OP-ED: How AI and Tokenisation are Redefining Finance
2025 made one thing obvious: finance is being rebuilt. More than $400 billion has flowed into AI, tokenisation, and digital infrastructure to cut costs and deliver more personal products. The decision is no longer "should we adopt?" It's "how do we win?"
AI's real capabilities (and where it actually pays)
AI use inside companies jumped this year, with most firms running at least one AI-enabled function and pushing it across departments. Many are already seeing earnings lift, with a meaningful share of EBIT tied to automation and smarter workflows. What's surprising to many teams: AI is often strongest in creative and product-adjacent work-campaigns, messaging, rapid prototyping-rather than old-school number crunching.
AI is also flattening the playing field. Smaller teams can ship at the speed that only incumbents could afford a few years ago. Yes, most organisations are still piloting, but the direction is set: by 2026, it will be unusual to find a product org without AI embedded in daily work. The question is who builds systems that compound, not pilots that stall.
If you want a benchmark for adoption patterns and impact, the State of AI research from McKinsey is a useful reference point. Read the report.
Practical AI moves for product teams
- Prioritise 3-5 workflows with measurable drag (manual reviews, QA, reporting, support triage). Set a 90-day target: cut cycle time by 40-60%.
- Adopt a "data first" stance: define data contracts, retention rules, and feedback loops before model selection. Good data beats clever prompts.
- Pick a simple stack you can support: retrieval for accuracy, lightweight guardrails, and clear human approval steps for anything customer-facing.
- Build vs. buy: buy generic layers (OCR, transcription, summarisation), build where you have proprietary data or a defensible workflow.
- Establish decision rights: what the model can auto-approve, what needs human sign-off, and what gets escalated.
- Track real business metrics: cycle time, error rate, approval rate, cost per action, and EBIT contribution-not vanity stats.
Tokenisation makes markets interactive
Tokenisation is pushing finance toward fully digital, programmatic assets. The market is set to multiply, and with it, the appetite for fractional access and automated execution. Buying exposure to a blue-chip stock as a low-cost token instead of a full share is one example. Another: token-wrapped certificates of deposit-simple, compliant, and tradable on digital rails.
As issuance, trading, and settlement turn into code, the work changes. Strategy libraries will trade on rules. Rebalancing and compliance checks will run on schedule. Some junior roles will shrink as routine analysis gets automated. The upside is clear for product teams: lower unit costs and the ability to launch faster with less operational overhead.
For a view of institutional experimentation, see MAS Project Guardian on asset tokenisation. Explore the initiative.
How to build a tokenised product the right way
- Start with one asset class and one region. Prove issuance, transfer, and redemption end-to-end before you scale.
- Choose your chain pragmatically: ecosystem support, custody options, auditability, and fees matter more than hype.
- Design for compliance from day one: KYC/AML gates, whitelists, transfer restrictions, and transparent audit logs.
- Handle keys like production secrets: HSMs, role separation, and clear recovery procedures.
- Use reliable oracles and clear pricing rules. Ambiguity in price feeds is a fast path to disputes.
- Make exits boring: predictable redemption, fiat on/off-ramps, and support workflows that don't break under stress.
AI x Tokenisation: the compound effect
AI and tokenisation work best together. AI cuts human time on intake, analysis, and review. Tokenisation converts assets and actions into code that can execute under rules. Combined, they turn slow processes into programmable systems with human oversight where it matters.
- Data plane: standardised data contracts for pricing, positions, clients, and compliance events.
- Intelligence plane: classification, summarisation, anomaly detection, and scenario testing.
- Asset plane: tokens with clear permissions, lifecycle states, and embedded compliance.
- Execution plane: rule engines that route, trade, rebalance, and escalate exceptions.
Humans make the final call
AI is trained on past data and can carry bias. In finance, you can't outsource accountability. Keep a human in charge of the final decision on anything with balance-sheet or client impact. That's not a constraint; it's how you move fast without breaking the business.
- Define approval thresholds: what AI can execute vs. what needs a human.
- Implement review queues with clear SLAs. Don't leave approvals stuck in limbo.
- Log every decision with input data, model version, and human approver for audit readiness.
- Run periodic red-teaming for prompts and rules. Kill switches are non-negotiable.
Your 90-day product plan
- Weeks 1-2: Pick one AI use case and one tokenised asset flow. Map data, risk, and success metrics.
- Weeks 3-6: Ship the smallest shippable version. Add human review, logging, and basic guardrails. Go live with a controlled cohort.
- Weeks 7-10: Measure cycle time, errors, and cost per action. Tighten prompts, improve data quality, and remove dead-end steps.
- Weeks 11-13: Expand scope or add a second workflow. Start building the strategy library and standardising components for reuse.
Career edge for product leaders
The teams that win won't be the ones with the most pilots. They'll be the ones that turn pilots into systems that pay for themselves and scale across the portfolio. If you need structured learning paths that map to real roles, this directory helps cut the noise: AI courses by job.
The path is clear: automate the routine, codify the rules, keep people responsible for judgment. Ship small, learn fast, and build the stack that compounds.
Your membership also unlocks:
