Best practices
Practical guidelines for effective work with the AI Advisor — depth selection, formulating questions, attaching files, handling citations, and recognizing budget exhaustion signals.
Selecting analysis depth for the query
Accurate selection ensures the advisor returns adequate results — quick for searches, comprehensive for analysis.
| Depth | When to select | When to avoid |
|---|---|---|
| Surface | The answer is a single fact or a short list, and the first reliable match is sufficient. | The query requires multi-source synthesis — Surface will search too shallowly. |
| Balanced | Default setting. The query requires several searches, and a full answer is expected in a reasonable time. | A Surface answer would suffice — Balanced generates unnecessary searches. |
| Deep | A multi-source query: comparison, briefing, identifying discrepancies, or „highlight anything relevant”. | The query is narrow — Deep will search too broadly and may return more than necessary. |
Formulating questions using corpus terms
The advisor is designed with specific skills — analyzing and searching Sejm proceedings, stakeholders, clubs, votes, and files. Questions formulated precisely using these terms generate better and more accurate responses.
- Less effective: „What is happening with energy policy?" — the scope remains undefined.
- More effective: „Summarize recent floor activity on the renewable energy bill, including the last vote and committee actions from the past week."
Not every question needs to be elaborate. Including at least one specific piece of information — a bill name, stakeholder, club, date range, committee, or file — is sufficient.
Attaching a file instead of describing it
When the answer resides in a document already available in the repository, attaching it is faster and more accurate than asking the advisor to search for it.
Entering @ in the input field opens the file selector. Searching is performed by name or — after switching the mode — by document content. Multiple files can be attached simultaneously.
The advisor treats the attached file as a primary reference and skips the step of searching across all files.
When not to attach a file: when the goal is to find the correct document. A query like „Which of our briefings discusses changes in public procurement law?" should be submitted without an attachment to trigger the advisor's hybrid search.
Reading citations, not just response content
Every advisor data source is exposed as a verifiable link. This enables verifying the generated data:
- Vote counts and results per club.
- Stakeholder identities.
- The content of a cited post.
- The source document at the cited fragment.
Generated responses generally align with facts, but if a generated response appears incorrect, regenerate it by selecting the arrow icon below the response. The second pass often returns correctly grounded citations.
Utilizing conversation continuity
The advisor retains the full conversation as context for follow-up questions. Three effective patterns:
- In-progress refinement — „Now break down that same comparison by voivodeship instead of club." — the advisor reuses the previous result.
- Deepening a specific finding — „Tell me more about the discrepancy in amendment 14."
- Transitioning to action — „Configure an alert covering the bills you just briefed." — the advisor extracts identifiers from the previous response.
Avoid jumping between unrelated threads in a single conversation. For different topics, open new sessions — the chat history makes this efficient.
Recognizing the budget exhaustion signal
When a query exceeds the single-request budget — usually a Deep analysis with a very broad scope — the response concludes with a note regarding the incomplete portion:
„...Out of the eleven bills you asked about, I discussed nine in detail; the query budget was insufficient for bills 10 and 11. Repeat the query narrowed down to those two specifically."
This is a valid result. The recommended action is to accept the partial response and submit a narrower follow-up query. If budget exhaustion occurs regularly for a given question type, the question is likely too broad — two queries with the Balanced setting yield a better combined result than one at the Deep level.
Responding to elicitation forms
When a tool requires a resolution (e.g., „Which Sejm term?", „Which of the two stakeholders named Kowalski?"), the advisor displays a form. Fill it out directly — entering the answer in the chat field leaves the form unresolved and initiates a new turn instead of continuing the current one.
When the advisor is the wrong tool
The advisor deflects several classes of questions:
- General conversation — questions about opinions and evaluations outside the legislative corpus.
- Out-of-scope analysis — international relations, market forecasts, election polls. The advisor will note the limitation and answer what is available from the corpus.
- Content creation — the advisor summarizes, briefs, and compares, but does not write press releases or legal documents.
- Bulk operations — the advisor executes queries, not actions. Existing actions (creating alerts, uploading files) are explicit and confirmed.
In these cases, the advisor explains the limitation and points to the appropriate part of the platform when possible.