Anonymized tier-level analysis — Q1 2024 through Q4 2025
This report examines how 49 of the largest U.S. banks publicly communicate their AI activity, using only publicly available sources such as SEC filings, earnings calls, press releases, and industry research. It evaluates the level and nature of AI-related disclosures—distinguishing between meaningful strategic commitment and generic, boilerplate language. It does not assess the actual quality or effectiveness of a bank’s AI capabilities.
Listen to a 2-minute audio brief
Every chart below is built on a quarterly AI signal rating assigned to each bank. Here's what those ratings mean.
What we mean by "signal": A publicly detectable indicator that a bank is engaging with AI — whether through SEC filings, executive commentary, partnerships, hiring, or industry coverage. A signal is not the same as capability. A bank can have strong AI programs with weak public signal (the disclosure gap), or a loud AI narrative with thin implementation. This dashboard measures signal, not capability.
| Rating | What It Means | Example Signals |
|---|---|---|
| AI is a dominant strategic theme | Named AI products, quantified ROI, CEO framing AI as strategic priority, multiple source types | |
| AI is present but not dominant | AI mentioned in 1-2 sources, directional language, pilot programs, governance buildout | |
| AI is tangential or defensive | Risk factor boilerplate, generic "technology investment," regulatory compliance language | |
| No AI signal detected | No AI-related language found in any of the 5 source types for that quarter |
Ratings are assessed per quarter across 5 source types: SEC filings, earnings calls, regulatory signals, announcements, and industry research. A bank can move between ratings quarter to quarter based on public disclosure activity. See full methodology below.
Tiers: Megabank (>$1T) — the 6 banks most people recognize by name. Large ($200B-$1T) — major nationals. Super-Regional ($80B-$200B) — multi-state footprints. Regional ($40B-$80B) — strong in specific geographies. Community (<$40B) — local and specialized. Tier boundaries are analytical groupings, not competitive sets.
Percentage of banks rated HIGH per quarter, by tier. Each line tracks what share of banks in that tier had AI as a dominant strategic theme.
A five-tier split that widens over time. In Q1 2024, only 2 banks in the entire dataset rated HIGH. By Q4 2025, 25 did — but the gains were concentrated at the top. Every megabank hit HIGH by Q3 2025. Nine of 10 large banks got there. Zero community banks did.
The Q1 2025 dip is real, not an error. Megabanks dropped from 83% to 50% and large banks from 50% to 30%. Ratings reflect the public signal in that specific quarter, not cumulative capability — a bank can be HIGH one quarter and MOD the next if that quarter's filings and calls had less AI emphasis. The recovery by Q2-Q3 confirms these were temporary dips, not reversals.
Why it matters: If you're benchmarking a bank's AI posture, tier matters more than any individual initiative. A super-regional at HIGH is outperforming 70% of its tier; a megabank at HIGH is just keeping up.
All 49 banks categorized by rating each quarter. Watch the color bands shift from left (NONE/LOW) to right (MOD/HIGH) over time.
The entire industry is moving right. In Q1 2024, 36 of 49 banks (73%) were rated NONE or LOW — AI was absent or defensive boilerplate. By Q4 2025, that dropped to 14 (29%). The green HIGH band grew from 2 banks to 25.
The LOW band is a holding pattern. It shrank from 24 to 8, but slowly. Banks don't jump from LOW to HIGH — they pass through MOD first, often spending 2-3 quarters there as they build governance, hire talent, and pilot projects before going strategic. The MOD band is the pipeline.
NONE hasn't hit zero. Six banks still had no detectable AI signal in Q4 2025. These aren't banks ignoring AI — some have internal programs — they're banks that don't discuss it publicly. The difference between "no AI" and "no AI disclosure" is the disclosure gap (see Findings).
