The world of generative AI is awash with bold claims and contrasting narratives. While a recent MIT study found that 95% of enterprise AI initiatives generate no measurable returns, it also noted that the remaining 5% followed a clear strategy to achieve genuine business productivity and adoption. The challenge is clear: how do we move beyond pilots and hype to deliver tangible, scalable impact?
To answer this, we were honoured to speak with Suraj Halai, Director of AI at Centrica Energy. Suraj leads a team delivering measurable value through AI, with over 20 uses cases in production, over £1 million saved, and thousands of manual hours freed for higher-value work. In this article, he shares his journey to building a multiple award-winning GenAI operation that delivers real impact.
From concept to reality
It was the summer of 2023, and I had just been presented with a demo of GPT-3.5 analysing George Orwell’s Animal Farm and extracting key themes and characters. People were impressed, but that quickly turned into the familiar question: “What’s the business value?”. With a background in commodity trading risk management, I had a use case in mind, though I doubted whether LLMs could deliver reliable results given concerns like hallucinations. There was only one way to find out.
After an afternoon of experimenting with a large dataset, the use case proved itself, and I went from being an AI sceptic to a convert. Within six weeks, my team moved the idea to production, fully automating a manual process and saving each business user 2–4 hours per workflow. The impact was immediate. Risk and position reporting in our Energy Trading Risk Management (ETRM) system became instant, and the use case went viral internally, shifting perceptions around GenAI. This success was not accidental. It was achieved by focusing on necessity-driven innovation, robust back-testing to ensure consistency, and by keeping a “human in the loop” to supplement AI with domain expertise.
We NEED a machine learning expert
Following our initial success, leadership called for a Director of AI to drive transformation and see how far we could take GenAI. The brief was clear: find a machine learning expert to lead the charge.
I disagreed.
Terms like “NLP”, “LLM”, and “hallucinations” mean little to our stakeholders. What was really needed was someone with deep domain expertise who could connect the dots across the entire trade lifecycle. That person happened to be me. As a hands-on Senior Business Analyst and Programme/Delivery Manager, I had 22 years of experience across oil, gas, power, freight, coal, and metals, including eight years implementing LNG on EOS and 14 years delivering Endur projects from greenfields to migrations. I knew exactly where AI could add real value.
I was so confident in this approach that I even told them they could fire me after six months if we didn’t see real progress. Luckily, it paid off, thanks to a team effort (and a fair amount of weekend work).
Small team, big impact
Our core team is lean. For most of this journey, our team’s been four developers. We’ve now since expanded to the excessive number of six, including myself! We own the entire lifecycle, from ideation and business analysis to development, testing, deployment, and ongoing production support. This startup mindset allowed us to fail fast, prioritise quick wins, and stay focused on measurable outcomes such as cost and time savings. Despite minimal investment, the return we’ve generated has been phenomenal, driven by my team’s relentless execution of our shared vision.
Governance & safety by design
Embedding governance from the start was crucial to our success, but we had to work hard to create the right forums without stifling innovation. Regular scrums and central oversight helped maintain standards, while mandatory security reviews and governance boards ensured clear prioritisation, leadership, and trust. Far from slowing us down, governance accelerated delivery and strengthened stakeholder confidence.
Stakeholder journey: From push to pull
Initially, GenAI adoption was IT-led, with centrally defined use cases and education programmes. Over 12 months, as success stories and measurable impact emerged, we saw a shift. Users began voluntarily adopting and experimenting with AI tools, motivated by productivity gains and a genuine desire to upskill. This pull from users, supported by active change management and accessible tools, created organic momentum and a culture of innovation.
Recently, we introduced a curated AI curriculum with self-paced courses for both technical and non-technical colleagues to develop their skills. The uptake has been fantastic, adding even more momentum to our journey.
Tangible outcomes: metrics that matter
We are relentless about delivering value for our stakeholders, measuring impact, and finding ways to do more with the same resources wherever possible. Here are some highlights from our 21 use cases in production:
- One AI application completed in 1.5 days what would have taken 81 years of full-time effort.
- Manual legal and compliance tasks reduced from days to minutes.
- Over £1 million saved in third-party licence costs since January 2025, with an app scaled to 37,000 colleagues.
- Thousands of hours of manual effort removed across the trading lifecycle.
- Agentic AI processes now live in production.
Conclusion: The road ahead
When I took on the role of Director of AI a year ago, there was no GenAI playbook for building an AI team, establishing governance, upskilling developers, or creating the right support structures. Nor were there historical use cases to draw from. Our journey has instead been defined by a unified vision, strategic focus, a scalable ecosystem, and an innovation mindset. With the right leadership, culture, and relentless focus on value, we’ve delivered positive disruption and set a new standard for AI in energy trading.
As a result, our team has earned both internal and external recognition. We’ve won two industry AI awards, hosted ministerial visits, and delivered C-suite training at Microsoft HQ. These milestones are important markers in our journey, but they are by no means the destination. The journey continues, driven by experimentation, a focus on outcomes, and a commitment to bringing our business stakeholders with us.
The views expressed in this article are my own and do not necessarily reflect those of my employer or any affiliated organisations.
Looking for more insights?
Get exclusive insights from industry leaders, stay up-to-date with the latest news, and explore the cutting-edge tech shaping the sector by subscribing to our newsletter, Commodities Tech Insider.
