Start your journey to becoming a superuser with our learning materials designed to cover all you need to know from analysing documents with the Assistant to building the best Playbooks on the Marketplace.
Discover the transformative potential of Playbooks through real-world success stories. This inspi...
Jylo Word Assistant (Beta)
Bring Jylo's AI directly into Microsoft Word with this quick video demo. The Jylo Word Assistant is a side-panel add-in that lets you chat with any open .docx file — asking questions, requesting summaries and instructing edits as tracked changes — all powered by the AI models enabled on your Tenant.
What You'll Gain:
How to ask questions about your document and receive answers with paragraph-level citations
How to generate summaries of a full document or specific sections
How to instruct the Assistant to draft new content, clauses or paragraphs in context
How to apply, review and accept or reject AI-generated edits as tracked changes
How to select your preferred AI model from the model picker
What This Demo Covers: This demo walks you through the core workflows of the Jylo Word Assistant: analysing, summarising, drafting and editing a .docx file directly from the side panel, with particular attention to the tracked changes workflow. The full Jylo Word Assistant Guide and Word Plug-In Deployment Guide are attached for reference. Note that the Word Assistant is currently in Beta, so chat history is not retained between sessions and Playbook connections are not yet supported. This is essential viewing for anyone looking to bring Jylo's drafting and analysis capabilities directly into their everyday Word workflow.
Jylo provides access to a curated selection of Artificial Intelligence (AI) models from leading providers. Each model has different strengths, and choosing the right one for your task can improve both the quality and efficiency of your analysis.
This guide explains the models available on the platform, what each one is best suited to, and what to keep in mind when reviewing AI-generated outputs. It is designed to support your organisation’s transparency, governance and compliance obligations—including the Algorithmic Transparency Recording Standard (ATRS).
Tip
Not sure which model to choose? For everyday tasks with well-defined instructions, start with Claude Haiku 4.5 or GPT-4.1—both are fast, reliable and cost-effective. For tasks that are less well defined and require reasoning or intuition, step up to Claude Sonnet 4.6 or GPT-5. For ambitious, complex work where you need the deepest analysis, choose Claude Opus 4.6.
How Jylo Uses AI Models
Jylo does not build or train its own AI models. Instead, it provides a secure, governed workspace in which commercially available models are accessed via API to perform document analysis tasks that you define.
You interact with AI models in two ways on Jylo:
Assistant: A conversational chat interface where you can ask questions about your documents, compare files and conduct ad-hoc analysis.
Playbooks and Flows: Structured, repeatable workflows where your prompts are applied to one or many documents individually, with results subject to human verification.
In both cases, you choose which model to use, what questions to ask and whether to accept the AI’s output. Your data is never used to train AI models, and all interactions remain within your organisation’s secure environment.
Available Models
Jylo offers several models across three providers. Your administrator controls which models are available in your region. Models convert text into tokens for processing; each model has a token limit. 100 thousand tokens are equivalent to roughly 75 thousand words.
GPT-5
Provider
OpenAI (hosted on Microsoft Azure)
Best for
Tasks requiring reasoning and intuition—regulatory analysis, nuanced legal interpretation, multi-step logic where instructions may be less well defined
Strengths
Near-frontier intelligence. Strong reasoning and multimodal capabilities. Excellent for complex, multi-step analysis.
Limitations
Significantly slower than other models. Affordable credit consumption, especially for a reasoning model.
Context window
400,000 tokens
Knowledge cutoff
September 2024
GPT-4.1
Provider
OpenAI (hosted on Microsoft Azure)
Best for
Everyday tasks with well-defined instructions—summarisation, question-answering, data extraction and lengthy document review.
Strengths
One of the fastest models on the platform. Largest context window (1M tokens)—ideal for very long documents. Strong all-round performer.
Limitations
Less suited to tasks requiring deep multi-step reasoning compared with GPT-5. May require tailored prompting for highly specialist terminology.
Context window
1,000,000 tokens
Knowledge cutoff
June 2024
Tip
GPT-4.1 and Claude Haiku 4.5 are recommended for everyday tasks with clear, well-defined instructions. GPT-4.1’s advantage is its enormous context window (1M tokens), making it the best choice when working with very long documents.
o3-mini
Provider
OpenAI (hosted on Microsoft Azure)
Best for
Fast, affordable reasoning tasks—structured analysis, logic-driven prompts and compliance checks where cost matters
Strengths
Reasoning model that “thinks through” problems step by step. Moderate end-to-end speed (approx. 11 seconds including thinking time). Cost-effective.
