Start here
What Kai can do
Everything Kai can do for you, in one place. Kai is one assistant that works across every part of Kaiday — this page shows what that means in practice, with the kind of sentences you'd actually say to make each thing happen.
Every interaction looks the same
You describe what you want, in your own words. Kai figures out how to do it, drafts the work, shows you the result, and keeps a record in Automations so you can always check what happened. You can stay involved at every step or let Kai run with it — your call.
One assistant, many things at once
The per-area lists further down are useful as a checklist. But the real story is what happens when Kai pulls several of them together in a single request. Every card below is one sentence you say to Kai — and what comes out the other end is the kind of work that used to need a whole team.
From portfolio data to a presentation
You say
"Pull last quarter's portfolio exposures, attribution, drawdowns, and benchmark deltas — and build a 12-slide review deck in our brand template. Save the raw numbers in a backup sheet too."
Kai does
Kai looks up the portfolio data, makes the charts, applies your brand template from the Knowledge Base, and saves both the deck and the sheet. Every chart shows its calculation and as-of date.
From research defect to reviewed fix
You say
A reviewer reports: "the temporal holdout includes future observations." You: "Verify it, fix it, and draft a reply once the tests pass."
Kai does
Kai reproduces the leakage, finds the faulty date boundary, drafts the fix with a regression test, and prepares the reviewer reply. You or your developer approves the change.
From scattered evidence to a review portal
You say
"Collect the evidence behind the quality-factor study, group it by claim, and build a review portal at research.example.com/quality."
Kai does
Kai organizes sources, links each claim to evidence and limitations, builds the MiniApp, and publishes it on your domain for reviewers.
From meeting to action items, without asking
You don't ask Kai to do anything. You just have a Kaiday meeting. By the time it ends, every attendee has a follow-up email with the decisions and next steps. The full transcript and action items are saved with the meeting.
From study state to reviewer follow-ups
You say
"Every Friday at 5pm, find studies waiting for review, send each assigned reviewer a limitation summary, and flag anything untouched for seven days."
Kai does
You set it once. Every Friday at 5pm, it runs. The Automations tab records every draft, notification, and escalation.
From review thread to investment memo
You say
A committee thread reaches the decision point. You: "Draft the investment memo from this thread, separate facts from assumptions, and include the dissenting view."
Kai does
Kai pulls the approved memo structure, fills it with cited evidence and risks, and saves a draft for committee review.
From review queue to a committee agenda
You say
"For Tuesday, schedule 45-minute reviews for every study awaiting committee decision. Order them by risk and attach the latest evidence and open objections."
Kai does
Kai finds open time, orders the agenda by risk, and attaches a concise study brief to each calendar event.
None of these are templates. None of these are workflows you have to build first. They are just sentences. Kai figures out which parts of Kaiday to touch and in what order.
Communicate
Draft replies in any thread with relevant study, evidence, and review context.
Handle multi-message conversations end-to-end. Track what's asked, answered, and still open.
Compose new emails from a one-line description.
Sort incoming emails into categories (data, research, risk, operations, noise).
Run scheduled follow-up email sequences (every 3 days, every Monday, whatever you describe).
Escalate threads that need human methodology or risk judgment.
Messaging
Summarize long threads into actual decisions and open questions.
Draft replies in any channel, in your voice.
Search every channel and DM by meaning, not keyword.
Turn a chat decision into an Email or Calendar invite.
Try saying to Kai
- ›Handle this data-provider thread, summarize the coverage gap, and stop before accepting new terms.
- ›Summarize the decision in #research and draft the reviewer follow-up.
- ›Ask reviewers for decisions on studies waiting more than seven days.
Schedule & meet
Calendar
Create events and recurring meetings from a sentence.
Find open time across calendars and schedule meetings.
Reschedule whole sets of meetings ("move all my Wednesdays to next week").
Prep notes before every meeting: who's coming, what changed since last time, what to ask.
Video meetings
Run meetings on Kaiday's own self-hosted video stack. No third-party app to install.
Record meetings automatically.
Transcribe with speaker labels.
Turn a recording into action items and follow-up emails.
Try saying to Kai
- ›Book a 30-minute intro with the lead from the intake form, prep me with their answers.
- ›Record this call and send the action items to the client afterward.
- ›Move all my Wednesday meetings to Thursday next week.
Write — documents, sheets, presentations
From a sentence to a saved Knowledge Base document
You say
"Write a one-page methodology guide for new reviewers: evidence standards, holdout requirements, escalation rules, and where to record decisions."
Kai does
Kai drafts the guide, saves it in the Knowledge Base, and shows you the link. It is immediately searchable, editable, and shareable with the research team.
GIF placeholder
Documents
Create documents from a description: memos, proposals, contracts, internal guides.
Edit existing documents, rewrite a section, restructure, tighten, expand.
Translate documents into other languages.
Summarize long docs into a one-page brief.
Sheets
Build spreadsheets from a description ("portfolio exposure by sleeve, monthly columns").
Run analyses you describe in English; chart the result.
Pull data from a connected database or a CSV file and reconcile it against your records.
Presentations
Generate decks: investment-committee reviews, research updates, board reports, and methodology training.
Edit slides by talking: "tighten the headline on slide 6, remove slide 4."
Generate decks from meeting notes, a memo, or a one-paragraph brief.
Restyle a whole deck to a new template or brand.
Try saying to Kai
- ›Generate a 10-slide investment-committee deck from the approved study package.
- ›Turn yesterday's methodology review transcript into a one-page decision memo.
