Why 94% of Banks See No Value From AI — And the Boardroom Equation That Fixes It

Banks are running the most expensive technology rollout in financial history. Productivity is going the wrong way. Here is the framework that fixes it.

Last month, three Vietnamese banks asked me the same question in three different boardrooms: where is the money in our AI program?

I do not blame them for asking. McKinsey's Global Banking Annual Review 2025 found that 94% of companies report seeing no significant value from their AI investments. The MIT Media Lab puts the number at 95% for generative AI specifically. And in banking, US productivity has actually fallen 0.3% per year since 2010 — at the same time global banking technology spending grew 9% annually.

Tech spending up nine percent every year. Productivity down. The largest peacetime technology investment in the history of finance has gone backwards.

This article walks through three brutal truths about banking AI ROI, then introduces the Boardroom Equation — the framework I use with banking executives to make every AI dollar auditable against a specific line on the P&L. If you are responsible for AI strategy at a bank — CEO, CFO, COO, CDO, CIO, or Head of Strategy — this is for you.

The state of banking AI ROI in 2026

Four data points worth pinning to your wall.

• The value pool is enormous. McKinsey estimates AI could add $200–340 billion in annual value to global banking — 9 to 15% of operating profits. BCG's November 2025 analysis goes further: retail banks alone could unlock $370 billion in additional annual profit by 2030, with AI agents as the primary driver.
• Capture is concentrated. Early-adopter banks could capture a 4-percentage-point increase in Return on Tangible Equity. Slow movers face what McKinsey calls "irreversible decline."
• Measurement is broken. BCG's 2025 study of 280 finance leaders found only 45% could quantify their AI ROI at all. Of those who could, one-third reported sub-5% returns; the median came in at 10% — half the 20% threshold most CFOs target.
• Even the success stories disclaim their measurement. JPMorgan's CFO told McKinsey in December 2025 that proving AI cost savings is "a very hard thing to do."

Truth #1 — Productivity is not profit

The first brutal truth most boards refuse to internalize: productivity is not profit. When AI helps a credit specialist finish reports 10% faster, that saved time does not automatically become revenue — the employee simply has more free time. Real profit appears only when the workflow is restructured around the new capacity.

Why the gap persists

Most banks deploy AI onto legacy workflows and call the result transformation. There are three structural reasons banks fall into this trap:

• AI is deployed on legacy workflows. Adding a copilot to a back-office process does not redesign the process — it makes existing inefficiency faster.
• Productivity gains are not measured back to the P&L. A 22% faster cycle time is a metric, not a cash flow.
• Time saved is not reallocated. Freed hours fill with low-value internal work instead of revenue activity.

McKinsey's contrast is blunt: banks that simply add AI to existing workflows capture about 5% productivity gains; banks that redesign their operating model around AI capture 15 to 20%. Three to four times the return for the same AI investment.

Truth #2 — Where AI return actually lives

The second brutal truth: AI return concentrates in three narrow zones. Everything outside these zones is, statistically, wasted spend.

The pattern is consistent across all three: high-volume repetitive work, compliance-driven outputs, and results that map directly to the income statement. If your AI portfolio does not concentrate here, you are probably in the 94% seeing no significant value — and you now know what to defund.

Truth #3 — Operating model redesign is the multiplier

The third brutal truth: knowing where the value lives is not enough. Operating model redesign is what multiplies the return. Same AI investment, three to four times the return — depending entirely on whether the bank had the courage to redesign the underlying process.

There is a counter-position every executive must internalize: pure AI cost-savings gains erode over time, because competition forces them to pass to customers, not shareholders. AI is a defensive necessity, not an offensive moat. The advantage is not whether you adopt — it is how you measure and deploy. Strict governance (model risk management, audit-grade documentation, regulator-ready evidence) is your competitive weapon, not your compliance burden.

The Boardroom Equation — making AI ROI auditable

The three truths converge on one financial equation:

The equation forces every AI initiative to trace back to a specific line on the financial statements. Auditable. Defensible at board level. Repeatable next quarter.

The three pillars

• Pillar 1 · Avoided hiring — planned headcount × fully-loaded cost per role, for roles AI removed from the hiring plan. Not layoffs: the absence of new headcount that would otherwise have been added. The most concrete and board-defensible pillar, because the hiring plan is a known, dated artifact.
• Pillar 2 · Vendor consolidation — legacy CRM, workflow-automation, document-processing, and research contracts renegotiated or terminated once an AI-augmented internal team delivers the same work. Each is a direct reduction on operating expense. BCG 2025 found top teams require vendor reductions to be proven before continued funding.
• Pillar 3 · Protected risk-adjusted cash flow — fraud and AML losses prevented, fines avoided, analyst headcount not added as alert volume grew. The hardest to measure, the most valuable; referenced to an independent methodology (Forrester TEI on HSBC + Quantexa).

Gartner's June 2025 maturity survey reinforces the discipline: 91% of high-maturity organizations have appointed dedicated AI leaders, and 63% run formal ROI analysis on every initiative. Discipline is the differentiator.

Vietnamese banking benchmarks

Vietnamese banks do not need to wait for global benchmarks. They have their own.

• Techcombank intends to become "the first fully AI-operated bank in Vietnam." Sixteen percent of staff specialize in data science or AI; the AI analytics platform contributes 100 billion VND in operating income directly; the eKYC platform processes 400,000 transactions a day at 99% success against deepfakes (The Asian Banker, 2025).
• Vietcombank's VCB Digibot handled 88.5% of customer requests in its first six months — roughly 2 million successful interactions — replacing planned hiring in customer service.
• BIDV's private-banking AUM grew 37% to 300 trillion VND in 2024, driven by an AI-supported digital wealth advisory platform.

These cases map cleanly to the three zones: Risk Intelligence (Vietcombank), Knowledge Operations (Techcombank), Decision Layer (BIDV).

The risks worth flagging

Algorithmic coupling. A University of Tokyo 2026 study of 809 US financial institutions found measurable systemic risk: when many banks deploy the same AI models, a single failure can trigger correlated shocks. Expect Vietnam's State Bank to require model-diversity disclosures.

Cost erosion. Pure AI cost-reduction gains pass to customers over time. Treat AI as a defensive necessity that maintains margin parity, not an offensive moat. Vietnam's AI Law 134/2025 takes effect March 2026 — the first comprehensive AI law in Southeast Asia. Governance is no longer optional.

What to do this week

Audit your AI use-case list against the three zones. Separate your AI budget from the IT line. And tag every initiative to one pillar before funding — with a twelve-month measurement plan written into the business case. When the numbers do not move, leaders should defund the spend or redesign the workflow before adding budget. This is mathematics that cannot be challenged: a clear positive number, or redesign.

Independent thought leadership · not affiliated with any current or past employer · compliant with Vietnam AI Law 134/2025 + PDPL.