Claude, Claude Code & Claude Co-work

Test your understanding of AI fundamentals. You get up to 3 attempts per question — hints will guide you if you get it wrong.

Test your understanding of Claude Code basics. Remember: up to 3 attempts per question with hints.

One of the most immediate ways Claude Code saves time is drafting professional emails and documents. Rather than starting from a blank page, you describe what you need, Claude produces a polished draft, and you review and send. Over a week, this can save hours of writing time.

Drafting Property Emails

When asking Claude Code to draft an email, always provide:

• Recipient and context — who are they, what is the relationship?
• Purpose — what do you need to communicate?
• Tone — formal/informal, urgent/routine?
• Key facts — dates, figures, property addresses, reference numbers
• Any attachments to mention — documents you will attach manually

Claude will produce a complete, properly formatted email. Before sending, always check:

Sending Emails via Google Workspace MCP

Claude Code can send emails directly through Gmail using the Google Workspace MCP server. This means you can draft and send without leaving the terminal.

Claude will ask for your confirmation before actually sending. You will see a permission request like:

Claude wants to use: Google Workspace > Send Gmail Message
To: sarah.chen@knightfrank.co.uk
Subject: Rent Review - Unit 4, Riverside Business Park, S2 4SL

Allow? (y/n)

Only type y when you have verified everything is correct.

Creating Google Docs and Sheets

Beyond email, Claude Code can create documents directly in Google Drive:

Managing Slack Messages

Claude Code can also send messages to Slack channels, which is useful for internal team updates:

Property work relies on accurate, current data. Claude Code can help you gather, structure, and analyse information from multiple sources — but you must understand where the data comes from and how to verify it.

Web Research with Claude Code

Claude Code has two tools for accessing the web:

Claude will search the web and compile findings. However, remember these limitations:

Working with Property Data Files

Much of the most valuable research involves analysing files already on your computer or in shared drives. Claude Code excels at this:

Comparable Evidence Analysis

A core property skill is gathering and analysing comparable evidence for rent reviews and valuations. Claude Code can help structure this work:

Xero Financial Data

Claude Code connects to Xero via MCP to access your firm's financial data. This is useful for checking invoices, tracking payments, and financial reporting:

Property consultants spend a significant amount of time producing reports — from brief client letters to full valuation reports. Claude Code can dramatically speed up this process by generating structured first drafts that you then refine with your professional expertise.

Common Report Types in Property Practice

Structuring Report Prompts

For longer reports, break the work into sections rather than asking for everything at once. This gives better results because:

• Each section gets focused attention
• You can provide section-specific data and instructions
• You can review as you go, catching errors early
• Claude's context stays focused and relevant

Working with Google Drive

Claude Code can interact with your Google Drive to find, read, and create documents:

This is particularly useful when you need to find a previous report, check what has been sent to a client, or locate comparable evidence files.

Template-Based Documents

Your firm may have standard templates for common documents. Claude Code can use these as starting points:

By now you have used Claude Code for several practical tasks. This section introduces prompt patterns — reusable structures that consistently produce high-quality results. Learning these patterns is like learning keyboard shortcuts: a small upfront investment that saves time every day.

The CONTEXT-TASK-FORMAT Pattern

The single most useful prompt pattern. Every good prompt contains three elements:

The ROLE Pattern

Telling Claude to adopt a specific role changes the style and depth of its response:

By assigning a role, Claude draws on more relevant knowledge and uses appropriate professional language.

The CONSTRAINTS Pattern

Adding constraints focuses Claude's output and prevents common problems:

The ITERATION Pattern

Complex outputs rarely come out perfect first time. The iteration pattern uses follow-up prompts to refine:

The EXAMPLE Pattern

Showing Claude an example of what you want is often more effective than describing it:

Test your understanding of practical property tasks with Claude Code. Remember: up to 3 attempts per question with hints.

Up until now, you have been working with Claude Code in a request-response pattern: you ask something, Claude does it, you ask the next thing. Co-work mode changes this dynamic. Instead of sequential requests, you and Claude work on the same task simultaneously — like having a colleague sitting next to you.

