Ten Structural Shifts in AI Adoption: Evidence, Frameworks, and Emerging Frontiers

Most investment firms now have cloud, models, and AI talent. Yet firm‑wide, durable impact is still rare.

Over 2018–2025, I’ve used structural frameworks to track how AI really scales in finance. My latest paper, Ten Structural Shifts in AI Adoption (2018–2025), suggests that what used to be “theory” is now visible in the data and in practice.

Here are three structural traps I see most often:

1️⃣ The real bottleneck is no longer cost — it’s time and leadership.

The FinTech Pyramid predicted that once cost and talent shortages eased, Leadership Vision + Time would become the binding constraints. Latest data shows that inference has become dramatically cheaper and that employees are often more ready for AI than their executives.

High‑stakes question:
Are your board and ExCo treating AI as a strategic, governed priority with clear timelines—or as an IT line item? Every quarter of delay creates a compounding competitive gap in learning, data, and client outcomes.

2️⃣ Siloed AI teams underperform — you need T‑shaped, cross‑functional teams

Many firms built AI labs or “Centers of Excellence.” They produced pilots, but struggled with adoption, accountability, and risk.

The T‑Shaped Teams model argues that durable value comes from cross‑functional squads: deep domain owners (PMs, product, risk) working with data, AI, engineering, and compliance around a shared P&L / risk / client objective.

Consulting data suggests ~70% of scaling obstacles are people and process, not model performance. Governance, incentives, and ownership matter more than another algorithm.

High‑stakes question:
Do you have empowered, cross‑functional teams that own specific outcomes end‑to‑end—or isolated experts handing off models into the void?

3️⃣ AI is seniority‑biased — it automates basic tasks first

The emerging pattern is that AI automates routine, junior tasks (collection, basic analysis, templated reporting), while making senior judgment, oversight, and client trust more valuable.

High‑stakes questions:
Are you redesigning analyst and associate roles so they still build judgment, not just monitor systems?
Are your senior leaders accountable and literate enough to oversee AI‑enabled workflows in front of clients, boards, and regulators?
Without intentional redesign, you risk hollowing out apprenticeship and weakening future leadership.

What’s coming next

Over the next few weeks, I’ll share a short roadmap series on:
1. T‑Shaped Operating Models: Beyond the AI lab; how to build integrated teams that own outcomes.
2. The Junior Talent Crisis: Starting a career when AI does the entry‑level work.
3. The Leadership Constraint: Governance and vision as the real bottlenecks to scale.
4. Beyond LLMs: Preparing for AI’s next evolution and regulatory expectations.

👉 If you want the full series (with frameworks and practical templates), subscribe to my AI in Finance newsletter: larrycao.substack.com.

For the complete list of AI trends in finance, read my SSRN article Ten Structural Shifts in AI Adoption (2018–2025).