Sydney and London: two very different markets. Rarely do commodities leaders get an intimate view inside both. Twenty-five years in technology leadership across firms including Lehman Brothers, RBS, and Macquarie gave Nilesh Khatri a front-row seat to three very different lenses to the same shift: a market rebuilding its relationship with risk. Having additionally worked as a fractional executive, advisor, and board contributor, he’s gained a perspective many execs never have access to: a cross-industry and pan-global view.

That vantage point, he says, has sharpened the picture. “I think for me, it was really about trust,” Khatri told us, reflecting on the throughline across Lehman’s collapse, RBS’s post-crisis rebuild, and fifteen years at Macquarie. “Technology’s role in this market throughout those events was really about how we earn trust and keep it. And it wasn’t always just about shipping new features.”

Two hemispheres, one directive

Khatri’s career gave him an unusual view before he ever went independent: Sydney from 2010, London from the tail end of 2016, several time zones apart. In Sydney, on Macquarie’s Futures Execution business, platform modernisation was inseparable from business strategy. It was a deliberate lever for lowering long-term costs, reducing the cost of change, and opening new markets; delivered, he’s clear to note, “without disrupting a live existing business.”

London added a different order of complexity: multi-asset, multi-regulator, a smaller relative team covering commodities and global markets that was growing quickly across EMEA. What that scale forced was proximity across the team. “We were much closer to the businesses. We partnered directly with COOs, desk heads, and trading desks,” he said. “What I took back to Sydney from London was that closeness around how we operated.”

Where regulation earns its place

Khatri’s later years at Macquarie centred on regulatory technology and non-financial risk. He’s unsentimental about what regulation is actually about. “Regulation earns its place when it forces genuine data quality,” he told us. Reforms like EMIR and MiFID, in his view, did something firms wouldn’t have funded on their own: forced real work on data lineage that ended up strengthening risk management well beyond the compliance case that justified it. That data lineage has become increasingly valuable as AI protocols have been introduced.

The friction, for commodities specifically, comes from the stack of regimes a firm runs in parallel — MiFID, EMIR, REMIT, and, where there’s a genuine US or Asian nexus, CFTC or ASIC — each with its own logic. REMIT’s carve-out for physically settled power and gas actually helps: it keeps physical and financial reporting separate rather than layered, so a genuine physical forward isn’t reported twice over.

That friction isn’t one trade being hit by every regime at once. It’s a firm building and maintaining several parallel reporting infrastructures for its overall commodity flow, because the regimes were never designed to share a data model. “You are kind of maintaining a complexity, not necessarily managing it,” Khatri said. “But there are huge benefits, and I think that outweighs the friction.”

What due diligence misses

Turning to commodities M&A and private equity, Khatri identifies the question that constantly arrives too late: which parts of the technology estate are genuinely irreplaceable, and why nobody has touched them. “That’s not really recognised before the deal structure is set, and it should be,” he said. The picture is incomplete when answered after terms are agreed; as a result, integration risk gets retrofitted rather than priced from the start.

His definition of a “clean” estate has little to do with modernity. He prefers the word legible: documented architecture, understood data lineage, a traceable line from trade capture to regulatory reporting without what he calls archaeology. Crucially, no single point of failure should live in one person’s head. “If it’s clean but only one or two people really understand it, then it’s not a clean estate,” he said. “It’s definitely a ticking time bomb.”

True legibility, in his experience, is rare. Most estates accumulate over ten or twenty years through individually sensible decisions that, viewed as a whole, no longer fully make sense. A buyer who does find a clean estate, he suggests, should ask harder questions before celebrating: is this genuine discipline, or have the difficult parts simply not been tested yet?

Cloud: the migration is only the starting point

At Macquarie, Khatri led the shift of over 80% of his team’s tech estate to cloud native, lifting platform availability from the high 90s to close to 99.5% — a jump made harder, not easier, by how high the starting point was. His central lesson from that work runs against a common assumption. “People assume that once you’re in the cloud, that’s your migration done,” he said. “But that often isn’t the end point. In some cases, that’s actually the starting point.”

The trap he’s watched other firms fall into is the “lift and shift” — moving existing platforms into a new environment without redesigning anything around them, then treating the relocation as the destination. What actually delivered the resilience gains, in his account, was the work that came after the move: rebuilding observability, adding redundancy, and instilling deployment discipline across teams. Firms that skip that step, he said, “haven’t really thought about what that means and how they might fundamentally benefit from the cloud.”

AI that actually changes the economics

Khatri also draws a firm line on AI’s back-office promise. Just as he cautions against a ‘lift and shift’ approach to the cloud, so too does he tread lightly when a firm wishes to integrate AI. Real economic shift, in his view, shows up specifically in pattern-based exception handling (reconciliation breaks, settlement fails, known and recurring issues) where volume has traditionally scaled with headcount. “AI is incredibly good at absorbing the volume,” he said. “It’s not just a productivity tweak.”

Outside that category, he is blunt: much of what gets called AI is automation wearing a new label, bolted onto processes that were never redesigned. The result is a faster version of the old workflow, not a cheaper one, because it still can’t scale without people. ‘His critique has a regulatory pedigree: years of transaction reporting taught him that a number only counts if it survives someone else’s follow-up question. He recently brought that instinct to moderating an industry panel on AI ROI, where the enthusiasm for AI often outruns the evidence for it.’ “Quite often it’s measured in hours saved on an existing process. And I think that’s probably the wrong metric,” he said. The better lens is whether the cost curve itself is changing shape — new economics, not faster automation. Success, he says, will be driven by good data governance.

Before the dashboard turns red

Asked what shows a commodities technology program is in trouble before the numbers confirm it, Khatri points to behaviour and language over metrics. Status reports that sit green for months and then flip suddenly to red mean nobody was tracking leading indicators, only lagging ones. A programme that gets re-planned repeatedly without any accompanying change in decisions or behaviour is, in his words, “also an indicator for me that something’s not working.” And if no single person close to the programme can give a straight, consistent account of where it actually stands, that’s a leadership or delivery gap — sometimes both.

His strongest signal, though, is people. “Where I’ve seen programmes that haven’t succeeded or have been failing is also where you’ve sometimes got your best engineers or your best team members quietly leaving,” he warned. “By the time that those attrition numbers show up, the programme’s probably been in trouble for a few months.”

The desk in 2029

We asked Khatri to look ahead: what does the next half decade look like across commodities technology? Khatri is candid that specifics will likely age badly at the pace things are moving. But the directions he’s confident in: reconciliation and exception management largely absorbed into AI-native workflows with real economic benefit, not bolted-on tools; real-time risk and regulatory reporting replacing end-of-day batches as regulators move toward T+1 and beyond; and richer pattern-based analytics becoming standard front-office kit rather than a differentiator.

What doesn’t change, in his view, is judgement and relationships, particularly in a regulated environment. “I don’t think AI will replace that,” he said. “It will allow leaders to make real judgement calls around where genuine knowledge can be used.” What he does expect is a thinner layer between decision and execution, freeing senior desk staff to spend more of their time on the calls that genuinely require experience rather than pattern recognition.

---

Nilesh Khatri spent 25 years in technology leadership across Lehman Brothers, RBS, and Macquarie, most recently as Global Head of Non-Financial Risk. His work as technology executive and board contributor spans commodities, regulatory technology, and platform modernisation.

---

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.