AI Strategy Endava: What Causes AI Adoption Gaps in the Finance Sector
Financial institutions say they're ready for agentic AI, yet most aren't funding it. In Endava's survey of 1,000 senior leaders across nine countries, 92% report readiness to deploy autonomous AI systems, but only 36% have budgets in place.
Leaders aren't questioning the technology. They're calling out missing infrastructure, unclear governance, and a shortage of operating capability to run AI at scale. As one executive put it: "Agentic AI represents a step-change in how financial services organisations operate and innovate. The opportunity is clear, but so is the responsibility."
The confidence-budget gap
There's a clear signal: ambition outpaces execution. The hesitation doesn't come from model accuracy or potential use cases; it comes from production reality-security, data residency, monitoring, auditability, and accountable decisioning.
Until those parts are funded and owned, CFOs won't move big numbers. That's rational. Autonomous systems without guardrails create operational, legal, and reputational exposure.
Agile is creating bottlenecks for autonomous systems
Three quarters of leaders say agile practices slow down AI at scale. Sprints are fixed windows. Autonomous systems learn, adapt, and operate on different clocks.
Two friction points top the list: regulatory complexity and integration across ecosystems (49% cite each). An autonomous system moving across borders, linking to multiple third-party platforms, and meeting AML/KYC variations doesn't fit neatly into a two-week cycle. Compliance paths differ by jurisdiction and require processes agile doesn't naturally support.
Still, 86% say agile remains useful-just not everywhere. The job now is to decide where agile fits and where an AI-native approach is needed.
AI-native operating models are coming fast
76% expect to implement AI-native operating models within two to three years, and 94% say this will influence competitive standing across the sector. Yet only 16% call themselves AI-native today.
Adoption by market: Italy 25%, US 24%, UK 16%, France 6%, UAE 4%. Most firms are still operating with legacy patterns while preparing to move.
AI-native means you design your operating structure around autonomous systems-decision rights, controls, tooling, and cadence-rather than forcing models to fit a sprint board.
Where leaders expect value
More than 80% believe agentic AI will open new markets or create additional revenue streams. Near-term focus is clear: fraud detection, financial crime prevention, and operational continuity.
Banks lose billions to fraud, and AML/KYC rules keep tightening. Even brief outages damage trust. Leaders expect 24/7 continuity, faster releases, and more personalized interactions. Routine tasks shift to machines; staff move to work that needs judgment. Part of this is defense-digital banks move fast while incumbents carry older systems and stricter oversight.
Top concerns: privacy, regulation, and explainability
Explainability sits at the center. If a model blocks a payment, flags an account, or scores a loan, customers and regulators want to know why. That gets harder as systems learn and adapt on their own.
Cross-border privacy makes it tougher, too. Surveyed actions reflect this reality: 47% build ethical guidelines into AI development; 47% implement transparency and explainability measures; 46% strengthen data privacy; 44% set governance frameworks; 42% align with regulatory requirements; 37% train employees on responsible use. Governance failures can flip AI from asset to liability.
"Our research shows that those who build AI-native operating models, backed by strong governance, will be the ones to lead the next era of financial services. At Endava, we're already adopting this approach with Dava.Flow, our AI-enabled engagement lifecycle. We know that success lies in adapting quickly, embracing multidisciplinary teams and balancing innovation with organisational health."
What finance leaders can do now
- Lock the first three use cases: fraud, financial crime, and continuity. Attach a risk rating and target KPIs (loss reduction, false positives, uptime).
- Stand up governance that can scale: model registry, lineage, policy-as-code, real-time monitoring, and incident playbooks. Consider aligning with the NIST AI Risk Management Framework.
- Fix data foundations: consent tagging, retention, residency, access controls. Centralize features and vault PII. Automate data quality checks.
- Shift delivery: keep agile for app changes; run AI on an always-on operations model with SLAs, shadow modes, and staged autonomy thresholds.
- Create a cross-functional "AI control tower": risk, compliance, model ops, platform, and product meeting weekly on metrics and escalations.
- Budget the plumbing: model observability, vector/search storage, event-driven integration, guardrails (policy checks, red-teaming), and human-in-the-loop tooling.
Where agile still fits-and where it doesn't
- Use agile for: UI tweaks, API wrappers, data pipelines, policy automation, integration stubs, and A/B testing.
- Use AI-native ops for: model orchestration, autonomy thresholds and kill-switches, human review workflows, decision logging, real-time explainability, and rollback plans.
Metrics that matter
- Fraud time-to-detect; loss per $1,000 processed; false positive/negative rates
- Decision explainability coverage; audit readiness SLA; policy drift incidents
- Model downtime minutes; mean time to rollback; model drift lead time
- Cost per 1,000 decisions; data quality defect rate; privacy incident count
Regulatory context worth tracking
- Global AML/KYC standards and guidance from the FATF Recommendations
Skill up the team
If your team is building capability around AI governance, model operations, and finance use cases, you may find this resource helpful: AI tools for finance.
The bottom line
Confidence is high. Budgets are waiting on infrastructure, governance, and accountable operations. The firms that move now-AI-native models, clear controls, and measurable outcomes-will set the pace for the sector.
Your membership also unlocks:
