Claude AI for Virtual Assistants: The Complete Guide (2026)

Claude AI for virtual assistants is the second most important AI tool in a VA’s stack after ChatGPT, and for a specific set of high-value VA tasks, it is the better tool. This guide covers what Claude does that ChatGPT does not, the VA-specific use cases where Claude produces consistently superior results, the prompt library organized by task type, the honest Claude vs ChatGPT comparison that tells you which tool to use for each specific task, and the five-day setup workflow that integrates Claude into your daily VA operations.

Most virtual assistants who try Claude do so because they have heard it produces better long-form writing than ChatGPT. That is true, and it understates the case. Claude’s advantages for VA operations go beyond writing quality. The 200K token context window processes entire client document libraries in a single prompt. The instruction-following precision reduces the revision cycles that consume time in high-volume content workflows. The reasoning depth handles nuanced client communication tasks (sensitive email responses, complex proposal framing, multi-stakeholder briefings) where ChatGPT’s tendency to default to predictable outputs creates more work than it saves.

The question most VAs ask is not “should I use Claude” but “I already use ChatGPT, do I need Claude too?” The answer is not universal. It depends on the service types in the VA’s scope and the specific task profile of their daily work. This guide provides the data and the framework to answer it correctly for your situation, and the prompt library to use Claude effectively from day one if the answer is yes.

What this guide covers:

• What Claude does and what it does not, the honest capability profile for VAs
• The VA-specific use cases where Claude outperforms ChatGPT
• Plans and pricing, what each tier gives a solo VA
• The Claude prompt library, 15+ templates organized by VA task type
• The Claude vs ChatGPT comparison, which tool for which task
• How to set up Claude for VA operations
• The five-day integration workflow
• Common mistakes and how to avoid them

👉 AI Tools for Virtual Assistants: The Complete Practical Guide — the full reference for every AI tool category in VA work.

1. What Claude AI Does and Does Not Do

Understanding how to use Claude as a virtual assistant correctly requires starting with an honest capability profile, what Claude does better than alternatives, what it does not do, and where the marketing positioning overstates the case.

What Claude does well:

Long-form writing and document generation. Claude produces longer, more coherent, and more tonally consistent long-form content than ChatGPT at equivalent prompting effort. Proposals, SOPs, client briefings, newsletter drafts, and multi-section reports all benefit from Claude’s stronger contextual consistency across extended outputs.

Instruction following at high precision. Claude follows multi-part, complex instructions more reliably than ChatGPT. A prompt with eight specific requirements (specific format, tone, length, exclusions, audience, structure, example, and CTA) produces output that addresses all eight more consistently from Claude than from ChatGPT, which tends to prioritize the most prominent instructions and drop lower-priority ones.

Long-context document processing. The 200K token context window, equivalent to approximately 150,000 words, allows Claude to process an entire client document library, a full project history, or a book-length research document in a single prompt. ChatGPT’s standard 128K context window handles most VA use cases but reaches its limit with very large client archives or multi-document analysis tasks.

Nuanced communication tasks. Claude handles sensitive, high-stakes, or tonally complex communication tasks with greater nuance than ChatGPT. Emails responding to difficult client situations, proposals that require careful framing, and briefings for multiple stakeholders with different priorities all benefit from Claude’s more careful, context-aware approach to language.

What Claude does not do:

Image generation. Claude cannot create images natively. For AI image generation, ChatGPT (DALL-E integration) or dedicated tools like Midjourney and Ideogram are required.

Real-time and live data retrieval. Claude includes web search on all plans, including Free, and uses it autonomously when a query benefits from current information. However, for tasks requiring live data feeds, financial tickers, or minute-by-minute news monitoring, ChatGPT’s more deeply integrated browsing or dedicated tools like Perplexity remain more reliable. Claude’s web search is optimized for research queries, not real-time data streams.

Deep third-party workflow integration. Claude supports MCP (Model Context Protocol) connectors, basic connectors (Notion, Slack, Google Workspace) on Free, full MCP server support on paid plans. For complex multi-step workflow automation connecting Claude to external tools, Make or Zapier via the Anthropic API provides the most flexible integration layer.

