PUBLISHED DATE: 2025-08-14 17:33:30

Beyond the AI Hype Cycle

VivaTech 2025 Financial Services Takeaways

What we saw on the Paris show floor that actually matters for banks and insurers
“At VivaTech, the shift was clear: clients no longer want GenAI POCs; they want GenAI outcomes. The conversations that stood out were about real impact—modernizing legacy systems, disrupting operations, and automating the SDLC with tailored intelligence from agentic AI. The future isn’t just about models—it’s about architecture, context and execution.”

— Pinak Kiran Vedalankar, Group VP of Technology and International Head of Engineering, Publicis Sapient

SDLC Transformation: From 18 Months to 12 Days

On the show floor, we demoed our Sapient’s Slingshot platform, where our engineers showed how we rebuilt a legacy trading system’s front-end in just 12 days. Not a prototype. Not a proof of concept. A production-ready system with full compliance documentation, automated testing, and real-time risk controls.

Traditional financial services SDLC runs 18-24 months for major releases. With AI-powered development platforms, teams are shipping weekly. One European bank reported cutting their mobile app release cycle from quarterly to every two weeks, while improving their defect rate by 40%.

Rakesh Rovuri, our CTO, demonstrated how their platform handles the unique challenges of financial services: embedded compliance checks, automatic generation of regulatory documentation, and built-in controls for data residency. When a developer tries to implement a feature that violates GDPR or MiFID II requirements, the system flags it before code review.

Legacy Modernization: Your COBOL System Just Became an Asset

NVIDIA’s partnership with major European banks revealed the killer application for enterprise AI: understanding and modernizing legacy systems. Their demonstration showed an AI system reading 40-year-old COBOL code, generating comprehensive documentation, identifying business logic, and creating modern microservices, all while maintaining regulatory audit trails.

Société Générale’s innovation lead shared their results: “We have 60 million lines of legacy code. Traditional modernization quotes came in at €200 million and 5 years. With AI-assisted modernization, we’re doing it in phases for €40 million total, and we’re already live with the first modules.”

The key insight? Stop trying to replace legacy systems. Transform them into well-documented, API-accessible services that can feed modern applications. Your mainframe becomes a feature, not a liability.

The "Show Me the Money" Moment Finally Arrived

After three years of breathless AI demos and PowerPoint promises, VivaTech 2025 marked a turning point. Financial services executives weren’t asking “what’s possible?” anymore. They were demanding proof of production deployments, cost reductions, and measurable outcomes.

One wealth management CTO put it bluntly: “I’ve seen enough chatbots. Show me how you’re cutting my release cycles from months to weeks.”

The good news? The vendors finally had answers.
Bonjour! Ask me how I can save you some of your time.

The Context Problem: Solved with Structure

Every financial services executive mentioned the same challenge: "How do I get the right context into the AI?" Generic LLMs don’t understand trade lifecycles, regulatory requirements, or why certain decimal precision matters for derivatives pricing.

The winning solutions aren’t using generic AI. They’re building what Mistral AI calls "sovereign models," or AI systems trained on domain-specific data with built-in understanding of financial services requirements.

Mistral’s demonstration for a French investment bank showed their system correctly identifying regulatory breaches that GPT-4 missed entirely. Why? Their model was trained on 10 years of regulatory decisions, enforcement actions, and compliance documentation specific to European financial services.

"Context isn’t just memory," explained their head of enterprise. "It’s structured understanding of how financial services actually works."

Quantum: From Theoretical to Tactical

The surprise announcement came from IBM and Crédit Agricole: quantum computing for risk modeling is moving to production pilots in 2026. Not 2030. Not "someday." Next year.

The use cases are specific:
An executive from Allianz noted: "We’re not waiting for quantum to arrive. We’re building quantum-ready architectures now. When the hardware catches up, we’ll be first to market."

Back-Office Disruption: Zero Ops as a Journey

The phrase that kept coming up in executive sessions was "zero ops as a journey." Not the elimination of operations teams, but the evolution from manual processing to intelligent automation with human oversight.

BNP Paribas showcased their KYC transformation. Previously, onboarding a corporate client required 15 different systems, 200+ manual checks, and 35 days.

Their new AI-powered system:
The head of operations was clear about the impact: "We redeployed 60% of our KYC team to relationship management. Same headcount, 5x more client interactions."
"Context isn’t just memory. It’s structured understanding of how financial services actually works."
"We’re not waiting for quantum to arrive. We’re building quantum-ready architectures now. When the hardware catches up, we’ll be first to market."
"We redeployed 60% of our KYC team to relationship management. Same headcount, 5x more client interactions."

Ready to Turn VivaTech Insights into Action?

Publicis Sapient’s Financial Services practice can help you:
Contact us to schedule your executive briefing and see how these solutions work with your systems, your data, and your regulatory requirements.

Dave Murphy
Head of Financial Services, EMEA and APAC
Publicis Sapient
david.murphy@publicissapient.com

Pinak Kiran Vedalankar
Group VP of Technology and International Head of Engineering
Publicis Sapient
pinak.vedalankar@publicissapient.com

publicissapient.com