When to Use Co-work Mode

Effective Co-work Techniques

To get the most from co-work mode:

One of Claude Code's most powerful features is agent teams — the ability to spin up multiple AI agents that work together on a task. Instead of one Claude doing everything sequentially, several specialised agents can work in parallel, each focusing on their area of expertise.

How Agent Teams Work

When Agent Teams Add Value

Agent teams are most valuable for tasks that benefit from multiple perspectives or parallel processing:

• Property research — One agent checks VOA data, another searches the web, another analyses local files
• Project audits — Multiple agents examine different aspects (code quality, security, documentation, performance)
• Fee proposals — Research agent gathers market rates, writer agent drafts the proposal, reviewer checks quality
• Board discussions — Multiple AI “directors” debate a strategic question from different perspectives

Example Team Commands

You can configure a library of custom skills (also called slash commands) tailored to your firm's specific business needs. These are pre-built prompt patterns and workflows that you can invoke with a single command.

Property Skills

These are the skills commonly useful for property professionals:

Business & Communication Skills

Using Skills in Practice

This single command triggers a comprehensive property intelligence report that checks up to 10 data sources: VOA ratings, EPC certificates, Land Registry ownership, flood risk, planning applications, and more. What would take you 30 minutes of manual research happens in seconds.

A significant limitation of AI chat is that each new session starts from scratch — Claude does not remember what you discussed yesterday. This is solved with Memory MCP and CLAUDE.md files, which give Claude persistent context across sessions.

How Memory Works

Types of information stored in Memory:

• Client information — preferences, history, contact details
• Project decisions — architecture choices, approach decisions with rationale
• Milestones — what has been completed, what is next
• Risk registers — identified risks and mitigations

CLAUDE.md Files

Projects typically have a CLAUDE.md file that gives Claude essential context about the project. When you start Claude Code in a project directory, it automatically reads this file.

Session Workflows

Structured workflows help maintain continuity across sessions:

Checkpoints and Handoffs

For long tasks that might span multiple sessions, Checkpoints help you pick up where you left off:

Test your understanding of co-work mode, agent teams, and custom workflows. Remember: up to 3 attempts per question with hints.

You have already learned the basic prompt patterns (CTF, ROLE, CONSTRAINTS, ITERATION). Now we go deeper into the principles that separate average prompts from excellent ones. These principles apply to every interaction with Claude Code.

Principle 1: Be Explicit, Not Implicit

Claude cannot read your mind. Information that seems obvious to you may not be obvious to an AI. Always state assumptions explicitly.

Principle 2: Provide Reference Material

When you want Claude to match a specific style or standard, show it an example rather than describing it abstractly:

Principle 3: Decompose Complex Tasks

Large tasks should be broken into smaller, manageable steps. This reduces errors and gives you checkpoints for review:

Principle 4: Define the Negative Space

Telling Claude what NOT to do is often as important as telling it what to do:

Principle 5: Request Reasoning

When you want Claude to show its working (not just give an answer), ask it to explain its reasoning:

Reusable prompt templates save time and ensure consistency across similar tasks. Many firms maintain a library of templates for common property work.

Building a Prompt Template

A good template has fixed structure with variable fields that change per use:

TEMPLATE: Rent Review Opening Letter

CONTEXT:
- Property: [ADDRESS]
- Tenant: [TENANT NAME]
- Landlord: [LANDLORD NAME]
- Current rent: [AMOUNT] per annum
- Review date: [DATE]
- Review basis: [UPWARD ONLY / OPEN MARKET]
- We act for: [LANDLORD / TENANT]

TASK:
Draft a formal opening letter for the rent review.
Include reference to the lease review clause.
State our initial position on ERV.
Propose a meeting to discuss.