- ›Add a sensitivity-analysis slide to the portfolio review deck from the Knowledge Base.
Build financial workflow apps
Build MiniApps from a description: hypothesis intake, evidence portals, calculators, risk trackers, and review dashboards.
Host financial workflow apps under your verified domain (e.g.,
research.example.com).Update an app by talking: "add transaction costs to the review form."
Connect apps to approved research data and namespaced app storage.
Change the app whenever you want, just by talking — Kai re-publishes it for you.
See the full MiniApps section.
Try saying to Kai
- ›Build a hypothesis intake with benchmark, costs, and rejection criteria.
- ›Add a limitation field and reviewer decision to the study dashboard.
- ›Publish the review app at research.example.com.
Code — Kai writes pull requests on GitHub
This section is for people who use GitHub
If you don't write code (or work with someone who does), you can safely skip this section. The short version: if you ever do connect a code repository, Kai can read it, fix bugs, and propose changes for your developer to review — never pushing anything live without approval.
Connect a GitHub repository and Kai writes code for you. Not autocomplete — real, end-to-end pull requests. Kai reads your code, figures out the problem, writes the change, runs the tests, and opens a draft pull request for you (or your developer) to review. You review, you merge. Kai never publishes anything to your main branch on its own.
GIF placeholder
Your review queue now drives fixes
You say
Reviewer note: "the temporal holdout contains observations from the training window." You: "verify it and draft a fix."
Kai does
Kai checks the study metadata, reproduces the leakage, finds the faulty date boundary, and opens a draft pull request with the fix, a regression test, and a link to the reviewer note. You review and merge.
Open real pull requests on GitHub. The code change, the tests, and a clear description — all written for you.
Diagnose bugs from a reviewer report, an error message, a screenshot, or a failed study run.
Read your code to answer questions like "where is survivorship bias controlled?" or "why is this study slow?"
Reorganize code on demand: "this file is too long, split it into smaller files."
Review pull requests your teammates open and flag anything risky before you merge.
Trace a research defect back to the line of code most likely responsible.
Never publishes code directly. Never touches your main branch without you merging it yourself.
A research tool, updated from your phone
You say
Airport. You message Kai: "add a log-scale toggle to the factor chart and remember the reviewer's choice. Draft only — don't ship it."
Kai does
You board. By the time you land, the change is waiting for you to review. You approve it from the rideshare. The feature is live before you reach the hotel.
Try saying to Kai
- ›The temporal holdout includes training-window observations. Find and fix it.
- ›Review pull request #421 and tell me what's risky.
- ›Why did the backtest runtime double? Find the cause and propose a fix.
- ›Add a log-scale toggle to the factor chart. Draft only — don't publish it.
Ask your data anything, get whatever you need back
Connect an approved portfolio, market, or research database and Kai treats it like part of your workspace. You ask in plain English. The answer comes back as a number, chart, slide, sheet, email, or all four at once.
The examples below use portfolio and research metrics. Every result should carry its source, calculation, assumptions, and as-of date.
Tuesday 10am, an answer in 30 seconds
You say
Committee email: "what drove active return over the last 12 months?" You: "Pull attribution by sleeve and month, chart it, and draft a reply with the calculation notes."
Kai does
Kai produces the sourced chart and a draft reply for your review. The calculation notes stay attached.
Answer questions about your data in plain English. No SQL, no BI tool.
Turn data into the format you actually need. A chart for a meeting, a slide for a deck, a sheet in the Knowledge Base, a chart inside a reply email.
Build recurring reports that run themselves on a schedule and land wherever you want them.
Compare two database sources and flag mismatched records.
Show exactly how each number was calculated, so you can trust it (and explain it to anyone who asks).
Read-only by default. Kai never changes your database unless you explicitly turn that on per connection.
Friday 5pm, the weekly digest writes itself
You say
You set it once: "Every Friday at 5pm, summarize exposure changes, drawdowns, benchmark deltas, and open risk exceptions. Post it in #portfolio-review and draft the committee email."
Kai does
Every Friday, the team gets the sourced review package and the committee email waits for approval.
Try saying to Kai
- ›Pull last quarter's attribution and risk metrics and build a 12-slide review deck. Show how each number was calculated.
- ›Chart active return by sleeve for the last 12 months and draft a reply with the source notes.
- ›Every Friday at 5pm, post a summary of exposure changes and open risk exceptions in #portfolio-review.
- ›Compare these two market datasets and flag mismatched observations.
Automate everything else
Build a workflow from one sentence. Kai turns "when X happens, do Y" into a real automation.
Schedule recurring tasks (daily data checks, weekly reviews, monthly risk reports).
Trigger an action when an app records a new hypothesis, review, exception, or submission.
Retry, cancel, or jump back to the chat for any task that didn't work.
Try saying to Kai
- ›When a study reaches review, DM the assigned reviewer with its main limitation.
- ›Every Monday at 7am, post a digest of completed studies to #research.
- ›Anything that fails twice in a row, escalate to Messaging and ping me.
Manage your team
Onboard new teammates. Send the invite, set their role, walk them through the first run.
Manage teammate roles and surface permissions in one workspace.
Hand off work between teammates. Kai keeps the thread of context.
Try saying to Kai
- ›Invite Jane Smith as an admin and onboard her into the workspace.
- ›Show me what's pending across the team.
- ›Hand off the momentum study to the model reviewer with all the context.
The boundary
What Kai will not do
Publish code changes directly without your review.
Process payments or execute trades from MiniApps.
Delete records permanently without a confirmation.
Talk to systems you have not explicitly connected.