How Knowledge Workers Use Shortwave's Semantic Search and Ghostwriter for Context-Aware Email

Knowledge workers spend hours reconstructing context from fragmented email threads before writing replies. Finding what a client agreed to six months ago, cross-referencing timeline changes across three colleagues, and responding in a personalized, professional tone usually turns a quick reply into a thirty-minute research task.

Shortwave tackles this workflow friction by connecting full-inbox vector semantic search with an adaptive drafting engine called Ghostwriter. Instead of forcing users to guess exact boolean keywords or rewrite robotic AI templates, Shortwave lets professionals locate past agreements using natural concepts and immediately generate replies that reflect their authentic voice and accurate history.

Here is the operational guide on how knowledge workers set up, search, and draft personalized communications in Shortwave.

1. The Core Problem: Why Keyword Search and Generic AI Fail

Traditional email tools separate search from composition. When managing high-stakes client relationships or internal projects, knowledge workers run into two persistent bottlenecks:

1. Keyword search brittleness: Standard search requires remembering exact words, dates, or senders. If an email referred to a contract discount as an "adjusted onboarding rate," searching for "pricing concession" returns zero results.
2. Sterile AI drafting: Generic AI assistants produce stiff corporate boilerplate ("I hope this email finds you well..."). They know nothing about previous conversations, specific constraints, or how the user actually communicates, requiring tedious line-by-line manual editing.

Shortwave eliminates this separation by treating email history as an interactive knowledge base. The AI Assistant searches conceptually across email threads and attachments, synthesizes the relevant facts, and passes that context directly to Ghostwriter to compose drafts.

2. Setting Up Shortwave for Deep Context Retrieval

To get reliable context retrieval, your account must index sufficient communication history and establish style parameters.

Importing Full Email History

When you link a Google Workspace or Gmail account, Shortwave automatically synchronizes your last 90 days of emails. For deeper historical retrieval:

• On paid tiers, navigate to Search in the app.
• Click Import full history on the top banner.
• Shortwave indexes your historical threads in the background. Once completed, your entire archive becomes searchable via vector semantics rather than just recent inbox items.

Note: Shortwave is designed exclusively for Gmail and Google Workspace accounts. It does not natively connect to Microsoft 365 or IMAP accounts without email forwarding.

Teaching Ghostwriter Your Voice and Boundaries

Ghostwriter learns continuously by analyzing your sent email patterns—studying your typical greetings, sign-offs, formatting preferences, and sentence cadence.

Knowledge workers can reinforce or constrain this learning through AI Memories in the AI Assistant settings:

• Writing preferences: "Remember I prefer casual, direct emails with bullet points" or "Remember to use Oxford commas."
• Business guardrails: "Remember our standard payment term is Net 30" or "Remember to check our Notion release notes before committing to deployment dates."
• Role context: "Remember I am Head of Client Success based in London; maintain British English spelling when drafting."

These memories persist across sessions, preventing the AI from reverting to generic corporate phrasing.

3. Retrieving Context with AI Semantic Search

Knowledge workers do not need to memorize complex search syntax to unearth buried discussions. Shortwave provides two complementary search pathways.

Conversational Search in the AI Assistant (Cmd-J)

Open the AI Assistant panel on desktop using Cmd-J (or the ✨ icon on mobile). Ask natural questions exactly as you would ask an executive chief of staff:

• "What did Sarah say about our Q3 onboarding milestone dates in late August?"
• "Find all customer feedback from last quarter complaining about export speed and list their specific comments."
• "What rate did we quote Marcus for the consulting retainer last year?"

The AI Assistant scans relevant threads and attachments (including PDFs, documents, and spreadsheets), extracts the answer, and cites the source threads so you can verify the underlying message with one click.

