HSBC
From Products to Outcomes
at Machine Speed
The convergence of AI agents and programmable financial infrastructure creates a structural shift in how value is created, priced, and delivered. The unit of design moves from the product to the accountable client outcome.
"The structure is the building, and the building is the structure."
Norman Foster
ENTER THE BUILDING
The Architectural Anchor
In 1979, Norman Foster was commissioned to design the new headquarters for the Hongkong and Shanghai Banking Corporation. The brief was simple: build the best bank building in the world. Foster's response was radical. He rejected the sealed glass tower, the conventional core, the standard floor plate. Instead, he designed a building that exposed its own mechanics, distributed light precisely through a sunscoop, and was assembled from prefabricated modules shipped from around the world.
Forty years later, the parallels to the bank's strategic challenge are structural, not decorative. The building is not a metaphor we have imposed. It is a metaphor that imposes itself.
FOSTER'S SUNSCOOP
The sunscoop captures a single beam of harbour light and refracts it to every floor. Each level receives exactly the light it needs. No uniform flood. No single spotlight. Standardised components, bespoke distribution.
THE PROGRAMMABLE BANK
Programmable rails capture a single client need and decompose it into atomic services. Each stakeholder receives exactly the outcome they see. No bundled product. No cross-subsidy. This is the Prism of Bespoke Outcomes.
PREFABRICATED MODULES
The building was assembled from prefabricated modules manufactured in the UK, Japan, and the US, then shipped and bolted together in Hong Kong. No module was bespoke. The composition was.
COMPOSABLE OUTCOMES
The outcome engine assembles from standardised, composable capabilities: payments, FX, credit, compliance, identity. No capability is bespoke. The client outcome is.
Foster's 1 QRC is literally a prism. He placed a giant mirror system at the base of the atrium that captures sunlight from the harbour and redirects it upward through the entire building. A single beam of natural light enters, hits the sunscoop, and is refracted and distributed to every floor. Every level receives exactly the light it needs. Not a uniform flood. Not a single spotlight. A precisely distributed, bespoke allocation of the same resource.
The Prism of Bespoke Outcomes = The Sunscoop.
The Structural Mismatch
THE THREE DANGEROUS PHRASES
"We have always done it this way."
"Nobody has done it, there must be a reason."
"We will do it after someone else does."
The past decade of Quantitative Easing expanded the jaws of the K by inflating asset prices faster than wages. It rewarded patience, duration, and concentration. The future, driven by AI, compresses the jaws of the K by collapsing service costs faster than pricing models can adapt. It rewards velocity, dispersion, and speed.
The old universal bank was organised like a department store. It offered payments, deposits, FX, lending, trade, securities, custody, markets and wealth under one roof, but the client still had to walk aisle by aisle. Even when the client relationship was strong, the internal logic remained product-led. That model created breadth, but it also created handoffs, delays and fragmented accountability.
CORE INSIGHT
AI is anti-size. Scale was never one advantage. It was a bundle of five: coordination, friction, opacity, inefficiency, and centralisation. AI attacks all five simultaneously. The strategic response is not to defend the department store. It is to rewire it into an outcome engine.
THE NEXT CHAPTER EXPLAINS WHY →
When Light Reaches Every Floor
Foster designed 1 QRC so that natural light reaches every floor. No dark corners. No hidden rooms. The sunscoop captures a single beam from the harbour and refracts it through the entire building. Every level is illuminated. Every structure is visible.
"The soul of money belongs neither to a big tech nor to an anonymous ledger. The soul of money is trust."
Agustin Carstens, General Manager, Bank for International Settlements (2022)
AI does the same thing to the large institution. It floods every layer with information, eliminating the shadows where rents, opacity, and coordination costs once hid. "Size" was never one advantage. It was a bundle of five structural advantages that thrived in darkness. AI attacks all five simultaneously.
Large institutions justified scale because coordinating complex, multi-party, multi-geography transactions required centralised command-and-control. Agents coordinate atomically. They do not need org charts, approval chains, or handoff protocols. The coordination equilibrium, centralised by necessity, gets disturbed and shifts toward decentralised orchestration.
What remains: Not coordination itself, but accountable coordination: the trust layer ensuring decentralised agents act within regulatory, fiduciary, and ethical bounds.
