Introduction

In the final part of our Liberis x Google blog series, we’re pulling together some of the sharpest perspectives from across the fintech and AI ecosystem, straight from the panel that closed out the day.

This wasn’t a product showcase. It was a candid, fast-moving conversation between three people building with AI at the edge of what’s currently possible:

  • Rob Straathof, CEO of Liberis
  • Camille Rougie, CEO of Plural.AI
  • Sufyaan Kazi, Head of Pre-Sales for Fintech at Google Cloud

Moderated by Bella Renney-Thwaites, Product Director at Liberis, the discussion tackled everything from overhyped buzzwords to regulatory bottlenecks, scaling strategies, and what the next five years might actually look like in practice.

Here are the big insights that closed out the day — and the series:

1. The Agentic Era isn’t on the horizon. It’s already here

Agentic AI isn’t a concept — it’s already powering production systems across fintech. Rob Straathof shared how Liberis has kept its headcount flat at around 240 people since 2022, but increased revenue nearly 10x in that time. That’s not a marginal efficiency gain — it’s AI multiplying human output across underwriting, operations, and product development.

Where tasks once took analysts days — like financial modelling or documentation review — tools now do the same in minutes with 80%+ accuracy. It’s not perfect. But for most junior analyst work, it’s already better than human first drafts. And it’s only getting faster.

2. AI adoption is accelerating, but maturity is wildly uneven

By the end of this year, the World Economic Forum predicts that 85% of financial institutions will have adopted AI. Of those, 75% are reporting cost savings of up to 83%, and 70% report improved customer experience.

But those headline figures hide a deeper gap. While 50% of fintechs are running 12 or more AI use cases, another 25% still only have two or three, with many stuck in the transition from proof-of-concept to scalable deployment.

Only 11% of fintech executives say they feel confident in their responsible AI frameworks, while 65% cite regulation as the main barrier to doing better. There’s progress, but it’s fragmented, and the risks of getting it wrong are real.

3. Agentic systems aren't a fit for every use case, and that’s fine

Not everything needs GenAI. Camille Rougie highlighted the importance of asking whether AI is even needed at all. In many enterprise workflows, especially regulated ones, reliability and predictability still matter more than autonomy.

Agentic systems shine where there’s unstructured data, variability, and complexity, but full autonomy introduces risk. A system that’s 95% accurate may sound great, but in a 20-step workflow, that can lead to catastrophic error rates without proper oversight.

The best outcomes today come from bounded systems, purpose-built agents with human fallback mechanisms and clearly defined decision scopes. They’re not black boxes. They’re tools that work with people, not instead of them.

4. The real value lies in learning, memory, and traceability

Systems like Ada, Liberis’ AI underwriter, don’t just make decisions. They explain them. Every approval or decline comes with reasoning, linked data, and full traceability. That’s crucial not just for transparency, but for auditability in regulated environments.

Google’s Sufyaan Kazi reinforced this with a critical point: explainability is non-negotiable. Their Explainable AI toolkit, available through Vertex AI, is designed specifically to give enterprise teams clarity around AI decisions. Especially in lending, fraud, or compliance — if you can’t explain the “why,” you can’t deploy it.

5. What’s coming next? AI co-pilots that drive real revenue

We’re on the verge of something bigger than chatbots and smart underwriting. Rob described the future of embedded finance platforms: AI agents acting as CFOs, CMOs, and COOs for small businesses, operating quietly in the background.

These agents won’t just respond to merchants. They’ll negotiate with platforms like Liberis in real-time on things like funding rates, inventory costs, and marketing ROI.

It’s already happening in prototype. Next comes agent-to-agent negotiation, with AI co-pilots on both sides of the conversation. For SMBs, that means faster, smarter decisions. For partners, it means stickier platforms, better performance, and a step-change in the value they can offer.

6. Scale matters, but so does safety

Camille and Sufyaan both emphasised the importance of designing for trust from the start. In one example, a client estimated a tool would take eight months of engineering time, but with Google’s Agent Developer Kit, they built it in three days. Speed is there. But without the right controls — audit trails, safeguards, and fallback protocols — speed alone is dangerous.

The most successful fintechs aren’t the ones racing to adopt every new AI tool. They’re the ones building clear frameworks, designing for explainability, and keeping humans in the loop, especially when real money and people’s lives are involved.

Final word: This is just the beginning

If there’s one message from the panel, it’s this:

The Agentic Era isn’t about replacing people. It’s about amplifying them.

The businesses that win won’t be the ones shouting the loudest about AI. They’ll be the ones that embed it thoughtfully into workflows, into platforms, and into the small moments that make a big difference for customers.

At Liberis, that means:

  • Building with partners, not just for them
  • Embedding smarter capital tools through the Liberis Capital Platform
  • Bringing funding to the point of need with Pay with Liberis
  • Speeding up complex decisions with explainable AI, via Ada
  • And working hand-in-hand with pioneers like Google and Plural.AI to push what’s possible

This blog may be the final part of the series. But the next chapter — the real one — is being written right now, one AI-powered decision at a time.