The practical implication for VAs: Claude is the correct primary AI tool for writing-heavy, documentation-heavy, and communication-heavy VA service types. ChatGPT is the correct primary AI tool for multimodal tasks (image generation), plugin-based integrations, and workflows requiring live data streams. Both tools now include web search and memory, the differentiator is output quality for depth-oriented tasks, not feature availability. Most experienced VAs use both, each for the task types where it produces the best output.

2. Why Claude AI Matters for Virtual Assistants

Claude AI for virtual assistants is not a ChatGPT replacement, it is a specialization. The data supports this framing. Anthropic reports that 13% of all Claude tasks fall into the Office and Administrative Support category, a figure that has grown 3% in four months, the fastest-growing task category on the platform. The tasks driving that growth are exactly the high-frequency VA work types: document creation, communication drafting, process documentation, and workflow automation support.

Claude completes college-level tasks 12x faster than the human average. For VAs whose service scope includes research, analysis, and structured documentation, tasks that map directly to college-level cognitive work, this speed multiplier translates to concrete hourly capacity.

The three reasons Claude matters specifically for VA operations:

1. The writing quality gap closes the revision loop. A VA who uses ChatGPT for long-form content drafting typically spends 30-50% of their AI-assisted writing time on revision, fixing tone inconsistencies, structural issues, and sections where the AI lost context. Claude’s stronger contextual consistency and more precise instruction following reduces this revision overhead to 10-20% on equivalent tasks. For a VA producing 5-10 documents per week, this is 1-3 hours of recovered time.

2. The context window handles multi-client complexity. A VA managing five clients accumulates significant per-client context: brand guides, SOP libraries, communication histories, project documentation. Feeding all relevant context into every AI prompt is essential for high-quality, client-specific output, but it quickly exceeds ChatGPT’s 128K context limit for clients with extensive documentation. Claude’s 200K context window removes this constraint for almost all practical VA use cases.

3. Projects replace the manual context-setting overhead. Claude Projects allow the VA to store client-specific instructions, documents, and style guides as persistent project context, accessible in every conversation within that project without re-pasting. The operational benefit is equivalent to ChatGPT’s Custom GPTs but with Claude’s superior writing quality applied on top of the stored context. A Claude Project per client eliminates the manual context-setting step that adds 5-10 minutes to every AI-assisted task in a contextless workflow.

3. Claude Plans and Pricing for VAs

The solo VA recommendation: Claude Pro at $20/month.

Claude Pro gives a solo VA everything needed for full daily integration: sufficient usage capacity for 4-6 hours of active AI-assisted work per day, access to Claude Opus 4.6 (the most capable model for complex reasoning tasks), Projects for per-client context storage, Google Workspace integration (read and edit Google Docs, Gmail, and Drive directly within the chat interface), Cowork for autonomous multi-step background tasks, and full MCP server support for external tool connections.

New in 2026 — Cowork. Cowork is Anthropic’s agentic feature launched January 12, 2026, that allows Claude to autonomously complete multi-step tasks in the background (file organization, multi-document analysis, project coordination) while the VA continues working on other tasks. Available on Pro since January 16, 2026. For VAs who run document-heavy workflows, Cowork eliminates the need to wait for Claude to complete long sequential tasks before moving on.

Claude Pro vs ChatGPT Plus — cost comparison:

Both cost $20/month. The decision between them as the primary AI tool is not a cost decision, it is a capability-fit decision based on the VA’s primary task profile. The comparison in Section 6 provides the task-by-task breakdown.

The Free plan reality for VAs:

Claude Free provides access to Claude Sonnet 4.6 with usage limits that reset every 5-8 hours. As of 2026, the Free plan includes basic web search, memory, and Projects, features that were previously Pro-only. This makes the Free plan genuinely more capable than in earlier versions of Claude. However, the usage limits remain the binding constraint: during peak hours, a VA can exhaust the Free plan’s message allowance within 1-2 hours of active work. For consistent, professional daily use, Claude Pro at $20/month eliminates the usage ceiling and adds Google Workspace integration, Cowork, and Claude Code that the Free plan does not include.

