Fireside Chat: Brian Moynihan

Session Focus

The Forum’s closing conversation moved outside healthcare to the macroeconomic environment: inflation and rate policy, the state of the consumer, AI adoption inside a large bank, and the capital markets environment facing the companies in the room.

Rates and Inflation

Moynihan described the Federal Reserve working to bring inflation in line with its 2 percent target, noting you want some inflation but not too much. Post-COVID dynamics made the real economy hard to read because of the cash that entered the system, and the Fed got behind it.

His expectation: rates rise a couple more times from the recent hike, after which inflation works its way down through 2027 into 2028 toward 2 percent.

He pushed back firmly on the assumption that rates then collapse. The unusual period in American history was the last 15 years since the financial crisis. Before that, a Fed funds rate of three and a half to 4 percent and a 10-year yield of four and a half percent were normal. The economy growing at 2 percent plus can withstand a higher rate environment and still grow.

He framed that as not a bad thing, given the years spent ensuring the economy avoided the disinflation that has been problematic in Japan.

On where tightening bites: small businesses borrowing on lines of credit, medium-sized businesses at market rates, and home mortgages.

He also identified an inflation channel people underestimate. Gas prices are visible at the pump, but petroleum products sit in every piece of plastic and furniture, so increases flow through much of daily life.

Asked whether a Middle East resolution is needed to control inflation, he separated the issues. Inflation has been controlled at $100 oil before, and the world economy can grow at 3 percent with oil at that level provided supply is known to be flowing. The current distortions in diesel stem from refining bottlenecks. A resolution is needed for a host of reasons, but technically $100 oil is not itself an inflation problem.

The Say-Do Paradox

On the consumer, Moynihan drew a sharp distinction between sentiment and behavior.

Bank of America data showed consumers spending 5 percent more in September 2026 than September 2025, with gas price impact accounting for less than half a percent, making the rest pure spending growth. The spending is on cruise bookings, out-of-home entertainment and somewhat more at restaurants.

His phrase for the gap: a say-do paradox. Networks report inflation and consumer stress daily, which reflects how people feel, but behavior has not changed. He was careful about the timeframe, saying he was describing September rather than predicting October.

He reported a reversal worth noting. Dividing 70 million customers into income thirds, lower income cohorts’ paychecks are now growing faster than middle and higher incomes. That was true after COVID in 2021 and 2022, then reversed, and has now returned, with paycheck growth in the mid to high threes in August.

He acknowledged the affordability problem is real, particularly for the lower third, since gas costs the same regardless of income. But behavior has not yet adjusted, and he noted that while demand stays strong, prices are harder to bring under control because sellers can sustain them.

AI Inside a Bank

Picker raised a product release that had caused bank stocks to fall the previous day on concerns an AI agent could direct consumers to move deposits to higher-yielding products.

Moynihan was unworried about deposits specifically. The ability to move money electronically has existed for years. People keep transactional balances because they use that money to conduct daily life, and moving it risks bounced checks and failed payments.

What does concern him is controlling the customer interface, and agentic commerce raises the question of whether the bank is involved in the payments.

His discussion of Erica, the bank’s assistant used by 20 million people, was the most instructive part for a healthcare audience. It has been deployed nearly 10 years and expanded from roughly 210 prompts to 700, deliberately slowly, because it has to be right every time.

His explanation of why: a single Erica query touches 110 systems and every account a customer holds. In a business handling someone’s financial life, as in a room full of people handling human life, you have to be perfectly right. You do not get to be right 95 percent of the time, because the customer loses trust, stops using it, and the next person stops too.

He separated that operational complexity from safety, security and cybersecurity, which are distinct problems.

On scale of investment: 150 AI use cases applied, $400 million invested against $800 million in benefits, with investment doubling next year. All 200,000-plus employees have unrestricted access.

His caution about consumer deployment covered the infrastructure requirements. Data has to be perfect, systems have to be perfect, responses have to be instant, and connectivity has to hold at 60 miles an hour on a highway without latency degrading the answer.

Capital Markets

Asked to characterize the environment for AI and innovation, Moynihan described it as infinitely receptive on one hand and increasingly discerning on the other.

For data center builds, investors want the revenue stream assured, and permitting and construction timelines are slowing things down.

On the IPO queue, he noted many companies have already raised what they need and cannot spend the cash on their balance sheets. They pursue public listings for other reasons: visibility with consumers, liquidity for employees, and investors wanting money out.

The constraint he identified is rates. Higher rates mean debt costs more, which slows activity.

Key Takeaways

1. The last 15 years were the anomaly. A 3.5 to 4 percent Fed funds rate is historically normal and compatible with growth.

2. Consumer sentiment and consumer behavior have decoupled, with spending up 5 percent year over year in September.

3. Lower-income paycheck growth now exceeds higher income, reversing the recent trend.

4. Erica took nearly a decade to expand from 210 to 700 prompts because it touches 110 systems and must be right every time.

5. Being right 95 percent of the time is not sufficient in finance or medicine, since trust does not survive the remainder.

6. Bank of America reports 150 AI use cases, $400 million invested against $800 million in benefits, doubling next year.

7. Capital markets are receptive but discerning, with higher debt costs the main brake on activity.