Five AI Trends Transforming Financial Services in 2025: From Reasoning Models to Autonomous Agents
AI in finance is moving from chat to reasoning agents, multimodal tools, and secure integrations. BBVA shows progress with no-code apps, pilots, and people-first adoption.
AI: Five Trends Already Changing the Financial Industry
AI adoption in finance is moving fast and getting practical. As Elena Alfaro, head of global AI adoption at BBVA, put it: "Artificial intelligence is constantly evolving, with impact across various sectors; anyone who feels it's moving incredibly fast is not mistaken-this is a full-blown revolution."
Her data point is hard to ignore. ChatGPT hit 100 million users in two months. Today, OpenAI is reportedly at 700 million weekly active users, with around 20 million paying users-and when you add major competitors, we may already be past 1 billion generative AI users worldwide.
Capital, Costs, and Jobs: The Signal Is Clear
Costs to run and apply AI have dropped since its launch three years ago. Demand is surging. In the U.S., AI-related IT job postings jumped 448%, while non-AI IT roles fell 9%. Meanwhile, the "Big Six" increased capex by 63% from 2023 to 2024, reaching $212 billion.
Profitability is uneven. NVIDIA benefits from GPU demand. Accenture's services line is performing. OpenAI is expected to reach $12 billion in revenue by the end of 2025. Even so, the largest firms continue to invest for the long term.
Five AI Trends Finance Leaders Must Track Now
1) Reasoning Models Are Getting Smarter
We're moving from bots that answer to systems that reason through tasks step by step. Alfaro highlighted models that can break down problems, follow logic, and flag doubts (examples cited: GPT-5, Gemini 2.5 Pro, Claude 3.7-4) versus faster, non-reasoning bots (GPT-4, Gemini 1, Grok 2, Deepseek Base).
For finance, this matters for risk analysis, policy interpretation, exception handling, and auditability. Better reasoning means fewer handoffs and more reliable automation of complex workflows.
2) Multimodality
AI is no longer just text. It now works across images, video, voice, music, or combinations. That unlocks document intake, KYC/AML checks with visual data, voice-first client service, and richer research summaries from multi-format inputs.
3) From Assistants to Agents
The shift is from prompt-response to goal-driven execution. Alfaro expects each person could have an "AI Chief of Staff" coordinating multiple tools and agents to complete multi-step tasks.
Think of end-to-end processes: pulling data, filling forms, drafting memos, executing in systems, and summarizing results-with human review at key control points.
4) Integration of Data and Tools
The unlock comes when AI connects to internal systems like Salesforce, Outlook, Drive, and core banking platforms. "Security and compliance teams will play a fundamental role in this integration."
With the right permissions, logging, and monitoring, teams can automate retrieval, reconciliation, and updates across apps without copying data into risky ad hoc workflows.
5) Growth of No-Code
No-code orchestration is maturing. Alfaro cited Google Flows as an example of chaining steps and tools without writing code. This lets analysts and operators ship prototypes quickly, then harden what works with engineering support.
What BBVA Is Building
BBVA's mobile app Futura adapts to each user, surfaces frequent actions, and provides shortcuts. It includes Blue, a chatbot that answers product and personal finance questions with context.
Inside the bank, a broad adoption program launched last May focused on people first: "This is a people project, not just a technology deployment project." The initiative has near 90% retention, over 5,000 functional apps built by non-coders, and about 1,000 high-value use cases moving into production using solutions from OpenAI and Google.
Practical Playbook for Finance Teams
- Split work into "AI does" vs "human does." Use AI for drafting, reconciliations, data pulls, and monitoring; use humans for judgment, prioritization, approvals, and client nuance.
- Start where variance is low and volume is high: account updates, policy Q&A, portfolio notes, call summaries, RFP scaffolds, risk alerts, and internal research briefs.
- Pilot reasoning models on complex cases with clear evaluation criteria. Compare quality, time saved, and error rates versus your baseline.
- Integrate securely. Connect AI to your CRM, document stores, ticketing, and comms under role-based access, audit logs, DLP, and model risk controls.
- Test agents on bounded workflows: ingest → analyze → draft → route for approval → execute. Add human checkpoints and monitor drift.
- Enable no-code builders. Provide templates, office hours, and a review board. Keep what works, retire what doesn't, and standardize patterns that scale.
Useful references
Career Upside: From Executor to Orchestrator
Alfaro's forecast is direct: humans aren't out-they're moving from task executors to task orchestrators. "We should always analyze tasks from the perspective of what makes sense for AI to do and what makes sense for a human to do."
Skills like organization, prioritization, and training still require strong human input. Roles will evolve, and new ones will appear. Continuous training is the edge.
Next steps
- Want structured upskilling paths? Explore finance-focused tools and courses: AI tools for finance and courses by job.
Better to be in than out. Start small, run controlled pilots, and compound what works.