FORMAT:
- Formal letter format (your firm's letterhead style)
- Maximum 1 page
- Professional but collaborative tone

Common Property Templates

Using Templates Effectively

Templates work best when you:

• Fill in all variables — empty fields lead to hallucinated content
• Customise for the situation — templates are starting points, not rigid forms
• Review the output critically — template-generated content still needs human review
• Update templates over time — improve them based on what works and what does not

For RICS-regulated firms, there are specific obligations around how AI is used in professional work. This section covers the intersection of AI tools and professional standards.

What AI Can and Cannot Do Under RICS Standards

Disclosure and Transparency

Mandatory position on AI disclosure under the RICS AI standard:

• Internal work — no client disclosure required, but maintain internal records of AI use
• Client-facing work — notify clients in writing that AI tools are used, specifying which parts of the process, and offer opt-out where practicable
• Formal reports — the signing surveyor is responsible for all content; document the AI review process
• General — be transparent and proactive about AI use; do not wait to be asked

The Human Oversight Framework

Every piece of AI-generated work should go through a human oversight process:

Using AI tools involves data security responsibilities and the need to handle situations when AI gets things wrong. This section covers both.

Data Security with Claude Code

When Claude Gets It Wrong

AI failures happen. Knowing how to identify and handle them is an essential skill:

Types of AI Failure

• Hallucination — Claude confidently states something incorrect. Example: inventing a comparable transaction that never happened
• Outdated information — Claude's training data has a cutoff. It may not know about recent legislation or market changes
• Misunderstanding context — Claude may apply the wrong jurisdiction (Scottish law instead of English) or wrong property type (residential instead of commercial)
• Arithmetic errors — Multi-step calculations can go wrong, especially with complex adjustments
• Tone misjudgement — An email may be too casual or too formal for the recipient

Handling Failures

This section introduces advanced Claude Code features that will become more useful as your confidence grows.

Playwright: Browser Automation

Claude Code can control a web browser via the Playwright MCP server. This enables automation of tasks that normally require manual web interaction:

• Navigating to a website and extracting data
• Filling in forms and submitting applications
• Taking screenshots of web pages for reports
• Checking the status of online planning applications

Bash: Running Commands

The Bash tool lets Claude run command-line operations. For property work, this is useful for:

• Processing multiple files at once (e.g., renaming, converting)
• Running scripts that connect to external APIs
• Managing Git repositories for version control of documents
• Installing or updating tools

Git: Version Control for Documents

Git is a system for tracking changes to files over time. It is commonly used for code projects, but it can also track changes to important documents:

Thinking Triggers

For complex analytical tasks, you can tell Claude to think more deeply by using specific triggers:

Test your understanding of best practices, RICS compliance, and advanced features. Remember: up to 3 attempts per question with hints.

Chain-of-thought (CoT) prompting is a technique that dramatically improves Claude's accuracy on analytical tasks by asking it to show its reasoning step by step, rather than jumping directly to an answer.

Basic Chain-of-Thought

The simplest way to invoke chain-of-thought is to add "Think step by step" or "Show your working" to your prompt:

Structured Chain-of-Thought for Property Analysis

For property-specific analysis, structure the chain of thought to match professional methodology:

When to Use Chain-of-Thought

Two powerful techniques that work together: few-shot learning (showing Claude examples of what you want) and persona engineering (defining who Claude should be for a specific task).

Few-Shot Learning

Zero-Shot vs One-Shot vs Few-Shot

Persona Engineering

A persona goes beyond a simple role assignment. It defines Claude's expertise, communication style, knowledge focus, and even personality traits:

Combining Techniques

The most effective prompts often combine multiple techniques. Here is an example combining persona, few-shot, and chain-of-thought:

This section provides tested prompt structures for common property tasks. Each has been refined through real use and produces consistently good results.

Rent Review Analysis Prompt

PERSONA: Senior RICS commercial surveyor, 20+ years experience,
specialising in [office/industrial/retail] rent reviews.

