Trending Articles
Trending Topics
Trending
This article is part of our Opinions section, where we invite industry professionals to share their views on the most pressing technology questions of our time.
A decade after the blockchain frenzy, we’re watching a similar story unfold in AI: on the one hand, optimism and overreach; on the other, making the same mistakes, namely underestimating the power of adoption, trust, and regulation.
Ten years ago, Australia explored the potential of blockchain. Teams across the public sector were exploring how the technology might streamline everyday infrastructure – vehicle registries, driver licences, even water management. The pitch was irresistible: decentralised, tamper-proof, and audit-friendly. If you could anchor identity to assets on a shared ledger, you could remove friction and fraud in one shot.
Only one of those projects survived – and even that with limited realisation. For most, the main barrier wasn’t technology; it was the cost of change. Replacing legacy systems or coordinating dozens of stakeholders proved far more expensive than the promised efficiency gains. Still, blockchain didn’t disappear; it found its place in well-defined, carefully scoped applications where the use case is clear and the economics make sense.
Today, I hear the same excitement and optimism about AI as about blockchain then. It’s familiar and a little uncomfortable. However, optimism without adoption is how great technology burns years and budgets. And while this time genuinely feels different, the risk remains the same.
Blockchain and AI didn’t promise the same things, but they sparked the same conviction that everything was about to change. With blockchain, that belief never translated into mainstream adoption. With AI, we’re watching entire industries move at speed because the value is visible. Yet the middle phase is still messy. Beneath all the big talk about transformation, the obvious questions are unresolved: What does security look like? Who carries liability? How will regulators respond? And what happens when something fails and needs to be rolled back?
Technologies don’t fail because cryptography is weak or models are mediocre. They fail because of a weak value proposition and/or enormous cost of change. The economics of a network beat the elegance of an architecture, every time.
Blockchain didn’t die. It matured quietly, inside finance. Tokenised cash equivalents and funds are the future. BlackRock’s BUIDL, a tokenised money market fund issued via Securitise on Ethereum, became the industry’s proof that tokenisation could be boring in the best possible way: regulated, auditable, and useful. Franklin Templeton’s FOBXX, which records ownership on-chain, is another example of incremental innovation that institutions can live with.
Market infrastructure followed the same logic. London Stock Exchange Group’s Digital Markets Infrastructure is being built for private markets and funds. This isn’t a publicity stunt, but because issuance and administration benefit from standardised, programmable rails. In post-trade, DTCC’s Project Ion runs a distributed ledger in parallel with existing systems to shorten settlement cycles and harden operational resilience. No heroics, no press-release promises to “replace everything by Q4.” Just parallel runs, controls, and a path to scale.
Why did these work? Because they kept three things in balance: regulation, incentives, and integration. They didn’t try to overthrow the system. They resolved real problems and became part of the ecosystem.
TradeLens, the IBM–Maersk supply-chain platform, was supposed to bring the world’s shippers onto a shared blockchain. Competitors refused to join. Without rival participation, the network effect never materialised, and the economics never clicked.
The Australian Securities Exchange tried to rebuild CHESS – the core clearing and settlement system – on a new DLT stack. Years and substantial expenditure later, the project was halted. The reasons cited were complexity, risk, integration, and liability. It turns out that replacing critical infrastructure mid-flight is not a hackathon.
Public-sector pilots such as land registries, legal documents, and water systems all looked great on slide decks. Even when the ledger is perfect, the governance and incentives often are not. If value appears only when everyone joins, almost no one will. And even then, it rarely lands evenly – it tends to skew to one side. These aren’t stories about bad technology. They’re stories about misjudged adoption.
Blockchain needed many parties to switch at the same time: banks, brokers, shippers, regulators, registries, and plenty more. The more stakeholders you require for value to show up, the lower your probability of rollout.
AI faces its own version of the same dynamics, but with one critical difference: this time, adoption is actually happening. We’re seeing massive investment, real deployments at scale, and momentum across industries, largely because AI delivers visible value, even if not always consistently. But the fundamentals still matter. Models rely on clean, permissioned data. Teams need to trust outputs and accept liability frameworks. Security and compliance demand evidence and controls. If marketing runs a brilliant AI pilot but risk and legal won’t sign off, you don’t have adoption. You have a demo. A technology without network economics is just a beautiful diagram in a slide deck.
Blockchain reached production not when we proved it was clever, but when we made it trustworthy. In finance, that means compliance. In compliance, that means regulators. And when regulators engage, the conversation changes from “cool” to “credible.”
The United Kingdom is moving tokenised funds into the mainstream with clear regulatory pathways, including the use of public blockchains. Market operators like LSEG are building within that clarity. In the United States, progress tends to come via market infrastructure and product approvals rather than a single omnibus law. Australia is pursuing a risk-based approach to AI, not a sweeping digital rights act, but proposed guardrails for high-risk uses, voluntary standards, and stronger privacy enforcement.
That last paragraph sounds dry. But this is what maturity looks like. Regulation doesn’t slow revolutions; it finishes them. In AI, trust is not a nice-to-have. It’s the only real currency.
We believed code could fix trust, and while it can certainly help, it cannot substitute for it. We imagined that decentralisation would remove bureaucracy. What it often removed was accountability. We thought immutability would settle every argument. It settled the wrong ones. The hard arguments, like incentives, power, and liability, stayed human.
That experience made me more optimistic about technology, not less. But my optimism now has conditions: design for adoption, align incentives, respect regulation, and measure trust. Blockchain was supposed to remove intermediaries. Instead, it taught us why they exist: because coordination is hard, trust is scarce, and shared accountability is valuable. AI won’t replace people. It will expose how well we cooperate, within teams, with customers, and with the institutions that make markets safe enough to scale. I still believe in technology. I just believe more in trust, design, and the patience to make adoption real. If blockchain was our first warning, let’s treat AI as our second chance.
