Experiment freely, scale deliberately: Tony Marron on engineering rigour

AI is making software easier to build, but it is also making strong engineering harder to fake. When a team can move from idea to working solution in a fraction of the time it once took, the strength of what lies beneath, from architecture to everyday engineering practice, shows up far sooner. 

That idea sits at the heart of a recent CIOnews feature, Why Liberty Mutual Keeps Engineering Rigor At The Core Of Its AI Strategy. In it, Tony Marron, our Managing Director and Liberty Mutual Insurance’s Global Engineering Capability Lead, talks about how our parent company, Liberty Mutual Insurance, a Fortune 100 global insurer, is scaling AI adoption across the enterprise. 

From how engineering work is changing to how the company decides what to scale, Tony’s view of AI strategy keeps coming back to engineering rigour. It’s a theme that also shapes the work of our engineers, who build technology used across Liberty Mutual’s global business. 

AI is exposing the fundamentals 

For more than 20 years, software engineering followed a fairly predictable rhythm. Cloud and CI/CD pipelines changed how infrastructure was managed and how quickly new technology reached the business, but the shape of the work stayed largely the same. AI is different, Tony explains, because it is changing the technology, the way teams operate, and the culture around them, all at once. 

As AI lowers the barrier to building software, it raises the stakes for the engineering excellence underneath. The faster teams move, the sooner weak engineering and weak architecture come to light.

It’s also why strong software engineering fundamentals sit at the heart of how some engineers start their careers with us, through our TechStart graduate programme. 

Speed is built on foundations 

One example from the interview shows what this looks like in practice. Liberty Mutual built and deployed a new conversational AI experience for auto insurance quotes in roughly three months. Tony credits that pace to work done long before, as teams spent years making their services API-first.

The same thinking shapes how technology is sourced. Rather than treating build vs buy as an either-or decision, Liberty Mutual builds, buys and composes solutions as part of a single strategy, and uses what teams learn from building with new AI tools to make better buying decisions later. A common, composable architecture lets teams match each workload to the reasoning power, cost and speed it needs, without being locked into a single vendor. As Tony puts it in the interview, a company can buy a tool, but it still needs to own the rules that tool enforces. 

Experiment freely, scale deliberately 

If AI has made building faster, it has also made experimenting cheaper. That’s where Tony’s advice to experiment freely and scale deliberately comes in. Liberty Mutual starts with the business problem, then runs promising options through formal experiments alongside existing systems to assess how they fit the wider ecosystem before anything is scaled. 

Three levels of AI adoption 

To give that discipline a clear shape, Liberty Mutual organises AI adoption into three levels: 

• Everyday AI: individuals get the tools and the room to experiment, building their understanding of AI and their confidence using it in their work. 

• Supercharge: AI is embedded into the workflows, systems and platforms that span multiple teams. 

• Rethink: processes are reshaped around agentic engineering, with people and AI agents working together across workflows. 

All three run within security controls, AI governance and responsible AI practices, with human oversight and accountability at the centre. For Tony, the real value of the approach is that it brings our people along on the journey, so skills and architecture can mature together, deliberately. 

Agentic engineering in practice 

The same mindset shaped Liberty Mutual’s agentic engineering factory, built after the team found that several agentic engineering platforms lacked the composability they needed. It is changing engineers’ day-to-day work, with more time spent framing problems, designing systems, evaluating outputs, and deciding what reaches production. We’ve explored in more detail what changes for engineers when AI agents join the delivery team, and agentic delivery was one of the big conversations at Forward Live, our annual employee conference. 

From building for people to building with them 

For years, many engineering teams built tools for users they rarely spoke to. As AI lowers the technical barrier to custom software, more companies are bringing development in-house, and engineers are increasingly expected to work side by side with domain experts across the business. Tony describes this as a shift from building a solution for someone to building it with them. 

That shift matters for Liberty IT. As part of Liberty Mutual, our teams support global digital enablement through engineering excellence, and the closer engineering sits to the problem it is solving, the more judgement, communication and an understanding of the business count alongside technical depth, and the more engineers can shape the outcome as well as the code. 

Why engineering rigour matters for what comes next 

Taken together, the three themes tell a consistent story. AI is accelerating how software is built, and it rewards the foundations already in place: architecture that can flex, controls that keep people accountable, and engineers who work closely with the business they build for. 

Because moving fast with AI depends on what sits underneath. Strong engineering. Deliberate choices. And people who build together. 

Read the full interview with Tony on CIOnews.