CONTEXT:
- Property: [address], [size] sq ft, [property type]
- Current rent: £[amount] pa (£[psf] psf)
- Review date: [date]
- Review basis: [upward only / open market]
- We act for: [landlord / tenant]

COMPARABLE EVIDENCE:
1. [address], [size] sq ft, [rent] psf, [date], [term], [incentives]
2. [repeat for each comparable]

TASK: Analyse these comparables and recommend an ERV position.

METHOD:
1. Assess relevance of each comparable
2. Adjust to effective rents
3. Weight by relevance
4. Apply adjustments for subject property
5. Recommend ERV range and preferred figure
6. Identify risks to our position

FORMAT: Structured report section with clear headings.
Do not include legal advice.

Lease Summary Prompt

Read the lease document at [file path].

Produce a structured summary covering:
1. PARTIES - Landlord, tenant, guarantor (if any)
2. PROPERTY - Description and demise
3. TERM - Commencement, expiry, any options to extend
4. RENT - Current rent, payment frequency, VAT position
5. RENT REVIEWS - Frequency, basis, assumptions/disregards
6. BREAK CLAUSES - Who can break, when, conditions, notice period
7. REPAIR - Landlord vs tenant obligations (FRI or internal only?)
8. ALTERATIONS - What requires consent?
9. ALIENATION - Assignment and subletting restrictions
10. USER CLAUSE - Permitted use, any restrictions
11. SERVICE CHARGE - Applicable? Cap? Sinking fund?
12. INSURANCE - Who insures? How are premiums recovered?

Flag any unusual or particularly important clauses.
Use plain English suitable for a non-lawyer client.

Fee Proposal Prompt

Draft a fee proposal for [client name].

SERVICE: [rent review advice / property management / valuation / etc.]
PROPERTY: [details]
SCOPE: [what is included and excluded]
FEES: [fee structure - fixed / percentage / hourly]
TIMELINE: [expected duration]
TEAM: [who will be involved]

Follow your firm's standard fee proposal format:
1. Introduction and understanding of instruction
2. Scope of services
3. Our approach and methodology
4. Team and qualifications
5. Fee schedule and payment terms
6. Terms and conditions
7. Contact details

Tone: Professional, confident, but not arrogant.
Emphasise your firm's local market knowledge and RICS credentials.

Market Commentary Prompt

Write a market commentary on [property type] property in [location].

PERIOD: [quarter/year]
AUDIENCE: [institutional client / private investor / general]
LENGTH: [word count]

Cover:
1. Market overview and key trends
2. Supply and demand dynamics
3. Rental growth/movement
4. Investment yields
5. Development pipeline
6. Outlook and risks

Do NOT include specific rental or yield figures unless I provide them.
Do NOT make specific predictions about future values.
Use British English throughout.

Professional efficiency with AI comes from having a curated library of prompts that you refine over time. This section guides you through creating and maintaining your personal prompt library.

What to Include in Your Library

Start with the tasks you do most frequently:

• Daily tasks — email drafts, meeting notes, quick calculations
• Weekly tasks — client reports, market updates, comparable analysis
• Monthly tasks — management reports, fee proposals, portfolio reviews
• Occasional tasks — valuation reports, expert witness prep, lease negotiations

Organising Your Library

Store prompts as text files in a dedicated folder. Claude Code can read these files directly:

/Users/yourname/prompts/
  /emails/
    client-update.txt
    rent-review-opening.txt
    meeting-confirmation.txt
  /reports/
    rent-review-analysis.txt
    lease-summary.txt
    market-commentary.txt
  /proposals/
    fee-proposal.txt
    terms-of-engagement.txt
  /analysis/
    comparable-analysis.txt
    yield-calculation.txt
    dcf-template.txt

Refining Prompts Over Time

Test your understanding of advanced prompting techniques. Remember: up to 3 attempts per question with hints.

Property consultancy involves vast amounts of document work — leases, licences, reports, correspondence, schedules, and plans. Claude Code can read and analyse these documents directly, saving hours of manual review.