The combined stack cost:

A VA using both Claude Pro and ChatGPT Plus pays $40/month, $20 for each. For a VA billing $35-50/hour, this represents less than one hour of billable time per month. The combined stack is the recommended configuration for VAs whose service scope includes both writing-heavy and multimodal tasks.

4. The Top 8 Claude Use Cases for Virtual Assistants

The eight use cases below represent the VA task types where Claude AI produces the most consistent, highest-quality output, ranked by the combination of frequency in VA operations and quality advantage over ChatGPT.

5. The Claude Prompt Library for VAs

The prompt library below covers the eight use cases from Section 4. Each prompt follows the Role/Task/Context/Format structure and is written for Claude specifically, leveraging Claude’s instruction-following precision with multi-part prompts that would produce less consistent output in ChatGPT.

You are a process documentation specialist writing an SOP for [CLIENT NAME], a [BUSINESS TYPE].

Write a complete SOP for: [PROCESS NAME].

Requirements — follow all of these exactly:
1. Length: 600-900 words
2. Format: use the following section headers exactly as written: PURPOSE / TOOLS REQUIRED / PREREQUISITES / STEP-BY-STEP INSTRUCTIONS / QUALITY CHECK / TROUBLESHOOTING / RELATED PROCESSES
3. Tone: clear, direct, instructional — write for someone doing this process for the first time
4. Steps: number all steps, include specific tool names, menu paths, and field names where relevant
5. Quality Check section: include 3 specific verification criteria, not generic statements
6. Troubleshooting: include 3 specific issue + solution pairs

Process description: [DESCRIBE THE PROCESS IN DETAIL]

Tools used in this process: [LIST TOOLS]

Do not add sections beyond those listed.
Do not use bullet points inside the Steps section — numbered steps only.

You are a professional services writer creating a client handbook for [CLIENT NAME], a [BUSINESS TYPE] serving [AUDIENCE].

Write the [SECTION NAME] section of their client handbook.

Tone: [VOICE — e.g., warm and professional, authoritative, approachable].

Requirements:
1. Length: 400-600 words
2. Open with a clear statement of what this section covers and why it matters to the client
3. Use H3 subheadings for each sub-topic
4. End with a clear next step or action the client should take
5. Write in second person ("you" / "your") throughout — addressing the client directly
6. No jargon — use plain language

Section topic and key points to cover: [DESCRIBE THE SECTION AND KEY POINTS]

You are a business development writer creating a service proposal for [YOUR NAME / BUSINESS NAME], a virtual assistant specializing in [YOUR SERVICES].

Write a complete service proposal for: [PROSPECTIVE CLIENT NAME] [THEIR BUSINESS TYPE AND BRIEF DESCRIPTION]

Their stated need / problem: [DESCRIBE WHAT THEY NEED]

Services you are proposing: [LIST YOUR PROPOSED SERVICES]

Investment: [YOUR PRICING / RETAINER STRUCTURE]

Requirements:
1. Open with a brief problem statement that demonstrates you understand their specific situation — not a generic intro about your services
2. Proposed solution section: explain specifically how your services address their stated need
3. What's included: use a clean list format
4. Investment section: frame the price in terms of time saved or value delivered, not just the number
5. Next steps: one clear, low-friction CTA
6. Tone: confident, peer-level — not subservient or over-eager
7. Length: 400-600 words total

Do not use: "I'm excited to", "I would love to", "please don't hesitate", or any similar filler phrases.

You are a project documentation specialist writing a Scope of Work document for [YOUR BUSINESS NAME].

