Answers
The important context is in my team's heads, not in any file. How does it reach the AI?
By being said once, in the work, and kept where every assistant reads it. Knowledge that lives in people's heads reaches an AI only when someone states it, and the moments people state it are ordinary ones: correcting an answer, explaining a rule to a new colleague, deciding something in a review. In Arroway the assistant records what it was told or what was decided as a proposal while it works, and a person on the project approves it, so the next session, whoever starts it, begins from that rule with its approver and the condition that ends it. Material that already exists, such as a list of rules or notes, can be imported once and reviewed in one sitting. Arroway does not interview people and cannot extract what nobody has said: a rule nobody has stated stays in someone's head until it is stated in a session. The install page is the first step.
Last updated October 30, 2026
Why interviewing everyone once does not fix it
The first idea is to sit down with the people who know the work and write it all out. It produces a document that is a snapshot: accurate that week, silent about the rules nobody thought to mention, and out of date as soon as something changes. The knowledge that matters most tends to be the least written, because the people who hold it no longer notice they are applying it. A year-end rule, a client who must never be called on Mondays, the reason a figure excludes refunds: these come out when someone is wrong about them, not when someone is asked for a list.
Let it enter at the moment it surfaces
The moment a rule surfaces is usually a correction. The assistant computes revenue with refunds included, the analyst says it never counts refunds booked after month end, and the assistant records that as a proposal with the reason the analyst gave. The analyst, or whoever on the project holds that area, approves it and says what would end it, such as a change in the finance calendar. From then on every assistant that opens the project is told the rule is in force, with its approver. For what is already written down, an import takes the rows you trust, and the review afterwards is often the first time anyone reads the whole list together. The memory grows from the work that exposes the gaps, and a person decides what counts.
When to use Arroway
Use Arroway when the team's AI keeps being corrected on the same unwritten things, or when one person's departure would take rules with them. Start by importing what exists and then let the assistants record corrections as they happen. Arroway does not interview anyone, does not infer what was never said, and does not know which of the rules in people's heads are the true ones; a person approves each. The first step is the install page.
What it looks like in practice
A four-person analytics team keeps its definitions in the heads of two senior analysts. A new assistant session reports monthly revenue, and the senior analyst corrects it: refunds booked after the month closes do not count against that month. In a plain chat, the correction ends with the conversation, and the next person on the team gets the same wrong number. In Arroway the assistant records the correction as a proposal. The analyst approves it, with the end condition that it stops when finance changes the close calendar. The next day a colleague's session reports the figure correctly without being told. The rule about refunds is the only one that came out this week; the rest of what the two analysts know is still unsaid. Nothing here was measured on a real team; the team and the rule are illustrative.
Questions people ask about this
- Does Arroway interview my team to capture what they know?
- No. It records what is said and decided in the work and what you import, and a person approves it. A rule nobody has stated stays unrecorded until someone does.
- What about notes and documents we already have?
- They can be imported once, and the review afterwards is a person's first look at the whole set. Whatever turns out to be dead is left out rather than carried over.
- How do I try Arroway?
- Install it in the tool you already use — the install page has one route for Claude, ChatGPT, Codex, Cursor or any MCP tool — and sign in when the tool asks. From then on, sessions open by reading the project and close by recording what was decided.
Where this is verifiable
Product documentation on this site (How it works, Install), the import and export tools that ship with the connection, the answers on tracking AI facts without a spreadsheet and on how team decisions reach AI, and the proposal, sanction and expiry-condition rules in the sanctioned product spec. Everything described here about Arroway is behaviour the tools apply today, not roadmap. The team, the figures and the rule in the example are illustrative.
https://www.arroway.app/en/answers/what-the-team-knows-but-never-wrote-down-for-the-ai