Average signal intensity by tier on a 0-3 scale (NONE=0, LOW=1, MOD=2, HIGH=3). Darker green = stronger AI signal.
| Tier | Q1 '24 | Q2 '24 | Q3 '24 | Q4 '24 | Q1 '25 | Q2 '25 | Q3 '25 | Q4 '25 |
|---|---|---|---|---|---|---|---|---|
| Megabank (6) | 2.0 | 2.33 | 2.58 | 2.83 | 2.5 | 2.5 | 3.0 | 3.0 |
| Large (10) | 1.5 | 1.7 | 2.15 | 2.35 | 2.1 | 2.7 | 2.85 | 2.9 |
| Super-Regional (10) | 1.0 | 1.5 | 1.3 | 2.2 | 1.8 | 2.1 | 2.3 | 2.6 |
| Regional (15) | 0.87 | 0.93 | 1.13 | 1.4 | 1.4 | 1.33 | 1.8 | 1.67 |
| Community (8) | 0.25 | 0.12 | 0.25 | 0.5 | 0.88 | 0.38 | 0.25 | 0.62 |
Read it top-to-bottom within a column to see the tier hierarchy. Read it left-to-right within a row to see acceleration. Megabanks started at 2.0 (solidly MOD) and reached 3.0 (pure HIGH). Community banks started at 0.25 and are still at 0.62 — barely above NONE.
The gap is widening, not closing. In Q1 2024, the megabank-to-community gap was 1.75. By Q4 2025, it was 2.38. The tiers in between aren’t converging either — super-regionals are pulling away from regionals at roughly the same rate that megabanks are pulling away from them.
Why it matters: This table is the clearest view of AI as a stratification force in banking. Size isn’t just correlated with adoption — it’s predictive. Banks should benchmark against their tier, not the industry average.
AI-related keyword mentions in SEC filings (10-K, 10-Q, 8-K) by tier per quarter.
The Q1 spikes aren't an AI trend — they're filing seasonality. Annual 10-K filings (filed in Q1) are 100+ pages of substantive narrative. Quarterly 10-Q filings are shorter, more formulaic, and less likely to mention AI even if the bank is actively deploying it. That's why Q1 2024 (206 mentions) dwarfs Q2 2024 (17).
The meaningful comparison is year-over-year Q1. Q1 2024: 206 mentions. Q1 2025: 528 — a 2.6x increase. That's real growth in how banks are framing AI in their most substantive annual disclosures.
Tier volumes can be misleading. The Large tier dominates mention counts (539 of 1,051), but that's driven by a few outlier banks with unusually high filing volumes. Mention volume doesn't correlate cleanly with AI maturity — it's a measure of disclosure verbosity, not capability.
The dominant AI vocabulary in SEC filings. These terms have been present throughout the study period.
"AI" and "artificial intelligence" are the workhorses. Together they account for 80%+ of all AI-related mentions. The pattern here mirrors the mention volume chart — Q1 spikes from 10-K filings, quieter mid-year.
"Machine learning" is fading. Not because banks stopped using ML, but because the language is shifting. "Machine learning" was the technical term of 2020-2023. By 2025, executives say "AI" to mean the same thing. The term's decline is a narrative signal, not a technology signal.
Low-frequency but narratively significant. These terms mark inflection points in how banks talk about AI.
"Generative AI" arrived in filings mid-2024 and peaked in Q1 2025 annual filings (21 mentions). This tracks the real-world timeline — ChatGPT launched late 2022, banks started piloting in 2023, and by 2024-2025 it was appearing in official SEC language. The term's presence in a filing signals the bank has moved past evaluation into deployment.
"Agentic" is the newest signal — and the most telling. Zero mentions before Q3 2025, then 2, then 4. Only 6 total mentions across all 49 banks. But this term marks a phase shift: from AI as a tool that assists humans (copilots, chatbots) to AI as an autonomous actor that executes multi-step workflows. Banks using "agentic" in their filings are signaling a fundamentally different AI architecture.
"RPA" is the old guard fading out. Steady at 2-4 mentions per quarter with no growth. Robotic process automation was the pre-AI automation paradigm. Its flat trajectory against the rising AI terms tells the story of a generational technology transition.