Client: [CLIENT NAME]
Project: [PROJECT NAME]
Project dates: [START DATE] to [END DATE]

Write a complete Scope of Work covering:
1. Project Overview (2-3 sentences)
2. Objectives (3-5 bullet points)
3. In Scope (specific deliverables — be precise, not general)
4. Out of Scope (explicit exclusions — at least 3 items)
5. Deliverables and Timeline (table format: Deliverable | Due Date | Format | Review Rounds)
6. Assumptions (conditions under which this SOW is valid)
7. Change Request Process (1 paragraph)

Format each section with its header in bold. Use a table for section 5.
Total length: 500-800 words.
Tone: professional, unambiguous, specific.
Avoid vague language in sections 3 and 4 — every item must be concrete and measurable.

You are a professional communication specialist drafting an email on behalf of [YOUR NAME], a virtual assistant.

Situation: [DESCRIBE THE SITUATION FULLY — include what happened, the client's likely emotional state, and the relationship context]

Email objective: [WHAT THIS EMAIL MUST ACHIEVE]

Requirements:
1. Tone: [TONE — e.g., empathetic but professional, direct but warm, firm but respectful]
2. Acknowledge the situation directly in the first paragraph — do not bury the lead
3. Do not over-apologize or use language that implies more fault than is accurate
4. Propose a specific resolution or next step — not an open-ended offer to "discuss further"
5. Close with one clear action item for the client
6. Length: 150-250 words maximum
7. Subject line: write 3 options

Do not use: "I apologize for any inconvenience", "please feel free to", "at your earliest convenience". These are weak — replace with direct, specific alternatives.

You are a professional business writer drafting an email for [YOUR NAME], a virtual assistant.

Context: [CLIENT NAME] has been requesting work outside the original scope agreement. [DESCRIBE SPECIFIC EXAMPLES OF SCOPE CREEP]. The email needs to reset boundaries without damaging the relationship or signaling willingness to continue absorbing extra work.

Requirements:
1. Open by acknowledging the project positively — 1 sentence only
2. Transition to scope: reference the original agreement specifically (do not say "as we discussed" — be concrete about what was agreed)
3. Name the specific out-of-scope items that have been added — list them
4. Offer two options: absorbing the extra work at an additional fee, or deferring it to a future scope
5. Close with a question that invites the client to choose one option
6. Tone: confident, collegial, not apologetic
7. Length: 200-300 words
8. Subject line: neutral, non-confrontational

This email must not read as a complaint or ultimatum — it is a professional reframing of the working arrangement.

You are a research analyst preparing a briefing for [CLIENT NAME], a [ROLE — e.g., CEO, founder, director] at [COMPANY TYPE].

[PASTE SOURCE DOCUMENTS OR RESEARCH NOTES]

Synthesize the above into a structured executive briefing.

Requirements:
1. Format:
   SITUATION SUMMARY (3-4 sentences)
   KEY FINDINGS (5-7 bullet points — most important insights only, no filler)
   IMPLICATIONS (3 bullet points — what this means for [CLIENT NAME] specifically)
   RECOMMENDED ACTIONS (2-3 concrete next steps, prioritized)
   OPEN QUESTIONS (2-3 questions the research does not answer but [CLIENT NAME] should consider)
2. Length: 400-600 words total
3. Tone: direct, analytical, peer-level — write for a busy decision-maker who will skim before reading
4. Avoid: summarizing what the sources say. Synthesize what they mean.
5. The Implications and Recommended Actions sections must be specific to [CLIENT NAME]'s situation — not generic observations.

You are a strategic analyst writing a competitor analysis summary for [CLIENT NAME], a [BUSINESS TYPE] targeting [AUDIENCE].

[PASTE COMPETITOR RESEARCH NOTES OR DESCRIBE WHAT YOU KNOW ABOUT EACH COMPETITOR]

Write a structured competitor analysis covering [NUMBER] competitors.

Format for each competitor:
COMPETITOR NAME
Positioning: 1 sentence on how they position themselves
Strengths: 3 bullet points
Weaknesses: 3 bullet points
Differentiator vs [CLIENT NAME]: 1-2 sentences

After all competitors, add:
STRATEGIC IMPLICATIONS
3 bullet points on what [CLIENT NAME] should do differently or emphasize given this competitive landscape.

