Paper Instruments

AI tools for investment memos, compared

An investment memo is written after the diligence is done, and the tools split that work between them. Some turn data rooms into cited facts, some supply the market evidence, and some build the model and the draft itself. The eight compared below, as of November 2026, include prices where they are public.

Checked on

At a glance

ProductBest forAvailable onStarts atFree access
RogoSourced market and comps research at bank scaleEnterprise web platformNot public, routed to salesNone documented
HebbiaExtracting cited facts from large data roomsEnterprise web platformNot public, routed to salesNone documented
Blueflame AIDiligence and memo analysis inside DatasiteEnterprise platform in the Datasite stackNot public, routed to salesNone documented
Ultramarine WorkMemos, models, cohorts, cap tables and tie-outsDesktop app (Mac and Windows) and browser$5 a month (Starter)7 days free on Starter and Plus
AlphaSenseMarket evidence: filings, research and transcriptsWeb and mobile platformNot public, annual subscriptionsFree start offered, trial terms not stated
ClaudeMemo drafting and file work, with finance pluginsDesktop, web and mobile$20 a month ($17 annual)Free tier, limited usage
ChatGPT WorkMulti-app tasks that end in finished filesDesktop, web and mobile$20 a month (Plus)Limited Work, desktop only
Perplexity ComputerCited market research by sub-agentsWeb, Comet, Mac app and mobile$20 a month (Pro)Free search, limited Computer turns

The tools to weigh

  1. 1. Rogo

    Rogo

    Good for
    Rogo is an AI platform for financial institutions, with purpose-built financial agents for research, comps and first-draft deal materials. It connects to FactSet, PitchBook, DealCloud and SharePoint, produces sourced public-market research and comps tables, and deploys with Forward Deployed Bankers.
    How it works
    Enterprise platform for banks, equity research and advisory teams.
    Runs on
    Not disclosed.
    Price
    Not public. Sales runs a proof of concept first, then quotes. No free access documented.
    Choose it when
    The memo’s market and comps sections need live public-market data inside the data sources and deal CRM your institution already runs.

    Sources: Rogo homepage

  2. 2. Hebbia

    Hebbia

    Good for
    Hebbia’s Matrix runs a question across a document set as a grid of companies or documents, with every cell cited to its source page, and Max works the whole corpus. For memos it supplies the evidence base: revenue by segment, customer concentration and retention details, each traced to the page a partner can check.
    How it works
    Enterprise web platform for multi-document extraction and research.
    Runs on
    Not disclosed.
    Price
    Not public, routed to sales. No free tier documented.
    Choose it when
    The data room is large and the committee asks where every number came from, so you need comparable, cited facts by company or question.

    Sources: Hebbia homepage

  3. 3. Blueflame AI

    Blueflame (Datasite)

    Good for
    Datasite’s AI agent for M&A covers sourcing, diligence and monitoring, with document question-answering under citations, workflow Blueprints and access to Grata’s market data. For PE teams it automates memo-input analysis inside the data-room stack, including IC-memo analysis as a standing workflow.
    How it works
    Enterprise platform, part of Datasite’s M&A infrastructure.
    Runs on
    Third-party multi-model architecture that selects and orchestrates models per task. Models not named.
    Price
    Not public, routed to sales. No free access documented.
    Choose it when
    Your deals already live in Datasite or Grata, and you run enough processes a year to standardize diligence and IC work as Blueprints.

    Sources: Blueflame AI homepage, Datasite acquires Blueflame (July 2025)

  4. 4. Ultramarine Work

    Paper Instruments

    Ultramarine Work fits the investor who writes the memo: growth equity associates and fund partners. Its Venture & Growth Equity mode stays selected for the workstream, with eleven skills covering cohorts, market sizing, valuation, cap tables and memos, and its skills have it cite the source file behind each figure. A person reviews every file it touched. Firms wanting governed, firm-wide data-room deployment start with the specialists instead.

    Each mode in Ultramarine Work configures a whole practice: its audience, recurring deliverables, and a curated set of skills with their methods and review standards. The agent picks the skills each step needs. The model is Ultramarine 1 Flash, post-trained on GLM-5.3-Flash for knowledge work. Paper Instruments writes the skills and trains the model, so the two improve together. Standard, the default, is Ultramarine 1 Flash doing the task. Assisted adds a frontier model as an advisor. Advanced hands the whole task to a frontier model.

  5. 5. AlphaSense

    AlphaSense

    Good for
    For the market section of a memo, AlphaSense is the evidence base: filings, broker and independent equity research, a large transcript library, Canalyst models, and SuperAnalyst plus Workflow Agents. Enterprise Suite adds an API and MCP, so a firm’s other agents can query its content.
    How it works
    Web and mobile market-intelligence and research platform.
    Runs on
    Not disclosed.
    Price
    Not public. Annual subscriptions: Market Intelligence is per-seat, Enterprise Suite is quoted separately. Trial terms not stated.
    Choose it when
    The thesis depends on evidence the deal team cannot produce itself, such as broker estimates, expert-call transcripts or a competitor set across filings.

