LaunchedEditorial Listing

Model ML

Model ML · Model ML: AI Agents That Build Models, Decks and Diligence for Banks, PE and Consultants

Open Model ML

Model ML is an enterprise AI agent platform for investment banks, consulting firms, private equity and credit funds, and asset managers: each professional gets an agent that builds financial models in Excel, pitch and valuation decks in PowerPoint, memos in Word and diligence outputs from data rooms, working from email, Teams, Slack, the Office apps or Model ML's own app. It is sold through demos with no public pricing, and can be deployed in a single-tenant environment or inside the customer's own Azure.

PricingCustom
Setuphard
Runs onWeb · API · Self-hosted
APIYes
Open sourceNo
DocsNo
CategoryFinance
Investment BankingPrivate EquityConsultingAsset ManagementFinancial ModelingPitch DecksDue DiligenceExcelPowerPointMicrosoft 365Custom AgentsDocument ReviewMeeting NotesMCPEnterpriseSelf-Hosted

Best for

Investment banks, consulting firms, PE and credit funds and asset managers on Microsoft Office that want an agent per deal professional to produce models, decks, memos and diligence in the firm's own templates, and that can run an enterprise security review and rollout

Not ideal for

Individuals and small teams wanting self-serve sign-up or published pricing, firms that work mainly in Google Sheets and Slides, buyers who need public documentation before a demo, and teams outside finance and professional services

Who it's for

Investment bankers, consultants, private equity and credit investors, and asset and wealth managers at firms that run enterprise software procurement

Capabilities

  • Agent Task Manager: reads requests arriving by Outlook, Gmail, Teams or Model ML's Notetaker, adds each to a to-do list with its context, starts the work on its own and stops to ask when it reaches a judgment call
  • Custom Agents described in plain language, which Model ML turns into instructions and a model choice, run on demand, on a schedule, when new files are uploaded, or when a filing, price move or earnings release occurs, and can be shared across a team
  • Multi-agent runs: larger tasks are split across focused agents working in parallel and combined into one output, and one agent can hand finished work to the next from Excel to PowerPoint to Word in a single run
  • Plugins for Excel, PowerPoint, Word and Microsoft Teams that carry the same context, memory and skills: push an assumption change across every file in a project, build a valuation book in the house template, or draft an IC memo from the deal's existing work
  • Excel change tracking: see every cell the agent changed, before and after, and undo any change
  • Document Review (AutoCheck) agent that flags calculation errors, numbers that disagree across pages, broken page references, text that contradicts charts, formatting slips and placeholder text, prioritised by severity, and compares two versions of a deck slide by slide
  • Notetaker with automated transcripts, summaries and pre-meeting prep notes, audio-only recording, retention controls and firm-level rules on which meetings can be recorded
  • Grids for running thousands of prompts across a document set with every source cited, and Sites for interactive web deliverables with per-recipient access, expiry, view analytics and scheduled refreshes
  • Financial data connectors including S&P Capital IQ, FactSet, PitchBook, LSEG, Preqin, Daloopa, Third Bridge, Crunchbase, Dun & Bradstreet and Bigdata.com, with the firm choosing which connectors are on, and outputs linked back to their sources
  • MCP in both directions: connect the firm's own systems and data to Model ML, or call Model ML's agents from Claude, Copilot or internally built agents
  • Model routing across providers based on Model ML's own Composite benchmark of finance tasks, with Claude and GPT models added as they release and a growing share of work on Model ML's fine-tuned open-source models
  • Proprietary filings dataset covering public companies worldwide and private-company accounts from UK Companies House, Singapore and Hong Kong, delivered to agents or to the firm by data feed, MCP or REST API

Limitations

  • No public pricing and no self-serve sign-up: access starts with a demo or contact form, and commercial terms are scoped per engagement
  • No public product or API documentation was found; Model ML says API documentation is provided through a request form
  • The in-app experience is built around Microsoft Office: plugins are documented for Excel, PowerPoint, Word and Teams, and no Google Sheets or Google Slides add-on is documented
  • Deployment inside the customer's own cloud is described for Microsoft Azure; customer-hosted deployment on AWS or Google Cloud is not documented
  • Many data connectors, including Daloopa, Third Bridge, Quartr, Harmonic and the SS&C Intralinks integration, work through the firm's own license or account, so the data an agent can use depends on the subscriptions the firm already holds
  • Performance and cost figures are vendor-reported: Model ML says its fine-tuned models run at 2% of the cost of frontier models on the workloads they handle, and its model comparisons come from its own Composite benchmark; no independent evaluation was found
  • Not every model is available to every customer: Model ML says access to Claude Fable 5.1 depends on the customer's compliance review because of Anthropic's 30-day data retention for that model
  • Built for financial and professional services deal work; teams outside banking, consulting, investing and wealth management are not its target users

Use cases

  • Replying to an MD's email with the agent CC'd and getting back a sourced company landscape, an Excel backup in the firm's template and a draft PowerPoint logo page
  • Reworking a DCF and football field to a tighter valuation range after a client email asks for it before a 4pm meeting
  • Running a weekly comps-spreading agent that pulls filings, spreads and calendarises comps with every number sourced
  • Screening every inbound CIM or teaser on arrival against a PE fund's mandate and its portfolio companies' bolt-on plans, logging it in the CRM and emailing the partner a summary
  • Checking a client deck for numbers that disagree across pages, broken page references and calculation errors before it goes out
  • Turning data room materials into a draft CIM or IC memo in the firm's own format

Our take

Model ML is aimed squarely at the junior-banker workload: comps, model refreshes, valuation books, CIM drafts and deck checks, handed over by email or Teams and returned in the firm's own templates. What lifts it above a finance chatbot is the plumbing around the agent: a task manager that picks up requests from the inbox and meetings, custom agents that fire on schedules and filings, Office plugins with cell-level change tracking, and connectors to the major data providers deal teams already license. Its security and deployment story (ISO 27001, SOC 2 Type II, no training on customer data, deployment inside the customer's Azure) is built for bank procurement, and its announced deployments at HSBC, PwC, FT Partners and Deloitte's M&A practice show large institutions are using it. The flip side is opacity: no pricing, no trial and no public documentation, so evaluation means a demo, a security review and a pilot, and the time-saving and model-cost figures are Model ML's own. For a firm on Microsoft 365 with the data subscriptions already in place, it is worth piloting on one recurring workflow, such as comps spreading or deck QA, and checking every output against what an analyst would have produced.

Who should use it

Investment banks, advisory boutiques, consulting firms, PE and credit funds and asset managers that work in Microsoft 365, already license data such as Capital IQ, FactSet or PitchBook, and want agents producing first drafts of models, decks, memos and diligence in their own templates, with deployment controls that will pass a bank-grade security review.

Who should skip it

Individuals and small teams that need self-serve access or published pricing, firms built on Google Workspace, buyers who need public documentation or a trial before talking to sales, and anyone outside financial and professional services looking for a general office agent.

Strengths

  • Built around deal deliverables (models, valuation books, CIMs, IC memos) in the firm's own templates rather than chat answers
  • Work can be handed over from email, Teams, Slack, meetings or the Office apps, and Task Manager starts on requests without being prompted
  • Scheduled and event-triggered custom agents for recurring work such as comps spreading and data room indexing
  • Broad catalog of licensed financial data connectors plus MCP for the firm's own systems
  • Strong enterprise posture: ISO 27001:2022, SOC 2 Type II, SSO, no training on customer data, and deployment inside the customer's Azure
  • Not locked to one model provider

Weaknesses

  • No public pricing, trial or self-serve access
  • No public product or API documentation
  • Microsoft Office-centred; no Google Sheets or Slides add-on documented
  • Customer-hosted deployment documented only for Azure
  • Speed, accuracy and cost figures are Model ML's own

Model ML pricing

Enterprise

Contact sales

  • Access through a demo or contact form; no self-serve plan or free trial
  • Single-tenant deployment or deployment inside the customer's own Azure
  • SSO (SAML or OIDC), JIT provisioning and MFA
  • Optional forward-deployed engineers for custom agent design, implementation and training

Note: Model ML does not publish pricing; commercial terms are scoped to each engagement through a sales-led process. Its security page refers to an Enterprise Plan for customers using their own SSO identity provider, but no plan list or prices are published. Checked on modelml.com and llms.modelml.com on 2026-11-18.

Technical specs

Modalities

Text, Documents, Audio

API pricing

Not published. Model ML says API documentation is provided through a request form, its agents can be called from other tools over MCP, and its filings dataset is available by data feed, MCP or REST API.