OPPORTUNITY GAPS
2-3 specific positioning or service opportunities the competitors are not addressing.

Total length: 600-900 words.
Be specific — no generic statements like "strong brand presence." Every point must be concrete and actionable.

You are a content writer for [CLIENT NAME], a [BUSINESS TYPE] with a [VOICE] brand voice targeting [AUDIENCE].

Write a complete email newsletter for the week of [DATE].

Topic: [TOPIC]
Key insight or takeaway: [MAIN POINT]

Requirements:
1. Subject line: write 3 options (one curiosity-driven, one value-driven, one direct). Mark the recommended one.
2. Preview text: 1 option (40-60 chars)
3. Opening: a 2-3 sentence hook that opens with the insight or a story — not a recap of the topic
4. Body: 3-4 sections with H3 subheadings, each 80-120 words
5. Each section: one concrete idea, one practical application
6. CTA: one specific action — not "reply and let me know your thoughts" unless that is explicitly the goal
7. Sign-off: 1-2 sentences in [CLIENT NAME]'s voice — personal, not boilerplate
8. Total length: 450-650 words
9. Tone: [VOICE DESCRIPTION]

Do not use: "I hope this finds you well", "excited to share", "don't miss out", or similar filler phrases.

6. Claude vs ChatGPT for Virtual Assistants — Task-by-Task Comparison

The claude vs chatgpt for virtual assistants comparison is most useful when structured by task type rather than by abstract feature categories. The table below provides a direct recommendation for each of the 12 highest- frequency VA task types.

The pattern in the table: Claude wins on tasks requiring depth, length, nuance, and tonal precision. ChatGPT wins on tasks requiring speed, multimodal output, or real-time data. For the middle tier (spreadsheets, brainstorming, routine communications) both produce equivalent results and the choice is a workflow preference.

The practical routing rule for VAs:

Use Claude when the task is: → Over 500 words of output → Tonally sensitive or high-stakes → Based on large amounts of input context → Requires consistent voice across sections → Involves synthesizing multiple existing documents

Use ChatGPT when the task involves: → Image generation or analysis → Live or real-time data (financial, news feeds) → Quick structured outputs under 300 words → Plugin or third-party tool integration

👉 ChatGPT for Virtual Assistants: Complete Guide — the full ChatGPT use case breakdown, prompt library, and workflow integration for VA operations.

7. The Claude + ChatGPT Combined Stack

The most effective claude ai virtual assistant workflow for an experienced VA with a diverse service scope is not a choice between Claude and ChatGPT, it is a combined stack where each tool handles the task types it is best suited for.

The combined stack configuration:

Claude handles:

• All long-form writing (SOPs, proposals, handbooks, newsletters, long articles)
• Complex and sensitive client communications
• Multi-document research synthesis and briefings
• Per-client context storage via Projects (one Project per client, stores brand voice guide, SOP library summary, communication preferences, and ongoing project context)

ChatGPT handles:

• Social media captions and short-form content
• Image generation and visual content briefs
• Quick task lists, planning, and brainstorming
• Research requiring live data (financial, real-time news, current prices)
• Workflow automation via ChatGPT API in Make and Zapier

The switching rule: The routing table in Section 6 provides the task-level decision. In practice, the daily workflow looks like this: morning planning and task breakdown → ChatGPT. Long-form writing sessions → Claude. Quick email drafts → ChatGPT. Complex communication or proposals → Claude. Research with live data → ChatGPT. Research synthesis from existing documents → Claude.

Setup cost of the combined stack: Claude Pro ($20/mo) + ChatGPT Plus ($20/mo) = $40/month. At a $40/hour VA rate, this represents one billable hour per month. The combined stack typically recovers 5-10 hours of operational time per week, a return of 20-40x the monthly tool cost.

8. How to Set Up Claude for VA Operations

The Claude setup for VA operations differs from a casual Claude user’s setup in one critical respect: Projects. As of 2026, Projects are available on all plans including Free, but the full operational value of Projects for multi-client VA work requires Pro, because Free plan usage limits constrain the volume of conversations a VA can run within Projects daily. Without a configured Project, every Claude conversation starts with zero context, the VA manually re-pastes client information, brand guides, and instructions at the start of every session. With Projects, the context is persistent, loaded automatically in every conversation within the project.

