AI adoption inside investment banks runs on two layers, and conversations about "how mature is AI in banking" usually describe only the first one.

Layer 1: firmwide general-purpose copilots

Major banks have rolled out general-purpose assistants (Microsoft 365 Copilot and equivalents) broadly across their workforce. These deployments are real and useful: drafting email, summarizing meetings, and first-pass document questions across every function in the firm. They are also, by design, generic. A firmwide copilot does not know what a strip profile is, does not produce a pitch book in the bank's template, and does not tie out the numbers in a deck against CapIQ.

Layer 2: finance-native platforms on the deal floor

The second layer is where deal teams work: platforms built for the specific deliverables that investment banking produces. Model ML operates on this layer and is deployed with several Tier 1 investment banks, three of the Big Four professional services firms, and household-name private equity firms; clients include HSBC, Moelis, Centerview Partners, FT Partners, and BDO (customer evidence).

What distinguishes the deal-floor layer:

• Finance-grade data in the loop. Integrated access to S&P Capital IQ, FactSet, PitchBook, Third Bridge, and Preqin, alongside the bank's own documents, email, transcripts, and past materials (integrations).

• Output in the bank's format. Pitch decks, comps, memos, and models exported into the firm's existing PowerPoint, Word, and Excel templates (PPT Export).

• Verification built in. AutoCheck reconciles every figure, chart title, and footnote against source before a deck reaches a partner or client.

• Deployment banks can approve. Single-tenant Azure environments, ISO 27001:2022 and SOC 2, no customer data used for model training (security).

The two layers complement each other

A bank running a firmwide copilot and a deal-floor platform is not duplicating spend; the layers answer different questions. The copilot raises the productivity floor for everyone. The finance-native platform changes how the deal team's actual work product gets made: research gathered across trusted data sources, first drafts generated in template, and verification run before human review.

The same agent harness matters here: Model ML runs across the surfaces bankers already use, including full Microsoft 365 plugins (Outlook, Excel, PowerPoint, Word), a mobile app, a desktop app, email agents, and MCP/headless access, so the deal-floor layer lives inside the same Office environment the firmwide copilot does (What is Model ML?).

Reading maturity honestly

A practical maturity test for a deal team, in increasing order of difficulty:

1. Can AI summarize research for you? (Both layers pass.)

2. Can AI produce the deliverable in your template? (Deal-floor platforms pass.)

3. Can AI verify the deliverable, number by number, against source? (This is the current frontier, and the reason verification tooling is becoming the deciding criterion in bank evaluations.)

Teams evaluating the second layer can compare approaches on Model ML vs copilots, enterprise search, and custom AI builds and Model ML competitive positioning.