Available models

Claude Opus 5.5Claude Sonnet 5.5Claude Fable 5.1GPT-5.6 SolGPT-5.6 TerraGPT-5.6 LunaModel ML fine-tuned open-source models

Where Model ML excels

Turning an MD's email into a briefing pack

A banker CCs the agent on the reply; it pulls from Capital IQ, PitchBook and precedent databases, returns a sourced landscape with an Excel backup, and builds the PowerPoint page and the draft reply from the plugins.

Screening inbound CIMs for a PE fund

A schedule runs whenever a CIM or teaser arrives, logs the deal in the CRM, screens it against the fund mandate and the portfolio companies' bolt-on plans, and emails a summary with a verdict.

Pre-send deck QA

The Document Review agent flags figures that disagree across pages, calculations that do not add up and broken page references, ranked by severity, before a client deck goes out.

Recurring comps and model refreshes

A scheduled comps-spreading agent pulls and calendarises filings each week, and the Excel plugin pushes updated assumptions through every linked file in the project.

Model ML vs. competitors

Model ML vs. AlphaSense SuperAnalyst

AlphaSense SuperAnalyst is a research agent inside an AlphaSense subscription, drawing mainly on AlphaSense's own library of broker research, filings, transcripts and expert calls to produce cited memos, models and decks. Model ML is a standalone platform that connects to third-party data providers and the firm's own systems, and puts more weight on producing deal deliverables in the firm's templates inside Excel, PowerPoint, Word, email and Teams, with scheduled custom agents and a task manager.

Model ML vs. Microsoft Copilot

Microsoft 365 Copilot is a general assistant built into Word, Excel, PowerPoint, Outlook and Teams for every kind of knowledge worker. Model ML works in the same Office apps but is specialised for finance deal work, with licensed market-data connectors, finance-specific custom agents, deck review for numerical consistency and deliverables in the firm's house templates.

Model ML vs. Claude Cowork

Claude Cowork is Anthropic's general task-execution agent for office work, built into the Claude app. Model ML is an enterprise platform specific to banking, consulting and investing that routes across several model providers (including Claude), adds financial data connectors and Office plugins, and can be deployed inside the customer's own Azure.

Frequently asked questions

What is Model ML?

Model ML is an AI agent platform for financial and professional services. It gives bankers, consultants and investors a personal agent that takes requests from email, chat, meetings, Teams, Slack and the Office apps and produces finished work such as Excel models, PowerPoint decks, Word memos and diligence summaries in the firm's own formats, using the firm's files and connected market data.

How much does Model ML cost?

Model ML does not publish pricing. It is sold to firms through demos and a contact form, commercial terms are scoped to each engagement, and there is no self-serve plan; no free trial is advertised.

Does Model ML work inside Excel and PowerPoint?

Yes. Model ML has plugins for Excel, PowerPoint, Word and Microsoft Teams. In Excel the agent can push an assumption change across every file in a project and shows each cell it changed with an undo option; in PowerPoint it can build decks such as valuation books in the firm's template.

Can Model ML run agents on a schedule?

Yes. Custom Agents can run on demand, on a fixed schedule, when new files are uploaded, or when a filing, price move or earnings release occurs. Examples Model ML gives include a weekly comps-spreading agent, a daily data room indexing agent and a buyer tracking agent that runs on new uploads.

Can Model ML be deployed in our own environment?

Yes. Model ML says it offers flexible deployment and often deploys within customers' own infrastructure so data does not leave their environment. Its published materials describe single-tenant Azure environments as standard and full deployment into a customer's own Azure; other clouds are not documented.

Does Model ML train AI models on customer data?

No. Model ML states that customer data is never used to train AI models. It is ISO/IEC 27001:2022 certified, SOC 2 Type II audited and supports SAML or OIDC single sign-on with JIT provisioning and MFA.

Which AI models does Model ML use?

Model ML is not tied to one provider. It routes each task to a model based on its own Composite benchmark of finance tasks, has announced Claude and GPT models as they became available (including Claude Opus 5.5, Claude Sonnet 5.5, Claude Fable 5.1 and GPT-5.6 Sol, Terra and Luna), and says a growing share of work runs on its own fine-tuned open-source models. Access to some models can depend on the customer's compliance setup: Model ML says Fable 5.1 is switched on only after review because Anthropic retains that model's traffic for 30 days.

Model ML vs AlphaSense SuperAnalyst: what is the difference?

