What Will Happen to Financial Sector AI Budgets in 2026?
AI spend across banking and capital markets is set to rise. Research from Capgemini indicates the share of total business budget dedicated to AI will move from 3% in 2025 to 5% in 2026, based on a survey of executive leaders in more than 1,500 organizations. The message is simple: the cost of stalling is now higher than the cost of moving.
Two-thirds of companies say lagging AI scale-up puts strategy and competitiveness at risk. For finance leaders, that's a budget signal, not a headline.
From "whether" to "how": leadership is the swing factor
The debate has shifted. It's less about proving AI works and more about building a system that makes it dependable at enterprise scale. Capgemini's Pascal Brier frames it around foundations-data, governance, and the way people and AI mesh-while emphasizing leadership readiness as the deciding factor.
Most teams have pilots. What separates results from noise in 2026 is a clear vision, ownership at the top, and operating rules everyone can follow.
Where the money will go
- Infrastructure: model-serving, observability, and integration with core systems.
- Data and governance: lineage, quality, access control, and model risk oversight.
- Workforce upskilling: frontline adoption, policy compliance, and prompt/reporting standards.
- Data sovereignty: keeping sensitive and regulated data under firm control is now a priority for more than half of organizations.
The top KPI is still the top KPI: return on investment. Capgemini notes 73% of respondents cite business value as the primary measure.
The KPI stack is changing
Cost takeout and efficiency still matter, but they're not the whole story. Finance leaders are adding metrics that connect directly to growth and resilience.
- Revenue growth: conversion, cross-sell, wallet share, and RMs augmented by AI assistants.
- Risk and compliance: fewer false positives, faster investigations, better audit trails.
- Knowledge management: retrieval quality, time-to-answer, and fewer repeat queries.
- Customer experience and personalization: response times, resolution rates, and NPS gains.
What leading banks are signaling
Goldman Sachs is treating this as a multi-year program. CEO David Solomon put the scoreboard in plain view: "This is a multi-year effort that we will build over time, and we plan to measure our progress across six goals: enhancing client experience, improving profitability, driving productivity and efficiency, strengthening resilience and capacity to scale, enriching the employee experience and bolstering risk management."
At JPMorgan Chase, Terah Lyons notes meaningful progress in detection quality and customer friction: "[AI has] significantly reduced false positives - there are fewer times our customers are having to make calls to try to make corrections. It supports enhanced security, cybersecurity, and fraud security for the firm." She also highlights the benefits of public-sector adoption for system stability and risk oversight.
Red Hat's Richard Harmon underscores the ongoing constraint in financial services: balance sophisticated models with regulatory requirements and explainability. Citi Taiwan's Christie Chang has emphasized that AI is materially changing how the bank operates.
Risk, compliance, and explainability: make it operational
The industry has learned that "good enough" AI doesn't pass audits. You need a clear, testable process from idea to production to monitoring.
- Model inventory and ownership: every model registered, risk-rated, and assigned.
- Controls: bias testing, stability tests, data provenance, adversarial checks.
- Explainability: consistent summaries for regulators, customers, and internal users.
- Policies: human-in-the-loop where required, escalation paths, incident response.
- Third-party risk: contracts that cover data residency, IP, and retraining rights.
If you need a reference framework for governance, the NIST AI Risk Management Framework is a useful starting point. View the NIST AI RMF.
Data sovereignty is now a design choice
AI plans must reflect where data lives and who can touch it. That means deciding between on-prem, VPC, and sovereign cloud setups, and keeping sensitive workloads ring-fenced.
Build for localization, residency, and auditability from day one-retrofitting later is where costs spike.
Your 2026 execution plan
- Q1: Set the scoreboard. Lock budget lines for infra, data, and people. Pick 3-5 use cases with clear KPIs (e.g., false positives down 30%, RM time-to-insight down 40%). Assign P&L owners.
- Q2: Build the factory. Stand up MLOps/LMMOps pipelines, monitoring, and model risk checkpoints. Implement data contracts, lineage, and access policies. Draft an AI policy staff can actually follow.
- Q3: Scale responsibly. Expand from pilots to at least two business units. Add role-based copilots for RMs, underwriters, and operations. Run red-team exercises and user acceptance testing.
- Q4: Prove value. Audit outcomes, publish ROI, and reallocate spend to the highest-yield use cases. Refresh controls and retrain models with new data.
Practical budget split (signal, not rule)
- 30-40%: data, governance, security, and sovereignty
- 25-35%: model lifecycle (training, fine-tuning, evaluation, monitoring)
- 15-25%: integration and change management
- 10-20%: workforce upskilling and enablement
Keep a small reserve (~5%) for fast experiments tied to clear business owners. No owner, no spend.
Upskilling that moves the needle
Focus training on everyday work: RMs using AI to prep meetings, risk teams reviewing model summaries, operations handling complex cases faster. Tie training to measured outcomes and compliance rules.
If you're building a structured plan for finance teams, this curated list is a helpful starting point: AI tools for finance.
Bottom line
Budgets are rising because the math supports it: targeted AI delivers measurable value, and delay costs more. The differentiator in 2026 won't be pilots-it will be leadership clarity, disciplined governance, and proof of ROI.
To go deeper on the research behind these trends, see the Capgemini Research Institute. Explore Capgemini research.
Your membership also unlocks:
