Managing client expectations often means reconciling conflicting information between email threads and internal project documentation. A new tutorial published on DEV.to by Alichherawalla outlines a workflow using OGAD (Off Grid AI Desktop) to compare client emails with local Obsidian Markdown notes. The guide emphasizes the importance of reading both sources thoroughly before drawing conclusions, noting that email previews are often insufficient for establishing specific conditions.

Narrowing the Scope for Accuracy

The tutorial advises users to define a single, specific question rather than asking for a general comparison of entire projects. Whether the issue is a delivery date, scope, owner, budget wording, or acceptance criteria, a narrow focus allows the AI to verify details more effectively. For instance, if notes indicate a Friday review but a client asks about Monday, the goal is to determine if the email proposes a change or confirms an existing agreement, preserving the distinction between a proposal and a decision.

Connecting Local and Cloud Sources

To perform the comparison, users must install OGAD and utilize its Pro Integrations workflow. This involves connecting the relevant Google or Microsoft account for email retrieval and selecting the local Obsidian vault folder. The guide clarifies that while Obsidian tools read local notes without needing a cloud API, they cannot access files that are merely unavailable cloud placeholders. Email retrieval still requires an internet connection, but the analysis can be performed locally, keeping sensitive data on the user's machine.

Structuring the Evidence Table

The core of the workflow involves prompting OGAD to generate a side-by-side evidence table. The recommended prompt asks the model to compare specific fields, such as delivery dates, showing short quotes from each source, the source date, and the exact difference. Crucially, the output must classify each difference as compatible detail, proposed change, direct conflict, or insufficient evidence. The tutorial warns against assuming one source is authoritative unless the documents explicitly state that hierarchy.

Verifying Resolutions and Drafting Replies

After identifying a discrepancy, users are instructed to search for subsequent replies or updated notes that might resolve the conflict. A newer date alone does not prove authority; the model must find explicit acceptance or clarification. If no resolution exists, the AI should help draft a focused question to the client rather than assuming an answer. The final step involves drafting a reply that repeats verified facts and asks only the unresolved question, ensuring the original notes remain untouched unless manually updated.

Key Takeaways

  • Use OGAD beta 122 to connect Outlook or Gmail with local Obsidian vaults for privacy-focused analysis.
  • Always request a four-category classification (compatible, proposed, conflict, insufficient) to avoid false positives.
  • Never treat file modification time as the decision date; look for explicit approval in the text.
  • Keep source filenames and message links to maintain an audit trail for future client interactions.

The Bottom Line

Using local AI to cross-reference client communications with project notes is a powerful privacy-preserving workflow, but it requires strict discipline in defining scope and verifying evidence to avoid misinterpreting proposals as decisions.