Inside Portfolio Companion: What Our AI Agent Actually Does Today

You’ve seen “AI analyst” claims before. This post explains, concretely, what Nagare’s Portfolio Companion can do for you right now.

We’ll walk through three workflows:

• Turning GP reports into updated actuals
• Answering portfolio questions in plain English
• Creating and running scenarios without rebuilding Excel

All of this runs on top of the same data and calculation engine that powers the rest of Nagare.

---

1. GP Reports → Updated Actuals in Minutes

Problem: Quarterly reporting season means 20+ PDFs, hours of manual data entry, and easy-to-miss mistakes.

With Portfolio Companion:

1. You upload a batch of GP reports to Nagare.
2. Portfolio Companion reads each document, identifies capital calls, distributions, and NAV updates.
3. It proposes a set of transactions per fund for you to review and approve.

Example prompt:

“Extract actuals from these 5 GP reports and update the underlying funds.”

The agent will:

• Parse each PDF (including scanned documents).
• Normalize dates, amounts, and currencies.
• Match GP fund names to your internal fund IDs.
• Propose a list of capital calls, distributions, and NAV marks.

You stay in control:

• You see a structured proposal.
• You approve or adjust.
• Once approved, Nagare updates the transaction and NAV data used everywhere else.

Result: the Funds and Fund Detail screens reflect the latest actuals without manual typing.

---

2. Ask Portfolio Questions in Plain English

Once your data lives in Nagare, Portfolio Companion can answer questions that would normally require building a report or exporting to Excel.

Examples:

• “Show me funds with IRR below 15% and TVPI below 1.5x.”
• “Compare my 2021 vintages to 2024 vintages on deployment, exits, and DPI.”
• “What’s my exposure to US growth equity across private funds and public holdings?”
• “Find funds where deployment is slower than expected vs their plan.”

Under the hood, the agent:

• Uses Nagare’s own APIs and query layer.
• Respects tenant boundaries and permissions.
• Returns answers grounded in the same definitions your team already uses (no shadow spreadsheets).

You get:

• Simple, readable answers.
• Links back to the underlying funds or scenarios.
• A natural way to explore your portfolio without building a new report each time.

---

3. Scenario Creation Without Rebuilding Excel

Portfolio Companion can also help you create and run scenarios on top of the existing modeling engine.

Examples:

• “Create a bear case with 20% slower deployment and delayed exits for my 2021 vintage funds.”
• “Model a $15M co-invest into Fund XYZ and show the impact on NAV and liquidity.”
• “Show downside if equity markets drop 30% over the next 18 months.”

What happens:

1. The agent translates your request into a scenario definition using the same parameters available in the UI.
2. It adjusts deployment curves, exit timing, or market assumptions as requested.
3. It triggers a calculation run and surfaces the results back in the chat and in the Scenarios UI.

You still decide which scenarios matter, but you don’t have to click through every parameter screen just to test an idea.

---

4. Guardrails: How We Keep AI Safe and Predictable

Portfolio Companion doesn’t get raw database access or free-form write permissions.

Instead, it works through:

• Tenant-scoped APIs and tools – every call is executed with your tenant context.
• Narrow tools – “extract fund actuals,” “create scenario,” “run Monte Carlo,” etc.
• Human approval – for actions like updating transactions or creating scenarios, it proposes changes before applying them.

This keeps the agent helpful and powerful, without turning it into an unbounded script engine.

---

5. How to Get Value in Your First Week

If you’re just getting started:

1. Pick 3–5 funds and upload their latest GP reports.
2. Ask Portfolio Companion to extract actuals and propose updates.
3. Validate one or two funds against your existing spreadsheet.
4. Ask 3–4 portfolio questions you’d normally answer in Excel.
5. Create one bear scenario and one co-invest scenario and compare them in the UI.

That’s enough to see how the agent can remove repetitive work while staying tied to real, auditable numbers.

---

Where to Go Next

• Learn how the extraction pipeline works in detail: How AI Extracts Data from GP Reports
• See how the modeling engine is validated: How We Validate Our Monte Carlo Engine
• Dive into the math: Inside Nagare’s Financial Modeling Methodology

When you’re ready, we’re happy to walk through your own GP reports live and compare Portfolio Companion’s output to your current process.