Limitations
Lower overall intelligence than GPT-5 or Claude Opus 4.6. Not recommended for the most complex, high-stakes analysis.
Context window
200,000 tokens
Knowledge cutoff
October 2023
Claude Opus 4.6
Provider
Anthropic (hosted on Amazon Web Services)
Best for
Ambitious, complex work requiring the deepest analysis—multi-document reasoning, high-stakes compliance, nuanced interpretation where maximum intelligence matters
Strengths
Highest intelligence of any model on the platform. Strong reasoning and low hallucination rates. Well suited to detailed legal and regulatory work.
Limitations
Highest credit consumption among Claude models. Slower response times for very large inputs.
Context window
1,000,000 tokens
Knowledge cutoff
Early 2025
Claude Sonnet 4.6
Provider
Anthropic (hosted on Amazon Web Services)
Best for
Tasks requiring reasoning and intuition at a moderate cost—contract review, regulatory analysis and less well-defined questions where the model needs to interpret intent
Strengths
Near-frontier intelligence at moderate cost. Strong reasoning capabilities without the premium cost of Opus.
Limitations
Does not match Opus on the most complex tasks. Slower than some OpenAI models.
Context window
1,000,000 tokens
Knowledge cutoff
Early 2025
Claude Haiku 4.5
Provider
Anthropic (hosted on Amazon Web Services)
Best for
Everyday tasks with well-defined instructions—document triage, classification, structured extraction, batch processing and cost-sensitive workflows
Strengths
Fastest model on the platform. Strong intelligence for its cost—outperforms GPT-4.1 and o3-mini on benchmark intelligence. Ideal when speed, volume and value matter.
Limitations
Smaller context window than GPT-4.1. Less suited to highly complex multi-step reasoning compared with Opus or Sonnet.
Context window
200,000 tokens
Knowledge cutoff
April 2024
LLama 3.3
Provider
Meta (hosted on Microsoft Azure)
Best for
Budget-friendly, high-volume tasks—simple classification, rapid triage and straightforward extraction where cost is the priority
Strengths
Open-source model. Fast and cost-effective. Good for large batches of straightforward tasks.
Limitations
Lower intelligence than proprietary models. Not suited to complex legal, regulatory or multi-step analysis. Best paired with thorough human review.
Context window
128,000 tokens
Knowledge cutoff
December 2023
Quick Comparison
Use this table to compare models at a glance:
Model
Provider
Intelligence
Speed (E2E)
Cost
Best For
GPT-5
OpenAI
★★★★
★
£½
Reasoning and intuition
GPT-4.1
OpenAI
★★★
★★★★★
£½
Everyday, well-defined tasks
o3-mini
OpenAI
★★★
★★★
£
Fast, affordable reasoning
Claude Opus 4.6
Anthropic
★★★★★
★★★
£££
Ambitious, complex work
Claude Sonnet 4.6
Anthropic
★★★★
★★★
££
Reasoning and intuition
Claude Haiku 4.5
Anthropic
★★★½
★★★★★
£
Everyday, well-defined tasks
LLama 3.3
Meta
★★
★★★★★
½
Budget high-volume tasks
What to Keep in Mind
All AI models have limitations. Understanding these helps you get the best results from Jylo and ensures your organisation uses AI responsibly.
Bias
AI models are trained on large bodies of text from the internet and published sources. This training data may contain historical, cultural or demographic biases that can surface in outputs. Models may also default to US-centric legal or regulatory assumptions—if you work within UK or other jurisdictions, frame your prompts accordingly and verify outputs against local requirements.
Hallucination
Models can sometimes generate responses that sound authoritative but contain factual errors. This is why Jylo’s verification workflow exists—every AI output in a Flow can be approved, modified or rejected by a human before it is treated as final.
Prompt Sensitivity
Small changes in how you phrase a question can produce different results. Jylo’s Playbook system helps standardise your prompts so that the same methodology is applied consistently across documents.
Important
AI outputs on Jylo should always be treated as a starting point for human review, not as a final determination. Use the verification workflow to approve, modify or reject every output before relying on it.
How to Select a Model
In the Assistant
Click the model dropdown menu in the top-left corner of the Assistant interface.
Select your preferred model from the list.
The Assistant will use your selected model for all subsequent messages until you change it.
In Playbooks
Each prompt within a Playbook can be configured to use a specific model. When creating or editing a prompt, select the desired model from the model dropdown in the prompt configuration form.
Administrator Controls
Administrators control which models are available across the organisation. To configure model availability:
Navigate to Admin > Settings > Models.
