Today we are launching AI Insights - a new AI-powered analysis feature for backtesting reports. AI Insights replaces AI Summary and represents a significant step forward in how we use AI to help you understand your portfolio.
AI Insights is currently in beta. This is the first public version and we will be continuously improving it based on your feedback and our own research.
What Is AI Insights?
AI Insights analyzes your backtesting report and produces three types of output:
- Portfolio Risk Score - a deterministic composite score from 1 to 7 that characterizes your portfolio's risk profile, from Capital Preservation to Aggressive Growth
- Characteristic Summary - a concise narrative describing how your portfolio behaves overall, covering growth, volatility, drawdowns, diversification, and benchmark relationship
- Key Findings - a prioritized list of observations across your report, categorized as Strengths, Warnings, and Observations, each backed by specific metrics from your report
This is meaningfully different from AI Summary. Rather than producing a single text summary, AI Insights reasons across multiple dimensions of your report in parallel and then synthesizes a structured analysis. The result is more detailed, more actionable, and easier to navigate.
Why It Takes Longer
AI Insights applies a multi-stage reasoning pipeline. It analyzes your portfolio across three groups - performance and risk, composition and diversification, and value projections - before synthesizing the results into a final structured output.
This requires multiple calls to frontier-class large language models, which are among the most capable AI models available today. We plan to continue using the latest models as they improve.
As a result, a single run takes approximately 30 seconds to a few minutes depending on the size and complexity of your report. We know this is slower than AI Summary, and it is a deliberate trade-off: deeper reasoning takes more time, but the analysis you get back is substantially more useful.
Multi-Language Support
AI Insights produces output in the language you have selected in the app. Whether you are using Portfolio Metrics in English, German, Czech, Japanese, or Chinese, the analysis will be generated in your language.
New Pricing: AI Credits
AI Insights introduces a new pricing model for AI features: AI Credits.
1 AI Credit = $1. Credits are consumed per run, and the cost of a run depends on the size of your report - more assets and more portfolios in a single report means more analysis and higher cost. A simple single-portfolio report will consume less than a complex multi-portfolio comparison.
This approach is more flexible than a fixed monthly limit. Rather than a hard cap on the number of runs, you get a credit budget that scales with how you use the feature.
Free plan: No AI credits included. AI Insights is available on Premium and Professional plans.
Premium: 2 AI Credits per month, enough for approximately 10 typical runs.
Professional: 10 AI Credits per month, enough for approximately 50 typical runs.
Availability
AI Insights is currently available in the Backtesting tool. Support for Optimization and Monte Carlo reports is coming soon.
How It Works
For those interested in the technical side, here is how AI Insights works under the hood.
Stage 1: Deterministic Risk Scoring
Before any LLM is involved, AI Insights computes a deterministic Portfolio Risk Score from four key metrics:
- Max Drawdown - the largest peak-to-trough decline (35% weight)
- Annualized Volatility - standard deviation of returns (25% weight)
- CAGR - compound annual growth rate, where higher return increases the score (20% weight)
- Ulcer Index - depth and duration of drawdowns combined (20% weight)
Each metric is independently scored on a 1-7 scale using fixed thresholds. The weighted total produces a composite score, also on a 1-7 scale, which maps to one of seven risk profiles:
| Score | Profile |
|---|---|
| 1 | Capital Preservation |
| 2 | Conservative |
| 3 | Moderately Conservative |
| 4 | Balanced |
| 5 | Balanced Growth |
| 6 | Growth |
| 7 | Aggressive Growth |
This deterministic score is computed first and passed as structured context into the LLM analysis. It anchors the AI's risk characterization to objective data.
LLM Analysis
The risk score feeds into a frontier-class large language model that reads your full report and produces the key findings. Each finding is categorized as one of three types:
- Strengths - notable positives worth building on, such as strong risk-adjusted returns, efficient frontier positioning, or high probability of positive outcomes
- Warnings - risks or concerns that deserve attention, such as severe drawdowns, high volatility, or tail risk
- Observations - neutral context that helps interpret the results, such as benchmark relationships, diversification patterns, or outcome dispersion
Each finding is backed by specific metrics from your report, so you can trace exactly what the analysis is based on.
The combination of deterministic scoring and LLM reasoning is intentional. Deterministic scores provide reliable, objective anchoring that keeps the analysis grounded. The LLM provides the reasoning and language to turn those numbers into meaningful insight.
We are actively working on improvements to AI Insights. If you have feedback, we would love to hear it at support@portfoliometrics.net.