After diving into Multiply's vision (the "Why") and our tech stack (the "How"), our "Beyond the Numbers" event brought together a panel of industry leaders to explore "What's Next?"
With insights from Bianca Yuen (Google), Matt Mair (Monzo), Aritra Chakravarty (Lloyds), Denish KC (Google) and our own Mike Curtis, a few powerful themes emerged about the real-world deployment of this technology.
1. Surprise finding: Customers prefer the AI
The old assumption was that chatbots were for "deflection" - a cheap way to avoid costly human interaction. Matt from Monzo shattered this myth.
He shared that roughly 50% of Monzo's support queries are now handled by LLMs. The truly shocking insight? They are seeing higher customer satisfaction and longer, more in-depth conversations with the AI than with human agents.
Rather than deflecting users, the AI is encouraging them to engage more. This led Monzo to a key realisation: their human agents were often constrained by overly restrictive, "computer-says-no" guidelines. The AI's success actually gave them the confidence to "loosen" the rules and empower their human team to be more helpful, closing the loop.
2. The "trust iceberg" is real
In his keynote, Denish from Google showed how easy it is to build a stunning, multimodal AI customer-service agent in minutes, with no code. The demo is the easy part.
Matt Mair called this the "trust iceberg." The demo - the delightful, human-like interface - is just the 10% you see above the water. The other 90%, hidden below the surface, is the massive, complex engineering required to make that experience reliable, safe, and trustworthy.
This underwater mass is all about evaluation frameworks, guardrails, audits, and, as Aritra from Lloyds put it, co-creating the entire system design (fallbacks, human-in-the-loop, deterministic layers) with your compliance teams from day one.
3. Advanced AI is about context and persona
How do you build that trust? Aritra Chakravarty (Lloyds) offered a brilliant technical insight. He used the ambiguous sentence: "The chicken is ready to eat."
Does that mean a hungry human or a hungry bird? An LLM doesn't know and often won't ask the clarifying question.
Aritra's solution for advanced AI systems is to engineer two things:
- A memory architecture: Explicitly structuring memory into "past, present, future" or "personal, professional, financial."
- A persona: He shared a personal example of giving his AI the persona of a Himalayan mystic, which fundamentally changed the pace and quality of the interaction.
For enterprises, this means we can’t just feed the AI data. We must give it context and a persona that aligns with the task. This is the difference between a simple Q&A bot and a true digital adviser.
4. The "long tail" of adoption is here
Finally, Mike Curtis shared a story about his mum, a non-technical 65-year-old, who was "buzzing" after she successfully used AI to help her write a speech.
For those of us in the tech bubble, it's easy to forget that most people aren't using these tools daily. This story was a wake-up call: the long tail of mainstream adoption is here, and it's happening faster than we think.
The future is a human-AI partnership
The event's key takeaway was clear. The technology is no longer the barrier. The challenge is in the implementation: building trust, ensuring safety, and designing systems that are fundamentally human-centric.
The future isn't AI or humans. It's a new partnership, and at Multiply, we're proud to be building the safe, auditable, and powerful systems that will power it.
Thank you for following our "Beyond the Numbers" series. If you're a technical leader looking to deploy safe, human-centric AI in a regulated space, learn more about Multiply's platform or get in touch with our team.
.png)