Once a bank shows real AI activity (first MOD), how quickly does it become a strategic priority (first HIGH)? Megabanks converted in ~1 quarter; regionals took a full year — and the gap compounds as early movers execute while others are still ramping.
The compounding gap: It's not just that bigger banks adopted AI first — they accelerated faster once they started. A megabank that hit MOD in Q1 2024 was at HIGH by Q2 2024. A regional that started in Q1 2024 might not reach HIGH until Q1 2025 — by which point the megabank has been executing at scale for a full year.
Reaching HIGH once could be a strong earnings call or a splashy announcement. Staying there means AI is structurally embedded. This chart separates the committed from the performative.
26 of 49 banks have reached HIGH at least once. But only 12 sustained it for 3 or more consecutive quarters. That gap is the difference between signaling AI ambition and demonstrating AI commitment. Sustained HIGH means AI appeared as a strategic theme in filings, earnings calls, and announcements quarter after quarter — not a one-time spike.
Megabanks don’t just reach HIGH — they stay there. 4 of 6 megabanks sustained HIGH for 3+ quarters, and 3 of those sustained for 5+. In contrast, only 1 of 10 super-regionals sustained for 3+ quarters, and zero regionals or community banks did.
2 banks hit HIGH exactly once and fell back — a flash of AI signal that didn’t persist. This pattern typically appears when a bank makes a major AI announcement or partnership in one quarter but doesn’t follow through with consistent strategic framing.
Why it matters: For investors and analysts, sustained HIGH is a credibility filter. A bank at HIGH for one quarter is interesting. A bank at HIGH for five consecutive quarters has structurally integrated AI into its strategy, narrative, and operations. This metric separates signal from noise.
How many banks were rated HIGH each quarter — the clearest single measure of industry-wide AI momentum.
The industry crossed a threshold in the second half of 2025. Banks rated HIGH nearly doubled from 13 to 25 in two quarters.
23 of 49 banks (47%) never achieved a HIGH rating across all 8 quarters. All 8 community banks and 11 of 15 regional banks remain below HIGH.
The bar chart tracks how many banks hit HIGH each quarter. This is not cumulative — a bank counted as HIGH in Q3 2025 might have been MOD in Q2 2025. The chart shows the industry's "high-water mark" quarter by quarter.
Two growth phases are visible. Phase 1 (Q1-Q4 2024): gradual climb from 2 to 13, driven by megabanks and early-moving large banks. Phase 2 (Q1-Q4 2025): acceleration from 6 to 25, as super-regionals and a few regionals crossed the threshold.
The Q1 2025 dip (13 to 6) is the same seasonal effect visible in Chart 1. Fewer banks produce HIGH-level AI signal in Q1 earnings calls, even if their programs are running. It snapped back by Q2.
Why 25, not 49: Half the industry still isn't at HIGH by Q4 2025. The 24 remaining banks are overwhelmingly regional and community tier — this isn't a temporary lag, it's a structural divide.
Anonymized observations from analyzing AI adoption patterns across 49 banks over 8 quarters.
Banks that publicly quantified AI-driven cost savings in early quarters showed accelerated adoption in subsequent periods. By Q4 2025, every megabank framed AI as self-funding — savings from automation directly financing the next wave of deployment. This flywheel dynamic creates a compounding advantage that mid-tier banks struggle to replicate without comparable scale.
A clear pattern emerged: banks that led with governance frameworks (risk committees, acceptable use policies, model validation) before scaling AI achieved more consistent signal trajectories. Banks that moved fast without governance often showed volatile ratings — HIGH one quarter, dropping to MOD the next as compliance caught up. The governance-first banks took longer to reach HIGH but stayed there.
"Agentic AI" appeared in zero SEC filings before Q3 2025. By Q4 2025, it appeared across multiple tiers. This term marks a phase shift: from AI as a tool (copilots, chatbots) to AI as an autonomous actor (agentic workflows, multi-step reasoning). Banks using this language are signaling a fundamentally different AI strategy than those still discussing "automation" or "machine learning."
