Iris Assistant¶

Iris is the in-app AI assistant. It helps you find your way around InvestLens, explains what the analytics mean, runs many of them directly in the conversation, and can propose changes to your portfolios and ideas that you approve before anything happens.

Iris provides platform guidance, analytics, reports, rankings, and model-based insights. It does not provide personalized financial advice or investment recommendations, and it will not tell you to buy, sell, or hold a security.

1. Where to find Iris¶

Iris has two surfaces, and they are two views of the same conversation — not separate chats.

Chat widget: a floating button in the corner of every signed-in page, so you can ask a question without leaving the page you are on. Click the button to open the panel; close it with the same button, the panel's close button, or the Escape key. On desktop the panel can be moved and resized, or expanded to full screen. On phones and tablets it always opens full screen.
Iris Terminal: a dedicated full-page view at https://app.investlens.ca/chat, reachable from the Iris Terminal entry in the sidebar (just below Contact us) or the Open in Iris Terminal button in the chat panel header.

Whichever surface you type in, the reply joins the same conversation and both views show the same transcript. Switching conversations in one place switches it in the other, including in a second browser tab. The conversation also stays put as you move between pages during a session — reopening Iris shows the prior exchange rather than starting over.

2. Iris Terminal¶

The Terminal is the same assistant with more room. Nothing is available there that the widget cannot do, and vice versa, so the choice is about space, not capability.

You may prefer it when:

Wide results are hard to read in a narrow panel — tables from style and factor analysis, optimal allocation, and rankings get the full page width.
You are scrolling back through a long session.
You want Iris in its own browser tab, open alongside a portfolio or report.

The widget's floating button is still present on the Terminal page. Clicking it does not open a second chat — it moves the cursor to the page's message box, since both would be typing into the same conversation.

3. What you can ask¶

What features are available, and where to perform a task.
How to analyze portfolios, review performance and risk, and compare against a benchmark.
How to read factor exposures and generate reports.
How to use backtesting, optimization, asset snapshots, rankings, sharing, settings, and account pages.
What a metric means and how it is calculated.

4. Analytics Iris can run in chat¶

Iris can carry out and summarize many analytics without leaving the conversation:

Portfolio performance and risk metrics — returns, volatility, Sharpe ratio, drawdown.
Style and factor analysis — Fama-French five-factor exposures.
Optimal asset allocation — your current weights against max-Sharpe, minimum-variance, and equal-weight, with the full risk and performance breakdown for all four variants.
Historical backtests over a chosen period, compared against a benchmark.
Idea generation via the Universe Screener — by theme, profile, or custom criteria.
Asset snapshots, recent news sentiment, and upcoming or historical earnings for a ticker.
ETF fund facts and top holdings — an ETF's largest positions, plus yield, expense ratio, assets under management, and the index it tracks.
Kicking off a written portfolio analytics report, delivered when it is ready.

All of these apply the same data-quality rule used on the portfolio pages: holdings without usable price history are left out, the remaining weights are rescaled to total 100%, and each exclusion is reported with its specific reason. If no holding has usable history, Iris explains that the market data may still be downloading and asks you to try again shortly rather than reporting figures.

5. Charts in chat¶

When figures read better as a picture than as a table, Iris draws a chart in the conversation — titled above the plot, with labelled axes. Charts appear on both surfaces and look the same in each.