What Claude Code Can Read

Reading Large Documents

Large documents like leases can be hundreds of pages. Claude Code handles these efficiently by reading in sections:

Reading Strategies for Different Document Types

Extracting structured data from leases is one of the highest-value applications of Claude Code in property. A task that might take 30–60 minutes manually can be done in 2–3 minutes, with you reviewing the output for accuracy.

Standard Lease Extraction Fields

The core data points to extract from any commercial lease:

Structured Extraction Prompt

Comparing Multiple Leases

When working with a portfolio, Claude Code can extract data from multiple leases and compile it into a comparison table:

Building on the document skills from Module 3, this section covers more advanced report and letter generation, including multi-section reports, formatted output, and integration with Google Docs.

Multi-Section Report Generation

For full-length property reports, use a staged approach:

Professional Letter Generation

Professional letters should follow a standard format. Claude Code can produce them quickly with the right template:

Document Formatting and Output

Claude Code can output documents in several formats:

• Google Docs — created directly via Google Workspace MCP
• Local text files — saved to your computer using the Write tool
• Markdown — useful for structured content that can be converted to other formats
• HTML — for web-based reports or email content

One of Claude Code's most powerful capabilities is processing multiple documents or data items in a single operation. This is especially valuable for portfolio management, where the same analysis needs to be applied across many properties.

When to Use Batch Processing

• Portfolio rent reviews — extracting key dates from 20+ leases
• EPC checks — looking up energy ratings for all properties in a portfolio
• Tenant covenant checks — researching multiple tenants' financial health
• Comparable analysis — processing multiple sources of comparable data
• Correspondence review — summarising a chain of letters on a dispute

Batch Processing Examples

Best Practices for Batch Operations

Test your understanding of document processing and analysis. Remember: up to 3 attempts per question with hints.

Effective property research combines multiple sources: web searches, local files, databases, and professional knowledge. Claude Code brings these together in a structured way, but you need to guide the research strategy.

Research Strategy Framework

Types of Property Research

Market analysis goes beyond data gathering. It requires interpreting trends, understanding context, and forming professional opinions. Claude Code is excellent at structuring analysis; you provide the professional interpretation.

Rental Market Analysis

Investment Yield Analysis

Claude Code can calculate and compare investment yields, but the interpretation requires professional judgement:

Comparative Analysis

One of the most powerful uses of Claude Code is comparing data across multiple dimensions:

Claude Code can access a wide range of data sources. Understanding what each provides and how reliable it is helps you build comprehensive, accurate analysis.

The Property Data Toolkit

Integrating Multiple Sources

The real power comes from combining data sources into a single, comprehensive view:

The ability to fact-check AI-generated research is the skill that separates effective AI users from dangerous ones. This section provides a systematic approach to verification.

The Verification Framework

Red Flags in AI-Generated Research

Asking Claude to Self-Check

You can ask Claude to review its own output for potential errors:

Test your understanding of research, analysis, and fact-checking. Remember: up to 3 attempts per question with hints.

A workflow is a sequence of steps that you follow regularly for a specific type of task. By designing workflows around Claude Code, you can dramatically reduce the time and effort required for routine property work while maintaining consistency and quality.

Workflow Design Principles

Workflow Example: New Instruction Setup

When your firm receives a new client instruction, a standardised workflow ensures nothing is missed:

Workflow Example: Monthly Rent Collection Review

Building Your Own Workflows

Identify tasks you do repeatedly and design a workflow:

1. Map the current process — write down every step you currently take
2. Identify AI-automatable steps — which steps can Claude handle?
3. Design review checkpoints — where should humans review before proceeding?
4. Write the workflow prompt — create a reusable prompt that executes the workflow
5. Test and refine — run the workflow with real data and improve it

Claude Code's power comes partly from its ability to connect to external services through MCP servers and APIs. Understanding how these connections work helps you use them more effectively.

How MCP Servers Work

MCP servers act as bridges between Claude Code and external services:

Common Active Integrations

Multi-Service Workflows

The most powerful workflows combine multiple services in a single operation:

While you may not write code as part of your surveying role, understanding scripts and version control helps you work more effectively with Claude Code and maintain professional accountability.