Several banks showed strong internal AI activity (job postings, patent filings, vendor partnerships) but minimal public signal in SEC filings and earnings calls. This disclosure gap is most pronounced in the Regional tier, where 60% of banks have some form of AI deployment but only 20% communicate it as a strategic priority. For investors, this creates an information asymmetry — the filing is not the full story.
Banks in active merger integration showed consistently suppressed AI signals. Technology integration, regulatory approvals, and cultural alignment consume the same executive bandwidth that AI transformation requires. Banks mid-acquisition averaged 1.5 rating points lower than size-comparable peers, with the suppression lasting 3-4 quarters post-close.
Zero community banks (<$40B) achieved a HIGH rating in any quarter. This is not merely a budget constraint — it reflects structural barriers: lack of dedicated technology leadership, regulatory proportionality reducing AI urgency, vendor ecosystems that don't yet serve this tier effectively, and a relationship-banking model that views AI as threatening rather than enabling. The gap is widening, not closing.
An unexpected finding: banks where the CFO (not CTO) led AI narrative on earnings calls showed faster adoption trajectories. CFO-led framing anchors AI to ROI and efficiency metrics, creating board-level urgency. CTO-led framing often stalled at "innovation" and "experimentation" — language that doesn't unlock budget. The most successful banks had both voices aligned, but CFO conviction was the catalyst.
Some banks have strong internal AI programs but weak public signals — and vice versa. Banks with robust AI hiring, patent filings, and vendor partnerships sometimes say almost nothing on earnings calls. For investors, this creates an arbitrage: the public signal underweights actual capability. Conversely, banks with aggressive AI narratives but thin implementation are overweighted. The gap between disclosure and reality is widest in the super-regional tier.
Of 1,051 AI-related mentions across all SEC filings, only 9 appeared in 8-K filings — the form reserved for material events like acquisitions, leadership changes, and significant business developments. No bank in the dataset considers an AI initiative material enough to disclose as a standalone event. That could change as AI spending hits materiality thresholds or regulators mandate AI-specific disclosures, but today the silence is telling: AI is discussed in annual narratives and quarterly updates, never as a triggering event.
Where does our bank fall on the Sustained vs. Flash spectrum? Do we have a named AI governance role — and if not, are we deploying faster than we’re governing? Is our AI language in SEC filings still defensive boilerplate, or has it shifted to strategic framing?
Key terms used throughout this dashboard, defined for readers across industries.
How the ratings work, what they measure, and what they don't.
| Rating | Definition |
|---|---|
| Multiple AI mentions across 2+ source types; named products, quantified metrics, or strategic framing | |
| AI referenced in 1–2 sources; directional but not dominant narrative | |
| Tangential or defensive AI mention (risk factors, generic “technology investment”) | |
| No AI-related signal found in any source type for that quarter |
Hybrid ratings (LOW-MOD, MOD-HIGH) are used when a bank falls between categories. For aggregation in charts, LOW-MOD counts toward MOD and MOD-HIGH counts toward HIGH.
49 of the 50 largest US banks by total assets, Q1 2024 through Q4 2025 (8 quarters). Each bank was assessed across all 5 signal sources per quarter. This dashboard shows tier-level aggregates only — individual bank data is intentionally excluded.
artificial intelligence, machine learning, generative AI, gen-AI, AI, large language model, LLM, natural language processing, NLP, deep learning, neural network, chatbot, virtual assistant, robotic process automation, RPA, intelligent automation, agentic, copilot, GPT.
A governance dimension will be added to this dataset: AI risk factor classification, governance structure assessment, and regulatory preparedness scoring across the same 49 banks. The goal is a Deployment vs. Governance maturity matrix identifying which banks are deploying AI faster than they are governing it.
Based on analysis of public SEC filings, earnings transcripts, and industry research. Ratings reflect publicly available signals only. Individual bank names are intentionally excluded from this public analysis. Not investment advice.