    Sources: AlphaSense pricing page

  6. 6. Claude

    Anthropic

    Good for
    Claude is a general agent for knowledge work, with Cowork folded in: file work across connectors, scheduled tasks, skills, and plugins that can bundle several skills at a time. Its marketplace has finance plugins for private equity and equity research, and ten financial-services agent templates.
    How it works
    General agent in chat and Cowork, with a plugin marketplace.
    Runs on
    Anthropic’s own Claude models: Opus, Sonnet, Haiku and Fable.
    Price
    Free tier with limited usage. Pro $20 a month ($17 annual), Max $100, Ultra $200 with no 5-hour limit for verified accounts, plus Team and Enterprise plans.
    Choose it when
    Your firm wants its memo process as editable plugins it can inspect, share and improve, with recurring tasks across connected apps.

    Sources: Claude plans and pricing, Claude Cowork help article

  7. 7. ChatGPT Work

    OpenAI

    Good for
    OpenAI’s agent for multi-step work that ends in finished files, ChatGPT Work gathers context across connected apps, produces documents, decks and web apps, and pauses for approval on consequential actions. The desktop app adds local folders and a browser, and Pro and Enterprise can run several agents in parallel.
    How it works
    General agent, desktop, web and mobile.
    Runs on
    OpenAI’s own GPT-5.6 and GPT-6.1 models (November 2026).
    Price
    Free and Go ($8) include limited, desktop-only Work. Plus $20 a month, Pro $100 to $500, Business $20 to $25 a seat, Enterprise custom.
    Choose it when
    Your team already works in ChatGPT and wants to test agent file-work on the free desktop access before any procurement decision.

    Sources: ChatGPT pricing, ChatGPT Business overview

  8. 8. Perplexity Computer

    Perplexity

    Good for
    Perplexity Computer breaks an outcome into tasks that sub-agents run in cloud sandboxes, and it is built for work where the deliverable is cited research: dossier reports and market studies that combine web sources, apps and files. A multi-step research report lands around 350 to 950 credits.
    How it works
    General agent, web, mobile, Comet browser and Mac app.
    Runs on
    Third-party model orchestration on Perplexity’s own harness. Models not named on official pages.
    Price
    Computer usage is metered: Pro $20 a month includes a weekly allowance (rolling out), Max $200 includes 10,000 credits, and 100 credits cost $1.
    Choose it when
    The deliverable for the research phase is a cited brief, and you want to price a small research job from published credit costs first.

    Sources: How credits work on Perplexity

Pricing and how to try it

Plans start at $5 a month, and Starter and Plus begin with 7 days free. On our internal PI Knowledge Work benchmark, Ultramarine 1 Flash costs about $0.13 a task, about a tenth of Claude Opus 5.5 or GPT-6 Sol at comparable scores.

Start free trial See every plan

Questions

Can an AI tool write an investment memo straight from a data room?

No single tool writes a committee-grade memo straight from a data room. Hebbia and Blueflame turn hundreds of documents into cited facts, AlphaSense contributes the market evidence, and drafting agents assemble the prose from those inputs. The deal team owns the thesis and verifies every figure before the memo circulates.

What does each tool actually do in the memo process?

Rogo and AlphaSense feed the market section with research, transcripts and comps. Hebbia and Blueflame extract and analyze the data-room documents. Agents that work in files, among them Ultramarine Work, Claude and ChatGPT Work, build the cohort tables, the operating model and the memo draft.

How can we keep every number in the memo traceable?

Require citations to the source file and page, run a tie-out pass before circulation, and keep an exception log for anything that does not reconcile. Ultramarine Work’s skills have it cite the source file behind each figure, and a person reviews every file the agent touched.

Are these tools safe for confidential deal information?

Not without a firm-level review, because data handling differs by product, plan and workspace settings. Ultramarine Work changes files only in the project’s folder. When a request needs it, it can read files elsewhere on your computer, never passwords, keys or app data, and you can turn that off in settings. The model runs on Paper Instruments’ servers, so the content it works on is sent there to be processed. Paper Instruments is SOC 2 Type II, and our security and compliance details are on our compliance page.

What will this stack cost?

The specialist platforms publish no prices and sell through sales, so budget for a procurement cycle around seat counts. Ultramarine Work starts at $5 a month. Claude and ChatGPT have free tiers, with paid plans from $20 a month. Perplexity Pro is $20 a month with metered credits for Computer tasks.

Should a specialist platform replace our general agent?

Choose a specialist platform when the bottleneck is extraction, curated content or standardized workflows across the whole firm. Choose a general agent when someone has to draft, model and check actual files. Most firms run both: the memo’s evidence comes from one layer and its prose from the other.

Sources

Checked on