Large banks benefited from high switching costs. Moving corporate treasury, trade finance relationships, or custody arrangements was painful and risky. AI agents comparison-shop, negotiate, and switch providers in real time. An agent managing a corporate treasury has no loyalty, only optimisation functions.
What remains: Distribution moats are always under threat. Every interaction must earn trust anew, because the agent has no memory of brand loyalty, only last-transaction performance.
Large organisations maintained information advantages. Internal information friction created layers, and those layers created jobs, budgets, and power structures. AI dissolves information friction. When every employee has an agent that can access the full institutional knowledge base, the layers that existed to relay and filter information become redundant.
What remains: Rent extraction based on knowing something the client does not becomes unsustainable. The client's agent already knows it. Transparency is not a choice, it is an inevitability.
Large banks tolerated inefficiency because margins absorbed it. AI enforces efficiency across three dimensions: atomic pricing replaces bundled opacity, agents eliminate resource overhead, and balance sheet velocity (throughput per unit of capital) matters more than absolute size.
What remains: Every rent, every margin, every fee is subject to agent-driven price discovery. The only sustainable premium is the trust premium, earned per interaction.
The coordination equilibrium in financial services was centralised by default: clearing houses, correspondent banks, central counterparties. When agents verify counterparty risk in real time and programmable ledgers provide atomic settlement, the rationale for centralised coordination weakens.
What remains: The bank shifts from being the central coordinator to being the most trusted node in a distributed network. Coordination decentralises. Accountability concentrates.
The only appreciating asset is verified trust, earned per interaction, not inherited from brand.
OLD MODEL
Size = Coordination + Friction + Opacity + Inefficiency + Centralisation
NEW MODEL
Value = Accountable Trust × Balance Sheet Velocity × Network Reach
Fast Brain, Fast Body
If an AI agent can spin up a marketing campaign, scrape global supply chain data, or rewrite a codebase in seconds, it cannot wait three days for a wire transfer to clear or navigate layered compliance approvals to move capital across jurisdictions. Without a native, programmable financial layer, AI remains a very fast brain operating inside a very slow body.
For our bank, the strategic opportunity is to become the fast body: providing the programmable settlement rails that allow capital to move instantly, deterministically, and without human coordination, matching the velocity of the underlying intelligence.
AI WITHOUT RAILS
Intelligence without execution. A fast brain trapped in a slow body. The agent can think but cannot transact.
RAILS WITHOUT AI
Dumb pipes awaiting instructions that never arrive at machine speed. The infrastructure exists but nothing intelligent drives it.
AI and programmable rails are not two separate initiatives. They are symbiotic. The board must resist the organisational instinct to assign them to separate workstreams. The bank that fuses them becomes the fast body that matches the velocity of the intelligence layer. The bank that silos them will build neither.
One common mistake in the current AI debate is to assume that human expertise becomes irrelevant as models become more capable. The better reading is subtler. Isolated expertise becomes easier to replicate or approximate. What becomes scarcer is accountable synthesis across domains.
"When execution becomes abundant, value shifts to the human who knows what to build."
Eric Schmidt and Dror Berman, "When Machines Build for Machines"
The Prism of Bespoke Outcomes
One structural shift. Six strategic truths. Like Foster's sunscoop, the move to bespoke outcomes is not one thing. It is a prism.
WHAT THEY SEE
Atoms, Not Bundles
I pay only for what clears. No product tax. No cross-subsidy. Every cent maps to a service I consumed.
Clients demand granular pricing and composable outcomes. The bank that offers atomic transparency wins the relationship.
THE OUTCOME IN ACTION
A micro-enterprise in Jakarta seeking trade finance for a $5,000 shipment to Ho Chi Minh City receives an outcome assembled in real time: the credit assessment drawn from transaction history on programmable ledgers, the FX conversion executed at the optimal moment by an agent monitoring rates, the compliance checks completed against tokenised identity credentials, the settlement confirmed on shared infrastructure. The outcome is bespoke. The cost structure is that of a utility.
The Abundance Hypothesis
When programmable infrastructure drops the marginal cost of assembling a bespoke client outcome to near-zero, the strategic choice for universal banks shifts fundamentally from value-driven economics to volume-driven economics. This shift is governed by three driving factors: Volume (the scale of the network), Complexity (the ability to orchestrate multi-party outcomes atomically), and Accountability (the institutional trust required to execute).