Understanding Scripts

You do not need to write scripts yourself — Claude Code can create them for you. But understanding what they do helps you verify that Claude is doing the right thing:

Git Basics for Non-Developers

Git is the version control system that tracks all changes to files over time. Claude Code uses Git to manage projects. You should understand the basics:

You do not need to run Git commands yourself. Claude Code handles version control automatically. But understanding the concepts helps you appreciate the audit trail that Git provides — which is valuable for professional accountability.

Things do not always go smoothly with AI tools. This section covers the most common issues and how to resolve them.

Common Issues and Solutions

When to Escalate

Most issues can be resolved by adjusting your prompt or restarting the session. Escalate to a senior colleague when:

• An MCP server consistently fails to connect (may need authentication refresh)
• Claude Code itself will not start or crashes
• You discover a security concern (data appearing where it should not)
• You need access to a new service or tool that is not currently configured

Maximising Performance

Test your understanding of automation, workflows, and troubleshooting. Remember: up to 3 attempts per question with hints.

Using AI tools with client data brings data protection responsibilities under UK GDPR and the Data Protection Act 2018. As a property professional, you must understand these obligations.

UK GDPR and AI: The Basics

Personal Data Categories

AI Data Policy Best Practice

Data Processing Agreement

Your firm should have a data processing arrangement with Anthropic (Claude's maker) that means:

• Data sent to Claude is not used for AI training
• Data is encrypted in transit and at rest
• Data is not shared with third parties
• Your firm can request data deletion if needed

This arrangement provides a level of protection, but it does not eliminate your responsibility to handle data carefully.

Building on the RICS principles from Module 5, this section examines how specific RICS standards apply to AI-assisted property work in practice.

RICS Global Professional and Ethical Standards

Five key RICS principles and how they apply to AI use:

AI in Valuations

Valuations are the most regulated area of surveying practice. AI use requires particular care:

AI in Client Advice

When Claude Code helps produce client advice (rent review recommendations, lease advisory, investment analysis), remember:

AI systems can reflect and amplify biases present in their training data. Understanding these risks helps you use AI responsibly in property practice.

Types of AI Bias in Property

Mitigating Bias

Professional Liability

AI use does not change your firm's professional liability. Key points:

• Professional indemnity insurance covers AI-assisted work, provided normal professional standards are followed
• "AI error" is not a defence — you are responsible for output you approve and send
• Duty of care is the same whether you used AI or not
• Record keeping must document that AI was used and what human review was performed. This is a mandatory requirement under the RICS AI standard
• If in doubt, escalate — discuss with a senior colleague before issuing advice you are uncertain about

Completing this course marks the beginning, not the end, of your AI journey. The property professionals who thrive in the coming years will be those who combine strong traditional skills with confident AI use. You are now one of them.

What You Have Learned

Your Ongoing Development Plan

The AI-Enhanced Surveyor

Your value as a property professional is not diminished by AI — it is amplified. The market needs people who can:

• Ask the right questions — AI answers questions; you decide which questions to ask
• Apply professional judgement — AI processes data; you interpret what it means for the client
• Build client relationships — AI drafts communications; you build trust and rapport
• Inspect properties — AI analyses documents; you see the building in person
• Take responsibility — AI assists; you sign your name and stand behind the advice

Resources for Continued Learning

• Anthropic documentation — docs.anthropic.com for the latest Claude features
• Claude Code release notes — new features are added regularly
• RICS (2025) Responsible use of artificial intelligence in surveying practice, 1st edition — mandatory from 9 March 2026. Available at rics.org
• Your team — ask colleagues for help and share what you learn
• This course — come back to refresh specific skills. The quizzes can be reset and retaken

Final quiz — test your understanding of AI ethics, data protection, and professional practice. Remember: up to 3 attempts per question with hints.

© 2026 Hillway Holdings Limited. All rights reserved.

Built with Claude Code — the tool you just learned to use.