A chart is drawn from the figures in the reply it accompanies, so it always shows exactly the numbers Iris just described. The written text stays the authoritative version of the answer.
Iris only charts figures it actually retrieved on that turn. Asking for a chart of something it has no data for gets an explanation rather than an invented picture.
The chart type follows the data: line or area for anything over time (a return series, a drawdown, growth of an amount), bar for summary figures compared side by side (Sharpe ratio, volatility, total return), and pie for a breakdown that adds up to a whole (allocation by holding, sector, or asset type). You can ask for a particular one.
Pie slices are labelled with name and percentage, ordered largest first, with a long tail grouped into a single Other slice — past about eight wedges a pie stops being readable. For values that are not parts of one total, Iris uses a bar chart, which reads more precisely.
Charts are meant for compact series: up to five series at a time, each a couple of hundred points. A longer series is condensed by sampling evenly across it, so the chart still covers the whole period at lower resolution rather than being cut short at the start.
Charts are read-only pictures — no zoom, pan, or hover read-out. A bar chart prints the value above each bar so it can be read directly.
One reply can carry up to five charts (a budget it shares with snapshot cards) when separate pictures read better than one crowded plot.
A chart belongs to the turn that produced it. It stays with that turn as you move between pages, and reopening the conversation from history redraws it in place.
If a chart cannot be drawn, a short note appears where it would have been. The written answer is unaffected.

6. Asset snapshot cards in chat¶

When a question is about a single security, Iris can place an asset snapshot card in the conversation — the same at-a-glance card the Asset snapshot action opens from My Assets, in a compact form that reads inside a chat message.

Each card is headed by the asset it describes (ticker, plus company name once available) and closes with a line crediting the data source. The heading matters when one reply carries several cards.
A card shows the analyst consensus rating and contributor count, the breakdown of recommendations, valuation and beta, earnings surprises over the past year, the 52-week price range with markers for current price, analyst target, and 50/200-day moving averages, and a News Sentiment reading.
A card holds a fixed readable width rather than stretching, so its rating dial and price bars stay legible on a wide display. Charts behave the other way round and take the full width, because a series over time genuinely reads better wider.
Figures are fetched fresh when the card renders rather than passed through the conversation, so a card always shows the current snapshot. Cards load independently, so several in one reply fill in as their data arrives.
Charts and snapshot cards share one budget of up to five visuals per reply, so a comparison across a few tickers can sit side by side.
If no snapshot is available for a ticker, the card says so plainly while still naming the asset.

7. What Iris used to answer¶

Each reply carries a small grey line underneath it naming the capabilities used to compose that answer — for example "3 data lookups · ticker search, news sentiment". Hovering the line breaks it down by source, with the number of calls to each.

Turns where nothing was retrieved say so plainly ("Answered from conversation context — no data lookups"). This is normal: resolving which portfolio you meant, explaining what a metric means, asking a clarifying question, or confirming an action all need no lookups.
One case is flagged in red with a warning icon: a reply that states several figures while retrieving nothing on that turn. The numbers were carried over from earlier in the conversation and may no longer be current — worth asking Iris to re-run the analysis if freshness matters.
Conversations recorded before this existed show no line at all, rather than claiming zero lookups.
The line is not part of the message, so it is left out of copied messages and downloads.

8. Actions Iris can perform¶

For anything that changes your account, Iris does not act on its own. It proposes the action as a confirmation card showing exactly what will happen, and the change is applied only after you approve it. Supported actions:

Create a new portfolio.
Update an existing portfolio — rename it, rebalance weights, replace the full set of holdings, or make targeted changes such as adding or removing specific holdings while leaving the rest untouched.
Save a generated Screener idea to your account.
Delete a portfolio you own.
Delete a saved idea you own.

How the confirmation works:

Iris presents a card describing the proposed action — portfolio name, holdings, weights, or idea details — with Confirm and Cancel buttons.
Nothing is created or changed until you click Confirm. Cancel discards the proposal.
When more than one item is proposed, you choose which to approve. Each approved action runs once.
The outcome of each action is reported back into the conversation. Anything that could not be completed (for example, a portfolio name that already exists) is reported with the reason, so Iris can follow up — such as suggesting a different name.
After a successful action, Iris can offer to take you straight to the result: the new or updated portfolio or idea, or the relevant list after a delete.

Proposing a portfolio or idea is portfolio construction carried out at your request and only after you confirm it. It is not investment advice.