HYPOTHESIS 1
Economics Invert
When agents operate on programmable rails, the marginal cost of assembling bespoke outcomes approaches zero. The bank must pivot from rationing bespoke outcomes to deploying them at scale across the dispersed long tail.
HYPOTHESIS 2
Pricing Power Inverts
In the old model, the bank held pricing power because structuring was scarce. In the new model, the bank holds pricing power because trust is scarce. Trust as the only appreciating asset. The long tail becomes the revenue engine.
HYPOTHESIS 3
Balance Sheet Velocity
Throughput matters more than absolute size. The bank that can clear the most outcomes per unit of capital deployed wins. Speed of capital rotation replaces size of capital base.
| DIMENSION | OLD MODEL | NEW MODEL |
|---|---|---|
| Revenue model | Product margin on bundled services | Infrastructure throughput + trust premium |
| Margin source | Opacity of bundled pricing | Volume at atomic transparency |
| Client access | High-value clients justify high-touch | Long tail at near-zero marginal cost |
| Growth ceiling | Finite pool of sophisticated clients | 3.2 billion people across four corridors |
| Competitive moat | Relationship (replicable) | Infrastructure density (compounding) |
THE SCARCITY HIERARCHY
Ideas have always been cheap. AI makes them cheaper still.
Execution, which was once hell, is becoming a durable commodity. Agents can write code, test systems, and deploy infrastructure at a fraction of the historical cost.
What remains genuinely scarce is vision: the ability to synthesise cross-domain capabilities across 60+ markets, and fuse them with emerging technologies into trust infrastructure backed by a real balance sheet.
The Growth Corridors
Four corridors represent a combined population of 3.2 billion people and approximately $14 trillion in GDP. The real prize is the six cross-corridors connecting them.
South Asia
POPULATION
0.0bn
GDP
$4.19tn GDP
DIGITAL ECONOMY
$3.5tn UPI (228bn txns)
OPPORTUNITY
$350bn MSME credit gap
CROSS-CORRIDOR
$51bn remittances from Middle East, $47bn bilateral trade with ASEAN, $125bn trade with Greater Bay Area. On programmable rails, every remittance becomes an opportunity to attach micro-credit, micro-insurance, or micro-investment.
ASEAN
POPULATION
0m
GDP
$3.8tn GDP
DIGITAL ECONOMY
$300bn digital economy
OPPORTUNITY
166m financially excluded
CROSS-CORRIDOR
Islamic finance bridges ASEAN to the Middle East. Malaysia is the global hub for sukuk issuance. Programmable rails can automate Sharia-compliant structuring. ASEAN-Greater Bay Area trade exceeds $300bn annually.
Greater Bay Area
POPULATION
0m
GDP
$2.09tn GDP
DIGITAL ECONOMY
Surpasses NY and SF Bay Areas
OPPORTUNITY
Cross-border wealth proving ground
CROSS-CORRIDOR
Sovereign wealth flows from the Middle East, with PIF, ADIA, and Mubadala actively investing. Middle Eastern borrowers raised a record $14.2bn through syndicated loans across Asia Pacific in 2025, a 175% increase.
Middle East
POPULATION
0m+
GDP
~$4tn GDP
DIGITAL ECONOMY
$5.6tn SWF assets
OPPORTUNITY
Vision 2030 digital transformation
CROSS-CORRIDOR
Centre of all four corridors. Commodity trade, sovereign wealth, Islamic finance, and labour mobility creating dense, multi-directional flows that programmable infrastructure can settle in real time.
THE PRIZE
Not the next thousand corporate mandates. The next hundred million client outcomes.
Machine-Native Capital Markets
Traditional equity markets were designed to value decades of stability, relying on quarterly earnings cycles, forward guidance, and discounted cash flow models. This architecture works when competitive landscapes shift slowly. However, when a company can launch a new vertical in days or a model upgrade can alter competitive positioning overnight, the quarterly earnings report becomes backward-looking by definition.
This is where prediction markets enter the conversation. They do not replace equity markets; they complement them. Structurally, they are better suited to pricing discrete events, pivots, product launches, regulatory outcomes, adoption curves, and execution milestones as they unfold.