Concept Filtering with about:"[topic]"

In the primary search bar, Shortwave supports the about: operator alongside traditional attributes (from:, to:, after:, has:attachment):

• about:"pricing pushback" surfaces threads discussing budget friction, discount requests, or payment term objections, even if the word "pushback" never appears in the text.
• from:David about:"architecture roadmap" isolates technical discussions from a specific person based on concept rather than file names.

Managing Thread Context Limits

Depending on your subscription tier, Shortwave processes varying depths of context per search:

• Business plan ($30/seat/month): Evaluates up to 50 threads per AI search with 5 years of AI search history.
• Premier plan ($45/seat/month): Evaluates up to 100 threads with unlimited AI search history and double context tokens.
• Max plan ($120/seat/month): Evaluates up to 150 threads with triple context tokens for cross-project research.

4. Drafting Personalized Replies with Ghostwriter

Once context is identified, Ghostwriter offers three ways to compose replies that sound human and remain factually anchored.

Method 1: The One-Step "Search & Draft" Prompt

Knowledge workers frequently handle incoming emails that require cross-referencing earlier discussions. Instead of doing the research in one tab and typing in another, combine both actions into a single instruction inside the AI Assistant:

Example Prompt:
"Search our email history with Liam regarding the API rate limit discussion from May. Draft a reply to Sarah's latest email explaining those limits, keep it under three paragraphs, and ask if Tuesday at 2 PM works for a quick sync."

The AI Assistant searches your correspondence history, pulls the exact technical rate limits agreed with Liam, crafts a draft adhering to your voice rules, and loads it into your compose window.

Method 2: Instant AI Replies for Rapid Triage

When processing incoming mail during morning triage:

1. Open the thread.
2. Look at the inline Instant AI replies suggested beneath the message.
3. Ghostwriter evaluates the sender, the conversation history, and your style, generating 2–3 contextual reply variations.
4. Clicking a suggestion inserts a fully formed draft with one click. You can send it immediately or make minor edits.

Method 3: Predictive AI Autocomplete & In-line Polishing

For messages requiring personal manual input:

• AI Autocomplete: As you type your reply, Ghostwriter predicts upcoming sentences based on your past phrasing and current thread context. Press Tab to accept suggestions, cutting typing time significantly.
• Improve Draft: Highlight a rough draft and click the composer's "Improve draft" button, or instruct the AI Assistant: "Make this friendlier without losing the firm deadline" or "Tighten this explanation into three clear bullets."

5. Daily Workflows and Quota Management

Shortwave meters AI usage via a 24-hour rolling quota. Each prompt burns quota based on conversation length and the underlying model level:

Best Practices to Maximize Quota

1. Start fresh chats often: The AI evaluates preceding chat turns with every prompt. Keeping a single thread running for multiple unrelated tasks consumes quota rapidly. Reset the assistant panel between distinct tasks.
2. Select the appropriate tier: Use Standard intelligence for routine replies. Reserve Advanced and Expert for multi-document synthesis or strategic client proposals.
3. Be explicit upfront: Prompts that specify length, recipient relationship, and key constraints avoid multiple revision cycles.

6. Practical Trade-Offs and Considerations

While Shortwave significantly accelerates email management, knowledge workers should account for several practical constraints:

• Google Workspace dependency: Teams using Microsoft Outlook or on-premise Exchange cannot use Shortwave natively without auto-forwarding setups.
• Verification of extracted facts: While vector retrieval minimizes hallucinations compared to standalone chatbots, users should click Shortwave's thread citations to confirm critical financial figures or legal commitments before sending.
• Cost structure: Advanced multi-thread semantic capabilities require paid tiers starting at $30/user/month, positioning Shortwave as a high-leverage tool for email-heavy professionals rather than casual email senders.

By pairing semantic search that understands intent with Ghostwriter's personalized tone matching, knowledge workers eliminate the twin drains of archive hunting and draft rewriting, transforming email from an administrative burden into an organized, responsive communication hub.

References

• The Shortwave AI Assistant
• Search
• Start your 14 day free trial
• Make the most of your quota
• Import full history
• The Shortwave Method