When Andrew Ng speaks, the AI world listens. He co-founded Google Brain, built Coursera into a platform serving 120 million learners, led AI at Baidu with 1,300 researchers, trained more than 8 million students, and now manages a $370 million AI fund. His track record gives him unique credibility.
Recently, Ng told his followers that the next big AI wave will not come from ever-larger models, but from Agentic AI—autonomous systems that can plan, reflect, critique, and improve their own work. He predicts the market will grow 13×, from $5.1 billion today to $69 billion by 2032, making it one of the most important technologies of the decade.
For financial services, where precision and trust are paramount, the implications are transformative.
What Is Agentic AI?
Traditional AI works like a calculator: ask a question, get one answer. Agentic AI, by contrast, behaves more like a team of analysts working through a problem. It relies on four key design patterns
Reflection – Critiques and revises its own outputs.
Tool Use – Connects to APIs, databases, and trading systems.
Planning – Breaks complex processes into smaller steps.
Multi-Agent Collaboration – Coordinates like a project team.
This makes workflows iterative and adaptive: outlining, researching, drafting, critiquing, and refining. In financial services, where mistakes can cost billions, this reliability is a breakthrough.
Pull Quote
“Agentic AI acts more like a team of analysts than a calculator.”
JPMorgan: A Case Study in Agentic AI at Scale
No financial institution has leaned into AI harder than JPMorgan. In 2025, it allocated $18 billion to technology, with agentic AI central to that spend.
The results:
30% lower servicing costs in consumer banking, with headcount reduced by 10% through automation.
83% faster research using Smart Monitor, which surfaces insights for traders in seconds.
3.4× boost in advisory productivity with Connect Coach giving real-time prompts to relationship managers.
40% lower onboarding costs in commercial banking.
90% less manual treasury work for over 2,500 clients using Cash Flow Intelligence.
Today, more than 200,000 JPMorgan employees use the bank’s generative AI platform, which has 100 tools in development and 175 use cases in production. The bank estimates $1.5 billion in cost savings and, during the April 2025 market turmoil, a 20% surge in asset and wealth management sales.
For a sector often accused of moving slowly, JPMorgan has proven that agentic AI isn’t hypothetical—it’s already a competitive advantage.
Why Smaller Models Are the Big Story
Ng’s second insight reinforces this: the real breakthroughs won’t come from massive, costly models, but from Small Language Models (SLMs)—compact, specialized systems built for specific industries.
The SLM market is projected to grow from $930 million in 2025 to $5.45 billion by 2032 (CAGR ~29%).
Token costs have dropped 90% since 2023, making SLMs cost-effective.
Edge computing investment is expected to hit $378 billion by 2028, enabling on-device AI.
For banks, this means trusted, explainable, domain-specific AI that regulators can scrutinize while still delivering speed and efficiency. Models like Phi-4, Llama 3.2, and Qwen 2.5 are already proving that smaller can be smarter.
Projections: What’s Coming Next
Industry forecasts point to exponential growth:
Agentic AI market: $5.1B in 2024 → $69B by 2032 (Ng).
Broader market: $7.4B in 2025 → $171B by 2034 (CAGR ~42%).
Finance-specific market: $490M in 2024 → $4.49B by 2030 (CAGR ~45%).
Enterprise adoption: Deloitte projects 25% of enterprises will run agentic pilots by 2025, 50% by 2027.
Scale: Barclays estimates 1.5 to 22 billion AI agents could be operating worldwide within a decade.
The trajectory is unmistakable: financial services will be one of the earliest and biggest beneficiaries.
Trust: The Real Moat
While speed and savings matter, Ng stresses that trust is the ultimate competitive advantage.
In finance, explainability, compliance, and auditability are non-negotiable. The winners won’t just deploy fast—they’ll deploy responsibly. JPMorgan’s Cash Flow Intelligence isn’t just a cost saver—it strengthens trust by providing transparent, data-driven recommendations.
The lesson: agentic AI will only scale in finance if trust is built into the system.
Pull Quote
“Trust is the real moat separating leaders from the rest.”
Leadership Checklist: Preparing for Agentic AI
For executives in BFSI, here are five questions to bring back to your teams:
Do we have clear budget allocation for AI, or is it buried in IT spend?
Are we piloting agentic workflows beyond chatbots—research, onboarding, compliance?
What governance frameworks ensure AI outputs are explainable and auditable?
How are we training teams to work effectively with AI copilots and agents?
Are we positioned to adopt Small Language Models for targeted, compliant use cases?
Finance’s Agentic Future
Andrew Ng predicts agentic AI will create more millionaires than any wave before. For financial services, the opportunity is not theoretical—it’s unfolding now.
JPMorgan’s $18B bet shows the payoff: billions saved, productivity multiplied, and stronger client engagement. Small Language Models promise to make these systems faster, cheaper, and more compliant. Industry forecasts show adoption will only accelerate.
The next great AI opportunity in finance isn’t abstract “general intelligence.” It’s targeted, trusted agentic systems solving real business problems at scale.
For BFSI leaders, the choice is clear: build with AI agents now—or risk being left behind in the age of intelligent workflows.
Continue reading in the FinScale Magazine
This insight was originally published in the first issue of FinScale Magazine by TrialScale. Download the magazine to keep reading.