Select the desired region tab.
Check or uncheck models to enable or restrict them.
Data Handling and Security
Your data is private: Jylo never shares your data with third parties. Your data is never used to train AI models.
Infrastructure: OpenAI and LLama models are hosted on Microsoft Azure. Anthropic Claude models are hosted on Amazon Web Services.
Access control: Role-based permissions govern who can access Projects, Playbooks and model configurations.
For further information on Jylo’s security posture and certifications, visit trust.jylo.ai.
Next Steps
Try different models: Experiment with different models in the Assistant to see which one works best for your documents and questions.
Access our Guides: For user guides and video demonstrations visit our support.jylo.ai.
Contact support: For any questions, reach out to support@jylo.ai.
Jylo has enhanced the User Preference Prompt feature by introducing Output Style Templates—pre-built configurations that instantly customise how the Assistant responds to your queries.
Previously, users needed to write their own preference prompts from scratch to achieve a particular response style. Now, you can select from five professionally designed templates that match common roles and industries, making personalisation immediate and effortless.
Note
Your preference prompt is personal to you and is not shared with other members of your organisation. It automatically applies to every Assistant conversation you have.
How It Works
The User Preference Prompt is automatically appended to every message you send to the Assistant. By selecting an Output Style Template, you define how you want the AI to communicate with you—whether that's formal legal analysis, friendly guidance, or executive-focused insights.
This personalisation layer ensures that every interaction with the Assistant is tailored to your professional context and communication preferences.
Accessing Output Style Templates
To set or change your Output Style Template:
Click your profile icon (showing your initial) in the top-right corner of Jylo
Select Edit User Preference Prompt from the dropdown menu
The "Edit User Preference Prompt" dialogue will open
Click to expand the "Choose from Output Style Templates" section
Review the available templates and their descriptions
Click Use Template on your preferred style
The template will populate in the text field below
Click Update Prompt to save your selection
Tip
You can edit any template after selecting it to further customise the instructions. Click in the text field to make adjustments before clicking "Update Prompt".
Available Output Style Templates
Jylo currently offers five specialised templates designed for different professional contexts:
Consumer
Description: Friendly, visual answers with emojis and guided follow-ups
Best for: General users who prefer conversational, approachable responses with visual elements and step-by-step guidance
Example use cases: Learning new concepts, general enquiries
Legal
Description: Precise, cautious legal analysis with structured reasoning
Best for: Legal professionals requiring formal analysis, citation-heavy responses, and careful consideration of edge cases
Example use cases: Contract review, legal research, regulatory compliance analysis
Government
Description: Policy-focused, neutral tone aligned with public sector standards
Best for: Public sector professionals who need politically neutral, policy-oriented responses that align with governmental communication standards
Example use cases: Policy development, public communications, regulatory guidance
Real Estate
Description: Market-aware insights with property metrics and deal implications
Best for: Property professionals requiring responses that incorporate market data, valuation considerations, and transactional context
Example use cases: Property valuations, market analysis, lease negotiations
Executive / Business
Description: Concise, decision-oriented answers for leadership and strategy
Best for: Senior leaders and business strategists who need bottom-line focused, actionable insights without unnecessary detail
Example use cases: Strategic planning, board reports, executive briefings
Creating Custom Preference Prompts
While templates provide instant customisation, you may want to create a fully bespoke preference prompt. To write your own:
Follow steps 1-4 above to open the User Preference Prompt dialogue
Instead of selecting a template, click directly in the text field at the bottom
Write your custom instructions describing how you want the Assistant to respond
Click Update Prompt to save
Custom prompt examples:
"Always provide responses in bullet points with no more than 5 key points"
"When analysing documents, prioritise risk identification over opportunities"
"Use UK English spelling and legal terminology appropriate for English courts"
"Structure all responses with an executive summary, detailed analysis, and recommended actions"
The Jylo Word Add-In enables users to upload and create docx templates in Playbooks ready to be populated with Flow results. This guide provides step-by-step instructions for deploying the add-in to your organisation through the Microsoft 365 admin centre.