Step 1 — Create one Project per client.

In Claude Pro, open Projects and create a new project for each active client. Name format: [CLIENT NAME] — [PRIMARY SERVICE]. Example: “Acme Corp — Content + Admin”

Step 2 — Configure the Project instructions.

Each Project has a “Project Instructions” field, the persistent system prompt for all conversations in that project. The instructions should contain:

CLIENT: [CLIENT NAME]
BUSINESS TYPE: [DESCRIPTION]
MY ROLE: [VA SERVICES PROVIDED]

BRAND VOICE: [3-5 voice descriptors with IS / IS NOT pairs]

VOCABULARY TO USE: [list]
VOCABULARY TO AVOID: [list]

COMMUNICATION PREFERENCES: [How the client prefers to communicate, any sensitivities, relationship context]

CURRENT PROJECTS: [Active projects and their status]

IMPORTANT CONTEXT: [Anything Claude should always know when working on tasks for this client]

Step 3 — Upload key client documents.

Each Project accepts document uploads. Upload for each client:

• Brand voice guide or style guide
• Key SOP reference documents
• Service agreement or scope summary
• Any ongoing project briefs

Claude references these documents in every conversation in the project, without the VA needing to paste them.

Step 4 — Create a VA Operations Project.

In addition to per-client projects, create a “VA Operations” project with instructions containing:

• Your service packages and pricing
• Your standard response templates
• Your working hours and policies
• Your tools and systems overview

Use this project for tasks that are about your own business, proposals, your own SOPs, business communications.

Step 5 — Build the prompt library.

Save the prompts from Section 5 as text files or Notion pages, one prompt per task type, with the variable fields clearly marked. Before each Claude session, copy the relevant prompt, fill in the variables, and paste into the project conversation. This eliminates blank-page time for every recurring task type.

9. The Five-Day Integration Workflow

The five-day workflow below is designed for a VA who is adding Claude to an existing workflow that already includes ChatGPT. It introduces Claude gradually, one task type per day, to build familiarity with Claude’s output patterns before relying on it for client-facing work.

Day 1 — Long-Form Documentation. Use Claude for one SOP or process document. Use Prompt 1.1 from Section 5. Compare the output to your typical ChatGPT output for the same task type. Note: where does the output require less editing? Where does it require more?

Day 2 — Complex Email Drafting. Use Claude for one sensitive or high-stakes client communication. Use Prompt 3.1 or 3.2. Compare to your ChatGPT output pattern for similar emails. The goal: identify whether Claude’s tonal approach matches your instinct for this type of communication.

Day 3 — Client Proposal or SOW. Use Claude for one proposal or scope document. Use Prompt 2.1 or 2.2. Pay attention to the opening and the framing of the investment section, these are the sections where Claude’s persuasive structure advantage is most visible.

Day 4 — Research Synthesis. Use Claude for one multi-document synthesis task. Paste 2-4 source documents and use Prompt 4.1. Test the context window, how large a document set can you include before the output quality changes?

Day 5 — Project Setup. Set up Projects for your top 3 clients following the instructions in Section 8. Run one task per client in the Project context and compare output quality to the same task without Project instructions.

After five days: the tasks where Claude produced output requiring less editing than ChatGPT are your Claude tasks going forward. The tasks where ChatGPT was faster or produced equivalent quality stay with ChatGPT.

10. How to Use Claude AI as a Virtual Assistant — Step by Step

11. Common Mistakes

12. Conclusion

Claude AI for virtual assistants is not a replacement for ChatGPT, it is a complement that closes the quality gap in the specific task types where ChatGPT’s limitations produce the most operational friction: long-form documentation, nuanced communication, and multi-document research synthesis. For VAs whose service scope is heavy in these task types, Claude Pro at $20/month produces a return that is visible within the first week of use.

