Bank of America Plans to Double Its AI Budget Next Year

Bank of America says it plans to double its AI expense budget next year, citing ~140 uses already live with claimed cost of $400M and benefit of $800M. Primary reporting: Banking Dive via Yahoo Finance — Bank of America plans to double AI spend (Caitlin Mullen, Wed Sep 23, 2026). Figures are company claims from BofA executive conference remarks (CEO Brian Moynihan, Co-President Jim DeMare, CTIO Hari Gopalkrishnan).

What shipped (claimed)

• Moynihan (Barclays conference last week): expects to double AI expense budget next year
• ~140 AI uses implemented; claimed cost $400M, benefit $800M
• Headcount ~213,000 start of year → 209,000; attrition ~8.5%; Moynihan: not laying people off — manage hiring carefully
• Erica (AI virtual assistant) handles work of ~11,000 people; considerable help-desk inquiry reduction
• ~95% of company has access to AI tools
• Bank spends ~$4B/year on new tech initiatives; AI is part of that
• Gopalkrishnan: employee ideas; AI investment interest more than twice last year; will probably spend twice next year vs this year on good ROI ideas
• Thousands using AI-powered CRM prep tools; next question = autonomy level for agents with omnipotent guardrails
• Moynihan: humans must check AI answers before clients — wrong answer loses the client

Caveat: Budget plans, cost/benefit figures, productivity percentages, and headcount framing are company claims from conference remarks as reported by Banking Dive / Yahoo Finance — not independent audits. Accenture (cited in the piece): only 20% of bank leaders seeing widespread sustained AI value.

How they ran it

• DeMare (BofA Securities conference Wed): most measurable returns in tech units / coding
• Across industries, coding productivity claimed ~15–20%; BofA seeing that with 20,000 software developers using coding agents
• Erica as customer/employee assistant reducing help-desk load
• Broad internal access (~95%) plus CRM prep tools for relationship teams
• Guardrail stance: agents with strong controls; human gate before client-facing answers

Scale and signal (claimed)

• AI spend: planned 2× next year vs this year’s AI expense budget
• ROI snapshot (claimed): $400M cost → $800M benefit across ~140 uses
• Developer footprint: 20k software developers on coding agents; ~15–20% productivity claim
• Workforce: 213k → 209k via attrition/hiring care, not framed as AI layoffs
• External context: Accenture cites only 20% of bank leaders seeing widespread sustained AI value

Operator takeaway

Read this as a task bundle, not a blank check: coding agents → help-desk deflection → CRM prep → client-facing answers with a human gate. Same pattern for operator stacks — leave / augment / automate per step, with humans owning judgment where wrong answers cost the relationship. Soft next step if useful: an Ada conversation to map which bank-ops or software tasks in your stack should stay human, which to augment with coding or CRM agents, and which to automate once evals and guardrails hold.

Source: Banking Dive via Yahoo Finance — Bank of America plans to double AI spend (Caitlin Mullen, Wed Sep 23, 2026). Figures are company claims from BofA exec conference remarks.