SuperAnalyst is a research agent inside AlphaSense that draws mainly on AlphaSense's own library of broker research, filings, transcripts and expert calls. Model ML is a separate platform that connects to third-party data such as Capital IQ, FactSet and PitchBook and to the firm's own systems, and focuses on producing deal deliverables inside Excel, PowerPoint, Word and the firm's communication tools.

Integrations & fit

Microsoft ExcelMicrosoft PowerPointMicrosoft WordMicrosoft OutlookMicrosoft TeamsGmailSlackS&P Capital IQFactSetPitchBookLSEGPreqinDaloopaThird BridgeCrunchbaseDun & BradstreetBigdata.comSnowflakeSS&C IntralinksQuartrHarmonicMicrosoft AzureMCPAPI
Good fit forEnterprise
Pricing modelCustom· Contact for pricing
See pricing on Model ML →

Alternatives to consider

About Model ML

Model ML pitches itself as the personal AI agent for every banker, consultant and investor, and its product is built around the deliverables of deal work rather than around answering questions. Requests can start in Chat, by emailing or CC'ing the agent, by tagging @Model ML in Microsoft Teams or Slack, from the Excel, PowerPoint and Word plugins, or from action items its Notetaker captures in meetings, and the agent returns work in the firm's own templates: DCF and LBO models with live formulas, comps and precedent transactions, valuation books, CIM drafts, IC memos, buyer logs and diligence summaries. Three pieces make it more agentic than a finance chatbot. Agent Task Manager reads requests arriving by Outlook, Gmail, Teams or the Notetaker, adds them to a to-do list with their context and starts the work, stopping to ask when it reaches a judgment call. Custom Agents are described in plain language and run on demand, on a schedule, when new files land, or when a filing, price move or earnings release occurs, and one agent can hand work to the next from Excel to PowerPoint to Word in a single run. Since September 2026, larger tasks are split across agents running in parallel by default, and in Excel you can see every cell the agent changed and undo it. Around the agent sit Document Review (the AutoCheck review agent, which checks decks for calculation errors, inconsistent numbers, broken page references and logic gaps, and compares versions), Grids for running thousands of prompts across a document set with every source cited, Sites for interactive, permissioned web deliverables that can refresh on a schedule, and a connector catalog covering S&P Capital IQ, FactSet, PitchBook, LSEG, Preqin, Daloopa, Third Bridge, Crunchbase and Dun & Bradstreet, plus the firm's own systems over MCP. Model ML is not tied to one AI provider: it routes each task to a model using its own Composite benchmark, has added Claude and GPT models as they ship, and says a growing share of work runs on its own fine-tuned open-source models. It is built for regulated institutions, with ISO/IEC 27001:2022 certification, a SOC 2 Type II audit, SAML or OIDC SSO, a commitment never to train on customer data and the option to deploy inside the customer's own Azure, and it offers forward-deployed engineers who sit onsite to build custom agents and run training. Announced customers include HSBC, which rolled it out globally in September 2026, PwC, which is scaling it globally, FT Partners, which rolled it out firm-wide, and Deloitte's M&A practice; the company raised a $250 million Series B in November 2026. The tradeoffs are the ones of an enterprise platform: no public pricing, no self-serve trial, no public product documentation, a Microsoft Office-centred workflow, and speed and cost figures that come from Model ML itself.

Updates from Model ML

Milestone$250 million Series B at a $1.5 billion valuation

Model ML raised $250 million in Series B funding co-led by General Atlantic and QED Investors, bringing total funding to $350 million. The company said annual recurring revenue grew 12x year on year in Q3 2026 and that it will expand research and engineering in London and New York.

New FeatureMulti-agent runs, Excel change tracking and Slack

Model ML's September 2026 release split larger tasks across agents running in parallel by default, showed every Excel cell the agent changed with undo, added an @Model ML Slack app, multiplayer chats, scheduled Site refreshes and Site analytics, and let the Document Review agent check against call transcripts.

MilestoneHSBC rolls out Model ML globally

HSBC rolled out Model ML across the bank to help with research, analysis and drafting content in HSBC's own formats and templates.

IntegrationMCP support in both directions

Model ML added Model Context Protocol support so firms can connect their own systems and data, such as Snowflake or LSEG, to Model ML agents, and call Model ML from other MCP-compatible tools such as Claude, Copilot or in-house agents.

New FeaturePlugins for Word, Excel, PowerPoint and Teams

Model ML launched plugins for Word, Excel, PowerPoint and Microsoft Teams that carry the same agent context, memory, skills and custom workflows into each app.

Are you the founder? Claim this listing →