The five-day integration workflow in Section 9 provides the empirical basis for a routing decision grounded in your actual work, not in abstract feature comparisons. The task-level routing table in Section 6 provides the reference guide once the routing decisions are established. The Projects setup in Section 8 provides the persistent context layer that makes Claude’s output quality consistent and setup overhead minimal.

The VAs who extract the most value from Claude are not those who use it for everything, they are those who use it for the right things, configured correctly, with prompts that leverage Claude’s instruction-following precision rather than treating it as a faster search engine.

Frequently Asked Questions About Claude for Virtual Assistants

Glossary: Key Claude Terms for Virtual Assistants

Claude Projects A feature available on all Claude plans that creates persistent context spaces, each Project stores system instructions, uploaded documents, and conversation history accessible in every new conversation within the project. The primary operational feature for VAs, replacing manual context-pasting at the start of every client-specific session. Full functionality (higher storage limits, priority processing, Google Workspace integration) is available on Pro.

Context Window The maximum amount of text Claude can process in a single conversation, measured in tokens (approximately 0.75 words per token). Claude’s 200K token context window processes approximately 150,000 words, equivalent to a full-length non-fiction book, in a single conversation. Critical for multi-document analysis and large client archive processing.

Claude Sonnet 4.6 The mid-tier Claude model available on the Pro plan, optimized for speed and cost efficiency. Recommended for routine VA tasks: email drafting, short content, task structuring, and any task where the primary requirement is fast, adequate output rather than maximum quality.

Claude Opus 4.6 The flagship Claude model, optimized for complex reasoning, nuanced writing, and deep analysis. Available on Claude Pro. Recommended for high-stakes VA tasks: client proposals, multi-document synthesis, sensitive communications, and long-form documentation where output quality directly affects client deliverable quality.

Token The unit of text measurement in AI language models. One token is approximately 0.75 words in English. A 1,000-word document is approximately 1,333 tokens. Claude’s 200K context window can process approximately 150,000 words of combined input and output per conversation.

System Prompt Instructions given to Claude that define its role, tone, and behavior for a conversation or project. In Claude Projects, the Project Instructions field functions as a persistent system prompt, applied automatically to every conversation in the project without the VA needing to include it in each individual message.

Instruction Following Claude’s ability to adhere to multi-part, specific instructions in a single prompt. Claude’s higher instruction-following precision compared to ChatGPT means that prompts with 6-10 specific requirements produce output that addresses all requirements more consistently, reducing the revision cycles needed to achieve the target output.

Hallucination An AI output that contains factually incorrect information presented with apparent confidence. Claude hallucinates less frequently than earlier AI models on most task types, but the risk is non-zero, particularly for specific factual claims, statistics, dates, and proper nouns. All Claude outputs for client-facing use require human review before delivery.

Model Selection The choice between Claude Sonnet 4.6 and Claude Opus 4.6 within a Pro account. Available via the model selector at the top of each Claude conversation. The correct selection is task-dependent: Sonnet for speed and routine tasks, Opus for depth and high-stakes outputs.

Memory A feature available on all Claude plans that automatically synthesizes past conversations into a persistent context summary, updated every 24 hours, that Claude references in every new session. Memory tracks the VA’s general working preferences, recurring contexts, and background details without requiring manual re-entry. Controllable from Settings > Capabilities: the VA can review what Claude has remembered, edit individual memory entries, or pause memory entirely.

Cowork Anthropic’s agentic feature launched January 2026, available on Claude Pro and above. Cowork allows Claude to autonomously complete multi-step tasks in the background (file organization, multi-document analysis, project coordination) while the VA works on other tasks. Particularly useful for document-heavy VA workflows where waiting for long sequential tasks creates operational bottlenecks.

Routing Rule A personal or team-level decision framework that determines which AI tool (Claude or ChatGPT) is used for each task type. The routing rule in this guide is based on the task-by-task comparison in Section 6. A documented routing rule produces more consistent output quality than ad-hoc tool selection and reduces the decision overhead of choosing a tool for every new task.