9. Dates and times¶

Iris uses the timezone set in My Profile, so "what day is it", "today", "yesterday", and "this week" mean what you mean by them — including for date-bound work such as earnings windows and intraday ranges.
If you have never saved My Profile, Iris falls back to the zone your browser reports, so "today" still means your local day rather than a server day. No setup is needed for this. Once My Profile is saved, that zone wins.
UTC is used only when no zone is known at all. In that case Iris says it is answering in UTC rather than guessing.
You can always ask about another zone explicitly — for example market hours in Tokyo — and Iris answers in that zone regardless of your own setting.

10. Conversation controls¶

Copy any message to the clipboard, formatted, with a plain-text fallback.
Download a single message as a Markdown file.
Download the entire conversation as a single Markdown file.
Start a new conversation, which clears the chat and begins a fresh thread.

Charts and snapshot cards are rendered pictures rather than text Iris wrote, so they have no copy or download button of their own and are left out of a downloaded conversation. Exports carry the written answer they illustrate.

11. Conversation history¶

Past conversations are saved to your account and can be reopened later. Both the widget and the Terminal have a Conversations panel that slides in over the left edge of the chat (full width on phones); it closes with its close button, the Escape key, or a click outside it.

In the panel you can:

Browse past conversations, grouped by when they were last used (Today, Yesterday, Previous 7 days, Previous 30 days, then by month). Each row shows when it was last used — a clock time for today and yesterday, a date for older ones — with the exact date and time on hover.
Search across your conversations. The search covers conversation names and the text of the messages inside them, and each result shows a short excerpt with the matched words highlighted.
Open a conversation to continue it. The full transcript is shown and the next message carries on from there.
Rename, keep the current name, or delete a conversation and its transcript, from the "..." menu on each row.
Start a new conversation.

Deleting asks for confirmation first. The prompt names the conversation and states that the transcript is removed for good; Yes deletes it, No leaves it alone, and Escape dismisses it. Deleting the conversation currently open clears the chat and starts a fresh one.

11.1 Naming¶

Conversations are named automatically, and the name keeps up with the discussion: Iris re-summarizes the thread as it develops, so an early name can be replaced by one that better reflects where the conversation ended up. If a name cannot be produced, the thread falls back to the first few words of your first message. Names do not include the date — when a conversation was last used is shown by the list's own date grouping.
Renaming a conversation makes the name yours: Iris stops retitling it and it stays exactly as typed. To hand the name back to automatic titling, rename it to an empty name.
Keep this name in the "..." menu does the same thing without retyping: checking it freezes the name Iris has given the thread so far, and unchecking it lets Iris resume renaming. Renaming checks it automatically. Toggling it only affects that thread's name — it does not open the conversation or change which one you are in.

11.2 What to expect¶

Reopening the chat, reloading the page, or returning to a tab picks up the conversation that is currently active — you are not dropped into an empty chat.
Conversations become read-only after 30 days of inactivity. The transcript stays fully readable, but the message box is replaced with a note and a New chat button. This is deliberate: after that long the assistant no longer retains the thread's context, so a follow-up would be answered without any memory of the earlier discussion. Read-only conversations are marked in the list and can still be renamed, downloaded, or deleted.
A very long conversation is remembered only in part: Iris works from the most recent stretch of the exchange, while the full transcript remains visible to you.
Date groups and per-row times follow the timezone set in My Profile, so "Today" means your calendar day. Changing the profile timezone shifts the groupings and times accordingly.

When Iris suggests a particular page, it can offer a Take me there button that goes straight to it, so you do not have to find it in the menu.
The same shortcut appears after a confirmed change, so you can jump to the result — the created or updated portfolio or idea, or the appropriate list page after a delete.
Contextual page tips work with Iris too: clicking a tip drops its starter question into the message box, ready to send or edit. Tips can be turned off in General settings.

13. Notes and limitations¶

Iris provides platform guidance, analytics explanations, reports, rankings, and model-based insights.
Iris does not provide personalized financial advice or investment recommendations, and does not tell you to buy, sell, or hold securities.
Charts, snapshot cards, and every other output are descriptions of data, never a signal.
Figures quoted from earlier in a conversation may be stale. The grey capability line under a reply tells you whether anything was retrieved on that turn.