The AI That Takes Notes While It Works
Most AI tools start every conversation from zero. You explain that "revenue" excludes refunds, that the status column has a quirk, that your fiscal year starts in April — and the next day you explain it all again.
Phaide AI's Memory removes that repetition. As the AI analyzes your data, it writes down what matters: data caveats, table quirks, business definitions, and company facts it discovers along the way. Each note is saved under a topic — for example "Company information" or "Analysis caveats" — and stored as part of your organization's context layer, the shared body of knowledge the AI consults before answering anything.
Memory is one of three things the context layer assembles for every request: the schema automatically extracted from your ingested data, the knowledge files your team uploads, and Memory — the running notes the AI keeps for itself. Together they mean the assistant is never guessing about your business from scratch.
Why a Memory Changes Every Answer That Follows
Memory is organization-wide. A note the AI records during one analyst's chat is available to everyone on your team, in every chat, dashboard, and automated exploration that follows. The context is injected, in full, into the AI's instructions on every request — so the knowledge compounds instead of evaporating when a session ends.
That produces three concrete benefits:
Consistency. When the AI has written down that "active users" means logged in within 30 days, every future answer uses the same definition. No more two dashboards quietly disagreeing because two people phrased the same question differently.
Less repetition. The caveats you would otherwise re-explain — a column that's null before 2023, a region code that maps to a legacy market — are remembered once and applied forever.
Knowledge that survives. Findings from your deeper investigations don't disappear. When someone uses dig-down to dive into a result, the AI's written summary of that answer is saved to Memory too, timestamped, building an institutional record of what your data has already taught you.
Because the notes are the AI's own working memory rather than a buried log, they are fully visible to you — and fully editable. If the AI records something imprecise, you correct it once, and every answer afterward inherits the fix.
Where to Find and Shape Your AI's Memory
Memory builds itself as you use Phaide AI, but you stay in control. Here's how to review and refine it:
- Open Knowledge from the left sidebar.
- Scroll to the Other context section at the bottom of the page.
- Click the Memory card (marked with a brain icon, next to Schema).
- A dialog opens listing every memory the AI has kept, each shown by topic with its content and the date it was last updated.
- To refine an entry, click Edit on any topic, adjust the text in the editor, and click Save. A confirmation appears, and the updated note takes effect on the AI's next response.
There's nothing to configure to get started. The first time the AI notices something worth remembering during a chat or exploration, the note appears here automatically. Your job is simply to curate — sharpen a definition, remove something out of date, or add the nuance only your team knows.
The same page is where you manage the rest of the context layer: the read-only Schema generated from your ingested data, and the knowledge files (.txt and .md) you can upload or write by hand. Memory is the part that grows on its own.
Give Your Data a Memory
The best analyst on your team is the one who already knows your business — its definitions, its exceptions, the lessons learned the hard way. Memory gives Phaide AI exactly that kind of accumulated understanding, and shares it across everyone you work with. Open the Knowledge page and see what your AI has already learned.
Frequently Asked Questions
Does Memory turn on automatically, or do I have to set it up? It's automatic. The AI records notes on its own as it analyzes your data — there's no setup step. You only need to open the Knowledge page if you want to review or edit what it has remembered.
Who can see and edit the AI's memory? Memory is shared at the organization level, so it's visible to everyone in your workspace, and any member can open the Memory dialog to review and edit the notes.
Will the AI actually use these notes when it answers? Yes. The full set of memories is included in the AI's instructions on every chat, dashboard build, and exploration — so the context shapes every response.
Can I correct something the AI got wrong? Absolutely. Open the Memory card, click Edit on the relevant topic, change the text, and save. Your edit replaces the note and applies to all future answers.
What kinds of things does Memory store? Typically data caveats, table and column quirks, business definitions, company facts, and summaries from your deeper dig-down investigations — anything the AI judges useful to remember for later analysis.
How is Memory different from uploaded knowledge files? Knowledge files are documents you provide; Memory is the notes the AI writes for itself as it works. Both live in the context layer and feed the AI's answers, but Memory grows automatically while files are added by your team.
Does Memory carry over between separate chats and dashboards? Yes — that's the point. Memory is not tied to a single session. Once a note is recorded, it's available across all chats, dashboards, and explorations in your organization.