I just published a case study on Bank of America’s AI strategy. Bank of America has been quiet about AI since it launched Erica back in 2018, but the bank opened up to us recently in a series of interviews with its executives, including Hari Gopalkrishnan, head of consumer, business, and wealth management technology, and Nikki Katz, head of digital. Here are some highlights: 

AI Is About The Basement, Not Just The Branch 

When Bank of America’s leaders talked about their approach to AI, they didn’t open with generative razzle‑dazzle; they talked about it as a foundation. “AI isn’t a project at the bank. It’s the infrastructure layer that powers everything.” That mindset — patient, data‑first, and deeply operational — explains how the bank lifted assets under management 90% while holding its workforce flat over seven years. 

But this kind of AI strategy takes preparation. Years before Erica debuted for clients, the bank had already aggregated the lion’s share of its customer interaction and transactional data, giving digital analytics and AI experts a clean runway for insight discovery and model training. Hari Gopalkrishnan’s pointed advice to peers? “Get started — don’t wait for data perfection.” 

Erica Is The Cornerstone 

Erica is unquestionably the cornerstone of their AI foundation. Launched in 2018, it has shouldered 2.7 billion client interactions at a 98% containment rate. Freed from routine calls, the bank could pivot many employees to higher‑value work, while the same natural‑language engine powers “Ask Merrill” and “Ask Private Bank.” The principle is simple: invest once, reuse everywhere — discipline that keeps budgets in check while building on the past to deliver the future. 

That mindset also shows up in how the bank continuously invests. Nikki Katz told us: “AI is never finished. You launch, and that’s when the real work begins.” Product teams have pushed continuous waves of algorithmic updates since launch, turning Erica into an enterprise platform, not a frozen product. Those evolutions protect the assistant from the one‑and‑done fate that sinks many experiments. 

How AI Turns Data Into Dollars 

Bank of America’s AI engine hums along a Data → Triggers → Insights → Treatments → AI loop. That flywheel enrolled 2 million new Preferred Rewards members and lifted digital sales 16%. Each interaction feeds new data back into the loop, sharpening the next recommendation and compounding value. 

Reuse Accelerates The Future 

The bank’s mantra for AI investment is reuse: The first project in a portfolio “pays the freight charge” — the foundational heavy lifting, such as enterprise‑grade data pipelines, a governed feature store, and reusable APIs. Every follow‑on effort moves faster at lower costs, like Ask Merrill and Ask Private Bank. 

The lesson? Treat foundational work as a shared asset, not a cost center. Reuse turns what looks like overhead into a compounding advantage, but it takes time and discipline. 

Three Takeaways For Banking (And Everybody) 

Our deep dive into Bank of America’s AI blueprint reinforced three key learnings for financial services and, as it turns out, everybody else: 

  • Unify your data before you over‑rotate on AI. Bank of America’s early investment in a clean, connected data layer made every downstream AI move easier. 
  • Boost AI’s value with insight. Raw data fuels models, but an insight layer multiplies return. 
  • Reinvest talent into the next wave. Your people, not the algorithms, remain your greatest asset. Bank of America didn’t trim headcount; it redeployed staff freed from rote tasks into higher‑value, AI‑building roles. 

What To Do Next 

Clients: Read the full case study for a deeper look at how Bank of America delivers value by unifying data, insights, and AI. 

Prospects: Contact Forrester to set up a proof‑of‑value session with me, and I’ll share more insights on aligning your data and AI strategy with value.