Prerequisites
Administrative access to the Microsoft 365 admin centre
Permissions to deploy custom apps in your organisation
Deployment Instructions
1. Access the Microsoft 365 admin centre
Navigate to the Microsoft 365 admin centre at admin.microsoft.com
2. Navigate to Integrated Apps
In the left navigation menu, click Show All
Select Settings
Click Integrated Apps
3, Upload the custom app
Click the Upload custom apps button
4. Select app type
When prompted for App Type, select Office Add-In
5. Provide the manifest file link
Enter the manifest file URL that corresponds to your organisation's region:
Region
Manifest File URL
United Kingdom
https://word-addin-001.uk.jylo.ai/manifest.xml
United States
https://word-addin-001.us.jylo.ai/manifest.xml
European Union
https://word-addin-001.eu.jylo.ai/manifest.xml
6. Configure user access
Select who should have access to the Jylo Word Add-In:
Specific users: Make the add-in available to individual named users
Groups: Make the add-in available to all members of specific groups
Entire organisation: Make the add-in available to all users organisation-wide
7. Deploy the add-in
Click the Deploy button to complete the installation. The add-in will appear for selected users within a few minutes.
Verifying Deployment
To confirm the add-in has been successfully deployed:
Open Microsoft Word (desktop or web version)
Navigate to the Home tab
Click the Jylo button in the right side of the ribbon
Sign in using your Jylo credentials
Support
If you encounter issues during deployment or have questions about the Jylo Word Add-In, our support team is here to help. Contact us at support@jylo.ai.
Master creating and managing Projects in Jylo with this comprehensive guide and accompanying video. Projects are the foundation of AI-powered document analysis in Jylo, providing secure collaborative workspaces where teams upload documents, run analyses, and manage workflows.
What You'll Gain:
How to create new Projects
How to upload Project files
Manage team members and permissions
Know where to launch a flow on your documents
Where to collaborate with the Projects Assistant
What This Guide Covers: This guide describes the concept of project centricity in Jylo, how to create and configure new Projects with appropriate metadata, add team members and manage permissions, detail Project settings and ongoing management options, and provide guidance on notifications and collaboration features. This is essential guidance for any new users to the platfrom.
Jylo is an AI-powered platform that allows you to automate your organisational intelligence.
The platform focuses on combining AI capabilities with human expertise, helping you work faster whilst maintaining quality through verification.
It comes in two parts – the AI Assistant, and Playbooks. Assistant is a classic chat-style AI tool. If you have a quick question, want a spur-of-the-moment document review or to perform a redline comparison, the Assistant can help.
Playbooks are automated workflows that use AI to not only extract your chosen data but also perform analysis and transform that base data into something completely different such as a letter or a report.
Everything in Jylo happens within Projects—organised workspaces that contain your files, analysis, team members, and results in one secure location.
Essential Setup
Each Project is typically dedicated to a particular department or initiative. Here, all related files and work are stored and shared amongst individuals chosen to form part of the team.
1. Create Your First Project
Navigate to My Projects in the "Me" section of the sidebar
Click the blue Plus button in the top-right corner
Enter a descriptive Project name
Provide a Project description
Complete any required metadata fields
Click Finish to create your Project
2. Upload Documents
Select your Project and go to the Files section
Click the blue Plus button
Select Upload files from the dropdown
Either drag and drop files or click browse to select from your computer
Assistant is perfect for answering quick questions, analysing various documents at once or having dynamic conversations which develop with every interaction. It's easy to use and can help you:
Prepare for meetings
Learn new concepts
Perform redline analysis
Compare documents of various types
you the overview with prompt labels and descriptions.
Note
Like a knowledgeable intern, the Assistant can be used to help you with ad-hoc tasks. However, you should always be sure to check its work.
To launch the Assistant:
Click Assistant in your Project navigation
Click the blue plus button to start a new conversation
Upload specific documents using the plus button in the input field, or ask general questions
Use natural language queries:
"Compare the agreement to the issues list"
"List the contractual milestones in table format"
"Differentiate between completion accounts and the locked box mechanism"
Always verify responses when analysing documents by clicking citation numbers to see highlighted source text
Playbooks and Flows
The Assistant provides rapid clarity and analysis whenever you need it. However, to consistently automate larger tasks across one or many documents, Playbooks are the best option. These templates enable users to apply high-quality prompts consistently and at scale. They are typically used to:
Spot trends across documents
Analyse documents using proven methodology
Structure large datasets before separating and storing files
Identify crucial information quickly from an abundance of data
Produce bespoke reports for single documents or entire datasets
Democratise methodology throughout your organisation
Tip
Once you are comfortable using Assistant, try using a Playbook to automate your most frequent tasks. By using a Playbook, you'll improve output consistency and save time.
