Tech & AI LIVE London: Victoria Grech, Head of AI at TAAP
Victoria Grech took to the AI Stage at Tech & AI LIVE with a mission: to spark vital questions about AI, how it’s evolving and where we’re taking it.
Her presentation, Beyond the Model: Architecting Multi-Agent AI for Human-Aligned Futures, was a provocative blend of technical insight, philosophical inquiry and practical advice drawn from her own experience building with AI.
The talk served not just as a showcase of emerging technologies, but as a call to action.
Victoria challenged the audience – comprising founders, SMEs, enterprises and developers – to think more critically about their relationship with AI and their role in shaping its trajectory.
Exploring AI's reality versus the hype
Victoria began with a warning: “Whether you're C-suite or deep in the weeds as a dev, it is not a magic wand.”
Instead, she likened it to a scalpel – a tool to be used with precision and intent.
Diagnosing the real business problem before diving into solutions is key, she stressed, warning that many organisations are still working from paper and Excel sheets rather than being digitally transformed.
She detailed the steps often missed: defining metrics, KPIs, and risk tolerance.
“AI is a very entrepreneurial game, failing in public is very difficult,” she added.
The insight was a timely reminder for enterprises still navigating early implementation phases.
When it came to agent setups, Victoria walked the audience through a range of architectures: from sequential and routing agents to more advanced supervisor-worker models and autonomous agents.
Yet she made it clear – regardless of sophistication, scalability remains elusive.
“This is break-fixing,” she said. “It's not quite there, but you can make it happen.”
Building responsibly with AI
While discussing her own multi-agent architecture, Victoria highlighted the persistent issues: memory fragmentation, data hallucinations, and vendor dependency.
Her chief concern? Memory.
“My Digital Chief of Staff needs to know every single conversation I have with every agent. Then we are there,” she said.
Victoria also underscored the complexity of using different models – like Gemini, Grok, OpenAI and Anthropic – across various agents for different tasks.
Choosing the right LLM was about intention, not hype: “Be in charge of what you’re creating.”
Crucially, she warned against overlooking the underlying data.
“If you don’t fix the groundwork, it’s built like your data in your enterprise. AI is never going to clean it up,” she said.
Small hallucinations at the start could spiral into compounded misinformation as data passes through agent workflows.
APIs, vendor licensing and client-side limitations were also pain points she flagged, calling for a shared protocol that allows for better orchestration.
“We’ve built machines to resemble thought, but we never designed them to think,” she remarked.
The consequence? Synthetic chaos in systems never designed to collaborate with each other – or with us.
Advocating for a future of shared intelligence
Victoria ended her talk with a stark philosophical pivot.
Referencing Bill Joy’s question, “Does the future still need humans?”, she challenged the room to think deeper than job losses or technological disruption.
Playing a tongue-in-cheek video that mocked the idea of a human-free, AI-driven future, she said: “I would go insane if I do nothing. Like, how do humans do that?”
She positioned herself as a ‘concerned optimist’ – someone who believes in a positive outcome but acknowledges we’re not there yet.
“We don’t need smarter agents. We need systems that think together with us, not for us,” she declared.
The key, she argued, lies in building systems that understand and care – ones trained not just on data, but on human values.
“The real race for me, it’s not data. It’s relational,” she said.
She concluded: “If we raise intelligence like a child – with ethics, with boundaries, with care and purpose – then maybe, just maybe, we would be able to build something that would change our humanity and future together.”
Essential diary dates for 2025
Discover the essential diary dates for Technology Magazine and AI Magazine and its sister publications Mobile Magazine and Data Centre Magazine.
To follow Tech & AI LIVE on LinkedIn, click here.
To enter for the Global Tech & AI Awards, click here.
- Tech & AI LIVE Singapore | 4 November
- Tech & AI LIVE New York | 18 November
Explore the latest edition of Technology Magazine and be part of the conversation at our global conference series, Tech & AI LIVE.
Discover all our upcoming events and secure your tickets today.
Technology Magazine is a BizClik brand