EQUITY MARKETS
Price decades of stability. Quarterly earnings cycles. Discounted cash flow models. Backward-looking by definition when execution velocity accelerates.
PREDICTION MARKETS
Price discrete events in real time. Agents scrape information, detect mispricings, and execute arbitrage in milliseconds. They process probability faster than humans can update their beliefs.
If equity is transforming into a shorter-duration call option on execution velocity, prediction markets represent a parallel information layer where that velocity is continuously repriced. They are not a replacement for capital markets. They are a preview of how real-time probability pricing begins to coexist alongside traditional duration-based valuation.
THE SUNSCOOP IN CAPITAL MARKETS
If AI is anti-opacity in banking, dissolving the information asymmetry that sustained layers and rents, prediction markets are anti-opacity in capital markets. They flood the market with real-time probability pricing, replacing inherited information advantages with verified, continuous price discovery. Foster's sunscoop brought light to every floor of 1 QRC. Prediction markets bring light to every discrete event in the market. Both replace darkness with transparency. Both make hiding structurally impossible.
The Bank That Banks Banks
Our bank is a $3.2 trillion balance sheet operating across 57 markets. In 2025, the bank delivered $71 billion in revenue, a record $36.6 billion profit before tax, and a 17.2% return on tangible equity — up 160 basis points year-on-year. The stock surged 50% in 2025.
The numbers that matter for the thesis: $1.787 trillion in customer deposits growing across all four segments simultaneously. $2.1 trillion in wealth balances, with over $1 trillion booked in Asia. $9.4 billion in wealth fee income, up 24%. 275,000 new affluent clients added in a single year. $900 billion in trade transaction value facilitated. Over 90% of non-interest income generated by transaction banking and wealth — businesses that expand the fee pool without proportional growth in risk-weighted assets.
$3.2T
TOTAL ASSETS
17.2%
RETURN ON TANGIBLE EQUITY
$2.1T
WEALTH BALANCES
57
MARKETS
$71B
REVENUE
$9.4B
WEALTH FEE INCOME (+24%)
$900B
TRADE VALUE FACILITATED
50%
STOCK SURGE IN 2025
Every segment delivered mid-teens or better returns: Hong Kong at 35.5%, the UK at 22.9%, CIB at 16.2%, International Wealth and Premier Banking at 19.0%. The CEO's own words: "We are becoming a simple, more agile, focused bank, one that moves with the speed our customers need to navigate the modern world." — Georges Elhedery, Group CEO. No other international bank maintains this breadth of on-the-ground presence across the Growth Corridors.
This model answers the Board's capital efficiency question. Serving 3.2 billion people across four corridors does not require deploying $3.2 trillion in capital. It requires building the programmable trust infrastructure and extending it through regulated partner-client banks. The infrastructure scales. The capital does not need to.
HSBC TDS
Real-time tokenised fund movement across HSBC locations. Demonstrates programmability in a bounded domain. A discovery wedge, not a blueprint.
HSBC ORION
Digitally native bonds in Hong Kong and Luxembourg. Full lifecycle from issuance to redemption. A learning engine for scaled architecture.
What HSBC Should Do Next
Fuse AI and programmable rails into a single programme. Do not silo them into separate workstreams.
Appoint an Outcomes Head for each corridor. The unit of accountability shifts from the product to the client outcome.
Redesign pricing from bundled opacity to atomic transparency plus trust premium.
Treat TDS and Orion as discovery wedges and learning engines, not final blueprints.
Build the programmable trust infrastructure for partner-client banks. The bank that banks banks.
Provide institutional access to prediction markets through Global Markets. Use them for precise hedging through Private Wealth.
THREE QUESTIONS FOR THE BOARD
Are we defending how we have always done it?
Are we waiting for proof that someone else can do it?
Are we planning to follow after someone else leads?
"The accountability of everything we do remains with humans."
Georges Elhedery, CEO, HSBC
Download the Full Horizon PaperCONCEIVED, CREATED AND AUTHORED BY
Rajeev Tummala
The Sage and the Magician
He does not forecast the future. He refracts it — a single beam of conviction into a prism of bespoke outcomes — and conducts the agents that deliver them at machine speed.
Imagination > Knowledge