To launch a Flow:
Go to Flows section in your Project
Click blue Plus button → Add Flow
Enter a Flow name and description
Select your chosen Playbook
Select your target documents
Click Finish to launch the analysis
Once the Flow is complete, review and verify results:
Use the left panel filters to find documents matching specific criteria
Click Evidence buttons to see AI reasoning highlighted in documents
Approve, reject, or edit responses using the verification controls
Use Tasks to assign specialist reviews to team members
Getting Results Out
From Assistant
Export PDF: Download the entire conversation thread
Copy the output: Save to Word or other applications
Continue conversations to refine analysis
From Flows
Click the download icon to access export options:
Export Spreadsheet: Comprehensive tabular data including verification status and comments for further analysis
Export with source documents: Complete package including original files
Generate flow assembly: Bespoke report comparing all documents based on the Playbook template
Generate document assembly: Individual reports for each document based on the Playbook template
Next Steps
Add team members: Go to Team in your Project to invite colleagues and assign roles
Explore more Playbooks: Browse the Marketplace to find analysis templates for different document types
Create custom workflows: Once you've identified effective prompts in Assistant, consider building custom Playbooks for repeated use
Visit our support portal at support.jylo.aifor comprehensive guides including the User Guide, Playbook Prompting, Assembly Building, and more.
Playbooks are reusable workflow templates that define how AI analysis is applied to documents within Jylo. They contain specific prompts and logic that determine what information to extract, how to process it, and how to present the results.
Benefits of Using Playbooks
Consistency: Apply the same analytical approach across all documents
Efficiency: Save time by reusing established analysis workflows
Customisation: Configure complex analyses with conditional logic
Quality Control: Build verification directly into the workflow
Collaboration: Share expertise through standardised processes
Working with Playbooks
Playbooks, Flows and Assemblies
It's important to understand how Playbooks fit into the Jylo workflow:
Playbooks define the analysis pattern (what questions to ask, what logic to apply)
Flows are where you apply these Playbooks to documents and carry out your human review of the AI process
Assemblies are how you bring together the AI output and use them outside of Jylo
You'll select which Playbook to use when you create a new Flow within a Project. This is when you'll decide which analytical approach is right for your specific documents.
The Playbook Marketplace
The Marketplace serves as a central repository where you can browse, access, and (with appropriate permissions) manage Playbooks.
To access the Marketplace:
Navigate to Marketplace in the left sidebar
Select Playbooks from the submenu
Each Playbook appears as a card displaying basic information. Clicking on a card shows you the overview with prompt labels and descriptions.
Note
The Marketplace is particularly useful for users with Builder access who create and manage Playbooks. Regular users will primarily interact with Playbooks when creating Flows within Projects although all users are encouraged to browse the marketplace to see what’s possible.
Playbook Access Control
Playbooks have specific access controls that determine who can use and modify them:
Builder permission: Enables a user to clone, edit and modify the Playbook after publication
Consumer permission: Makes the Playbook visible to users who can deploy it in Flows
Your access to specific Playbooks depends on the permissions assigned to you by the Playbook creator.
Creating Playbooks
You can create custom Playbooks from the Marketplace by following these steps:
Access the Marketplace Navigate to Marketplace > Playbooks and click the blue plus button in the top-right corner
Metadata Enter the Playbook Name and Description
Prompts Set the Label, Description, Question, Answer Type, and Model Upload test and source documents using the right panel
Assembly Use flow assembly to consolidate results from across the dataset Use document assembly to create a report for individual documents
Testing Create a test Flow with the blue plus button
Access Control Assign Builder or Consumer permissions to specific users
Summary Review your Playbook overview
Tip
Use clear, descriptive names that reflect the Playbook's purpose and include version numbers for easy identification and tracking
Next Steps
Maybe now you you’d like to have a shot at building your own Playbook. For full guides on Playbook building visit our Playbook Prompting and Assembly Building guides.
Discover the transformative potential of Playbooks through real-world success stories. This inspirational guide showcases how organisations across different industries have automated complex workflows and achieved significant productivity gains.
What You'll Gain:
Inspiration from real-world Playbook implementations
Understanding of diverse use cases across industries
Ideas for workflows you could automate in your practice
Confidence that AI can handle complex, expert-level tasks
Practical examples of problem-solution-result frameworks
What This Guide Covers: This guide presents anonymised case studies from actual Jylo implementations, including contract reviews, compliance assessments, due diligence processes, building safety case compliance, statements of work generation, and traineeship application reviews. Each success story follows a clear structure showing the challenge faced, the Playbook solution implemented, the results achieved, and practical insights for your own practice. You'll see how organisations have automated everything from simple checklists to complex multi-document assessments, demonstrating that with the right expertise and structure, almost any repetitive analytical workflow can be templatised.