Assistant Tips & Troubleshooting

Refreshing Conversations

If the Assistant takes too much time to respond or start reasoning, refresh your conversation:

‍

When to refresh:

The Assistant is taking unusually long to begin reasoning or generate a response

The page appears unresponsive or stuck

Reasoning has stopped progressing through phases

‍

How to refresh:

Click the refresh button in your browser

The conversation will reload with all prior context preserved

You can continue from where you left off without losing your work

‍

Avoid Long Chats

Start a new chat when your current conversation has accumulated very high context. Extended conversations with extensive content can impact response time and performance.

‍

When to start a new chat:

The conversation has 20+ exchanges with lengthy responses

You've attached thousands of pages across multiple queries

The Assistant has performed numerous searches in a single conversation

You've asked many complex follow-up questions building on previous answers

Performance seems slower than usual

‍

When to continue in the same chat:

Following up on the same question or matter

Iterating on a draft or analysis where the Assistant needs prior context

Working on related aspects of the same issue

The conversation is still relatively short and focused

‍

Handling Context

Provide context strategically to optimize the Assistant's performance.

‍

Prioritize the most relevant sources:

Attach or select only the documents directly relevant to your query. More documents doesn't mean better results—focused, relevant sources produce higher quality analysis.

‍

❌ Too many sources:

Attaching 20 documents "just in case" when only 3 are directly relevant to your question

✅ Relevant sources only:

Attaching the specific contract you need reviewed, your template for comparison, and the relevant correspondence about disputed terms

‍

File size vs. content:

The number of pages and amount of text matters more than file size in MB. A 5MB PDF with 50 pages of dense text requires more processing than a 10MB presentation with 30 slides and many images.

Focus on text-rich, relevant content rather than file size alone.

‍

Don't reattach documents:

The Assistant remembers all documents attached within a conversation. If you've already uploaded a contract in your first message, don't attach it again in your follow-up questions—simply reference it.

‍

Example:

First message: [Attach contract] "Review this MSA for data processing terms"

Follow-up: "Now check the same contract for liability caps"

[No need to reattach — the Assistant already has it]

‍

Pictures and OCR considerations:

Picture-heavy documents (presentations, marketing materials, infographics) contain less processable text and may not yield as much analytical value

Poor OCR quality from badly scanned documents can affect accuracy — if a scanned document isn't processing well, copy and paste the text directly instead

Charts, tables, and diagrams in images may not be fully captured — describe key data points in your prompt if they're critical to your question

‍

Common Issues & Solutions

‍

Legal Research: Case Law Retrieval

For legal research, especially case law, the ideal results are sometimes not publicly available on the web. Full-text judgments, older decisions, or cases from certain jurisdictions may not be accessible through open sources, which affects the Assistant's retrieval capabilities.

‍

Solution:

Provide relevant case law as attachments. If you have access to the full judgment through Westlaw, LexisNexis, or your court filing system, attach it directly. The Assistant can then analyze the case in detail, extract key holdings, and apply the reasoning to your specific question.

‍

Example:

[Attach judgment PDF from Westlaw]

Analyze the court's reasoning in this judgment on the doctrine of legitimate

expectations. How does it apply to our client's situation where the regulator

changed guidance after our client made substantial investments?

‍

Enterprise Search: Library Quality

The performance of enterprise search is directly affected by the quality and organization of sources in your Library. Well-maintained, high-quality Library content produces better search results and more relevant answers.

‍

Best practices for Library sources:

‍

Avoid conflicting or outdated sources:

Remove superseded templates or policies when new versions are added

Archive outdated legal memos that no longer reflect current law

Don't keep multiple conflicting versions of the same document unless version control is intentional

‍

Add repeatable, high-quality sources:

Upload your firm's current standard templates and playbooks

Include well-organized precedent files with clear naming conventions

Add approved internal memos and legal analysis

Ensure documents have clear metadata (client, matter, date, author)

‍

Ask descriptive questions with semantics and keywords:

The Assistant searches your Library using both semantic understanding (meaning and context) and keyword matching. Effective Library search queries combine both.

‍

❌ Vague query:

Find the IP clause

✅ Descriptive query with semantics and keywords:

Find our standard intellectual property assignment clause for software development agreements, specifically the provision on work-made-for-hire and assignment of moral rights

‍

The descriptive query includes:

Semantics:"intellectual property assignment," "software development agreements," the concept of work-made-for-hire

Keywords: "IP," "assignment," "moral rights," "work-made-for-hire"

Context: The type of agreement and specific provision needed

‍

This combination helps the Assistant retrieve the most relevant Library sources even if your firm uses different terminology in different documents.