Documentation - Cadence IP

What is the knowledge base?

The knowledge base is a collection of domain knowledge that the AI assistant draws on when working within a specific matter type. Each matter type has its own knowledge base, so the AI's understanding of PCT applications is different from its understanding of trademark oppositions.

Think of it as your firm's institutional knowledge, codified in a way the AI can use. Instead of the AI relying solely on its general training data, it references your firm's specific practices, standard approaches, fee structures, and procedural knowledge.

Without knowledge base

The AI drafts a generic examination response that misses your firm's standard structure, doesn't reference the correct fee schedule, and uses American terminology instead of Australian.

With knowledge base

The AI follows your firm's reporting template, quotes fees from your schedule, references the Patents Act 1990 (Cth), and uses your preferred tone and structure.

Manual content editing

Navigate to your matter type configuration and open the Knowledge Base tab. The editor lets you write free-form content that teaches the AI about this type of work.

Good knowledge base content includes:

Procedural knowledge — how you handle each stage of the workflow, typical timelines, what triggers the next step
Standard approaches — your firm's preferred strategy for responding to common objections or office actions
Fee structures — what you charge for each type of work, official fees, typical disbursement ranges
Terminology preferences — jurisdiction-specific terms, preferred phrasing, words to avoid
Client communication standards — how formal to be, what to include in reports, standard disclaimers

Building a PCT Application knowledge base entry

You add content describing your firm's standard approach to PCT written opinion responses:

Written Opinion Response Process:

When we receive a Written Opinion from the ISA, we follow this process:

1. Review the opinion and identify each objection (novelty, inventive step, industrial applicability, clarity).

2. For each cited document, assess relevance and identify distinguishing features of the invention.

3. Report to the applicant within 5 business days, including: (a) summary of each objection, (b) our recommended response strategy, (c) estimated fees for preparing the response, (d) the Chapter II deadline if applicable.

4. Our standard fee for a written opinion response is $2,500-$4,500 depending on complexity. Simple amendments are at the lower end; arguments requiring detailed technical analysis are at the higher end.

Auto-learning settings

Beyond manual content, the knowledge base can learn automatically from the work you do within Cadence IP. Configure auto-learning settings on each matter type's knowledge base:

Train on emails

When enabled, outgoing emails sent from matters of this type are analysed to learn your firm's communication style, standard phrases, and response patterns. The AI picks up how you structure reporting letters, deadline reminders, and filing confirmations.

Train on documents

When enabled, documents generated from templates are analysed to learn the structure and content of your standard documents. This helps the AI understand your document conventions when drafting new content.

Privacy consideration

Auto-learning processes content through your configured AI provider. Ensure your AI provider's data handling policy is compatible with your firm's confidentiality obligations. Content is not stored by the AI provider beyond the processing request.

Start with manual, then enable auto-learning

Begin by writing manual knowledge base content for your most important workflows. Once you are satisfied with the AI's baseline performance, enable auto-learning to gradually improve its understanding from real work.

Training instructions

Training instructions tell the auto-learning system what to focus on when processing emails and documents. Set these in the Training Instructions field on the knowledge base settings.

Training instructions for a Trademark matter type

When learning from emails, focus on:

- How we report examination results to trademark applicants

- Standard objection categories we encounter (distinctiveness, deceptive similarity, prior marks)

- Our recommended strategies for overcoming each type of objection

- Fee estimates we provide for different response scenarios

- The structure and tone of our client-facing correspondence

Do NOT learn from:

- Internal notes or casual messages between team members

- Automated notification emails from IP offices

How the AI uses the knowledge base

The knowledge base is automatically included in the AI's context when you interact with it on a matter page. The AI uses this knowledge in several ways:

Email drafting

When proposing emails, the AI follows the communication patterns and structure it learned from your knowledge base and past emails.

Chat responses

When answering questions about procedure, fees, or strategy, the AI references your firm's specific knowledge rather than generic information.

Action proposals

When proposing workflow actions (step changes, tasks), the AI uses knowledge base content about your standard procedures to suggest the right next steps.

AI guidances

Guidances are named instruction documents within the knowledge base that you can invoke on demand using @mention syntax in the AI chat. Unlike the general knowledge base content (which is always included), guidances are only activated when you explicitly reference them.

This is useful for specialised instructions that should not apply to every interaction. For example, you might have a guidance for reporting written opinions that includes a very specific email template — you only want the AI to use that template when you are actually reporting a written opinion.

1

Create a guidance

In the knowledge base editor, click + New Guidance. Give it a short name (this becomes the @mention handle) and write the instructions.

2

Write detailed instructions

Be specific about what the AI should do when this guidance is activated. Include template structures, required sections, tone guidance, fee references, and any other detail the AI needs.

3

Use in chat with @mention

When chatting on a matter page, type @ followed by the guidance name. The guidance's instructions are added to the AI's context for that interaction.

Creating a @reporting_written_opinion guidance

You create a guidance called reporting_written_opinion with these instructions:

When reporting a PCT Written Opinion to the applicant, structure the email as follows:

1. Opening paragraph: Acknowledge receipt of the Written Opinion, state the ISA, and note the date.

2. Summary section: For each objection, provide a subheading (e.g., "Novelty - Claim 1"), cite the relevant document(s), and summarise the examiner's position in plain language.

3. Recommendations: For each objection, state our recommended approach (amend, argue, or both) and explain why.

4. Fees: State our estimate for preparing a response. Standard range is $2,500-$4,500. Simple amendments: lower end. Complex arguments: higher end.

5. Deadline: State the Chapter II demand deadline (22 months from priority) and the Article 34 amendment deadline.

6. Closing: Ask the applicant to confirm instructions by [deadline minus 4 weeks].

Now when you type @reporting_written_opinion in chat and ask the AI to draft the report, it follows this exact structure.

Build a library of guidances

Over time, build guidances for every common task: @reporting_examination, @fee_estimate, @filing_confirmation, @deadline_reminder. This creates a library of firm-specific AI expertise that any team member can invoke.