MICHELLE CONWAY - Lead Data & AI Scientist,Lloyds Banking Group

Some careers are mapped out early. Others reveal themselves gradually through curiosity, depth and persistence.

Michelle Conway’s path into data and AI was never about following fashion. It began with mathematics, sharpened through statistics and matured into leadership inside one of the UK’s most regulated institutions.

This is not a conversation about hype. It is about resilience, technical credibility, rejection, reinvention and what it really means to build AI at national scale.

Being the only woman in the room

JP: You were the only woman in your Maths Science degree at Dublin University. That is not just a headline, it is a lived experience. What did that feel like at the time, and how has it shaped the way you operate in data and technology today?

MC: I went to all-girls primary and secondary schools, so arriving at university and realising I was the only woman in my Maths Science course was a shock. It was the opposite of what I had known. In the first few weeks it felt daunting. I was not only adjusting to university life, I was adjusting to being visibly different in the room.

Very quickly, though, that feeling eased. I realised I was surrounded by people who were just like me in the ways that mattered. They were curious and they loved mathematics. That shared focus made the environment far less intimidating.

It was not entirely unexpected. Even in secondary school I watched the number of girls taking higher-level maths decline year after year. My year started with over a hundred girls. Around 35 chose higher-level maths initially. By the time we reached the Leaving Cert, the Irish equivalent of A Levels, there were just 12 of us left. University felt like the continuation of a pattern I had already seen.

Looking back, I am grateful for that experience. It prepared me for the reality of working in data and technology, where I have often been the only woman in the room or the only woman presenting. University taught me something simple and important. What matters is not who else is in the room, but whether everyone is there to solve problems and move the work forward.

Discovering statistics andthe logic behind the noise

JP: You have said your move into data came from genuine interest rather than a fixed career plan. When did it become clear that statistics was your direction? What drew you in?

MC: I studied Mathematical Science for four years and from the start I gravitated towards the statistical modules. By my final year, the choice felt obvious. Around 80 per cent of my courses focused on data analysis, data mining, modelling and statistical methods. I enjoyed every part of it.

What captured me was not just the numbers but the explanation behind them. I liked taking something abstract such as a distribution, a model or an algorithm and turning it into something that could be understood and applied. Statistics sits at the intersection of mathematical theory and structured communication.

I enjoyed presenting complex ideas and breaking them down so others could see how the models related to real behaviour. Whether exploring Bayesian distributions or the mechanics of linear regression, I was fascinated by how these tools could explain patterns that otherwise felt random. It felt like uncovering structure beneath apparent chaos.

At that stage I did not have a clear job title in mind. It was my professors who pointed me towards analytics roles in industry. At the time it was called analytics. Today we refer to it as data science, but the core principles remain the same.

From individual contributorto leadership

JP: Your move from delivery into leadership was not rushed. What were the decisions that pushed you forward, and were there moments when you questioned whether you belonged at a more senior level?

MC: The transition was gradual. I spent many years as an individual contributor because I wanted technical depth. I enjoyed being hands-on and staying close to the technology.

I moved deliberately across industries, from investment banking to insurance, then Amazon, BT and later data consulting. Each move expanded my perspective and exposed me to different operating models.

The real turning point came when I decided to deepen my engineering capability. I did not want to be seen solely as a statistician. I returned to study for an MSc in Data Science so I could build stronger foundations in Python, software engineering and DevOps. That shift allowed me to operate more confidently in end-to-end data science roles.

After more than a decade in the field, I became increasingly interested in leadership. Not because of status, but because of the opportunity to shape direction and support others’ growth.

I also faced rejection. I applied for senior roles and was not successful the first time, or the second, or the third. It took seven applications before I progressed. Each rejection was difficult, but each one strengthened my resilience. Senior leadership demands persistence. That lesson has stayed with me.

The years spent as an individual contributor now inform my leadership decisions. Technical credibility matters. It creates trust with teams and stakeholders. Although I am still early in my formal leadership journey, I value the balance between depth and direction that those years provided.

Learning leadership properly

JP: You have spoken about the shock of moving into management without formal preparation. What did you misunderstand at the start, and what do you know now that you did not then?

MC: Initially, I believed technical expertise would carry me. I thought leadership meant having the answers and being the most capable person in the room.

It quickly became clear that this was not enough. Leadership is less about technical precision and more about understanding people. Emotional intelligence is critical. You need to understand what motivates different individuals, what blocks them, how they respond to pressure and what they need from you. That awareness does not come from a programming manual.

I had to learn to adjust my communication style. Some team members need clarity and structure. Others need autonomy. Some need reassurance. Others need to be challenged. Listening became more important than speaking.

I also learned the importance of psychological safety. Teams perform best when they feel safe to raise concerns, suggest ideas and admit uncertainty.

A colleague once said that progression requires a PhD in psychology rather than a PhD in STEM. At the time it felt exaggerated. Now I understand exactly what she meant.

Building with Gemini

JP: Lloyds is working with Google Cloud and Gemini as part of its strategy. Setting aside corporate language, what are you actually building and why does it matter?

MC: Lloyds Banking Group works with Google Cloud as a major cloud provider, and Gemini plays an important role in our generative AI strategy.

Across the Group we combine build and buy approaches. On the buy side we use tools such as Microsoft M365 Copilot and GitHub Copilot. On the build side, engineering teams use Gemini APIs to create tailored generative AI products that operate securely on customer datasets.

These are not generic deployments. They are controlled, tested and deployed through rigorous MLOps frameworks so they can operate safely in production. That distinction is important in a regulated bank.

I cannot share commercially sensitive detail, but the impact is practical. We are seeing measurable improvements in speed and productivity. Processes that once took months can now be delivered in days. Repetitive, manual work can be automated or augmented.

The value is not simply efficiency. It is about releasing capacity for higher-value thinking and decision-making.

AI in a regulated environment

JP: There is a significant gap between the media narrative around AI and the reality inside a regulated UK bank. What does that reality look like in Consumer Lending?

MC: The media narrative focuses on speed and disruption. In a heavily regulated bank, the emphasis is different.

We work with sensitive financial data and operate under regulatory oversight. That changes the risk calculus. There is no appetite for uncontrolled experimentation. Everything must be explainable, fair, secure and robust.

That means structured experimentation, ring-fenced testing, strong governance and comprehensive security controls. It does slow development. But the reason is simple. When something is deployed at Lloyds, it must operate reliably for more than 30 million customers.

The scrutiny is high and the tolerance for error is low.