attributes (age, address) are stripped from training sets, with two independent teams policing drift.
- Human feedback loop—underwriters can override and the system relearns.
Lloyds Banking Group: Payments Without a Checkout
Payments executive Samantha Emery warned that agent-to-agent commerce blows up familiar KPIs like checkout conversion and top-of-wallet. Liability models must contemplate a fridge ordering butter and a bot buying a bespoke sofa. Financial literacy gaps could widen if agents obscure the money layer.
Softcat: Ethics Starts With Data Hygiene
Chief Strategist Katharine Wooller argued that trustworthy AI begins with “rubbish-free” data and relentless iteration. Boards should fund AI Centres of Excellence owning policy, metrics and continuous optimisation. Without enterprise-wide literacy, Wooller said, “AI won’t take your job—someone who knows how to use it will.”
Regulation: Gaps and Overlaps
All three panellists stressed that sector-specific regulators aren’t set up for cross-industry agentic flows. Emery called for “open-living” policy frameworks; Straathof cautioned that any decision an agent makes must withstand the same three-lines-of-defence audit a human decision faces.
Boardroom Scorecard
| Theme | Hard Metric Quoted on Stage | Implication for CFOs |
|---|---|---|
| Uptime & Scale | 99.99989% reliability on $1.4 trn volume (Stripe) | Button up vendor-risk assessments; five-nines is the new threshold. |
| Fraud & Risk | 20% lift in fraud-score efficacy (Mastercard) | Quantify fraud-loss reduction before approving new AML spend. |
| Productivity | 28 hours removed from onboarding (Mastercard) | Bake hour-savings targets into AI vendor SLAs. |
| Cloud Readiness | Only 15% of core workloads in cloud (IBM) | Treat infra modernization as multi-year capex, not optional opex. |
| Advisory Consistency | 94% exam score for AI mortgage agent (Koodoo) | Pilot regulated advice bots where human variability is highest. |
The Forward Agenda
Day Two promises sessions on programmable money, CBDC pilot results and buy-side sustainability metrics, but the opening salvo already set a demanding bar.
Finance chiefs now have a shortlist of questions to ask their teams before quarter-end:
- What uptime SLAs do our payment partners actually meet?
- Which fraud KPI will AI improve—false positives, approval rates or both?
- How many hours can GenAI agents demonstrably remove from onboarding or reconciliation?
- Are we tracking cloud migration as a compliance imperative or a tech nice-to-have?
- Where can AI deliver regulator-approved consistency in customer advice?
The message from Amsterdam’s first day is unambiguous: flashy demos matter less than metrics with line-item impact. As Stripe, Mastercard and IBM each showed in their own way, the winners in 2025 are those who treat AI as an ROI engine sitting on bullet-proof rails—not as an experiment bolted onto yesterday’s spaghetti.
Check back tomorrow for continued reporting from Money 20/20 Europe 2025.
Was this article helpful?
YesNo