Skip to main content

Intelligence Orchestration

New

Multi-model consensus, persistent AI memory, intelligent model routing, notebook collaboration, and custom AI instructions — your research team, automated.

Overview#

Intelligence Orchestration transforms your research experience from a single-model chatbot into a multi-model research team. Five capabilities work together to give you higher-confidence answers, persistent context, automatic model selection, team collaboration, and project-specific AI instructions.

Model Council

Query 3 frontier AI models simultaneously. See where they agree and disagree with consensus scoring.

AI Memory

The AI remembers your preferences, expertise, and instructions across all sessions automatically.

Auto-Routing

Automatic model selection based on your query type — regulatory, financial, academic, code, or general.

Shared Notebooks

Invite team members to collaborate in research notebooks with role-based access control.

Custom Instructions

Set notebook-specific AI guidelines that apply to every chat message in the notebook.

Model Council#

Model Council runs your question against three different frontier AI models simultaneously and shows you where they agree and disagree — giving you higher-confidence answers backed by multi-model consensus. Find it on the Search page toolbar as the "Council" toggle.

1

Enable Council Mode

Navigate to the Search page and turn on the Council toggle (Users icon) in the toolbar. It stays on across sessions.
2

Submit Your Query

Type your question and submit. The system queries three different AI models in parallel.
3

Read the Consensus

You receive a synthesized consensus answer with an agreement score. Green (80%+) means high agreement, amber (50-79%) means moderate, and red (below 50%) means significant disagreement.
4

Explore Individual Responses

Expand "Individual Responses" to see what each model said, along with its response time and cost.
5

Review Divergence Points

If models disagreed, you'll see divergence points highlighting exactly where and how their answers differed.

AI Memory#

AI Memory automatically learns your preferences, domain expertise, and instructions as you use the platform. Your AI experience becomes personalized over time — no setup required. Visit AI Memory in the sidebar (Explore section) to view and manage your memories.

1

Automatic Learning

As you chat in Search and Notebooks, the AI extracts key facts about you — your domain expertise, preferred citation style, industry focus, and more.
2

View Your Memories

Click AI Memory in the sidebar to see everything the AI has learned. Search, filter by type (Entity, Fact, Preference, Instruction, Domain), and browse paginated results.
3

Manage Memories

Delete individual memories by hovering and clicking the delete button. Clear all memories to start fresh. Add manual memories for specific instructions you want the AI to always follow.
Memory TypeColorWhat It StoresExample
PreferenceBlueYour formatting and style preferences"Always use bullet points"
DomainPurpleYour areas of expertise"Specializes in FDA regulatory affairs"
EntityGreenPeople, organizations, products you work with"Works at Pfizer"
FactAmberSpecific facts the AI has learned"Reports quarterly to the board"
InstructionRedStanding instructions"Always cite APA 7th edition"

Auto-Routing#

Auto-Routing automatically selects the best AI model for each query based on the topic, complexity, and your subscription tier. Enable it with the Auto-Route toggle (purple Zap icon) in the Search toolbar. A 7-layer routing engine analyzes your question and picks the optimal model.

Query TypeRouting StrategyOptimized For
RegulatoryCompliance-specialized modelsLegal accuracy, citation precision
FinancialNumerical reasoning modelsData analysis, financial modeling
AcademicResearch synthesis modelsCitation quality, comprehensive analysis
CodeProgramming-specialized modelsCode generation, debugging, technical docs
NewsFast synthesis modelsCurrent events, real-time information
GeneralHigh-quality standard modelsBalanced quality and cost

When auto-routing is active, a small badge appears below the AI response showing which model was selected (e.g., "Auto → Claude Sonnet 4"). Hover over the badge for the routing rationale. Auto-routing is bypassed when you manually select a model.

Shared Notebooks#

Invite team members to collaborate in your research notebooks. Click the Share button in the Notebook chat toolbar to manage collaborators. Only the notebook owner can invite, remove, or change roles.

1

Open the Share Panel

In any Notebook, click the Share button (Users icon) in the chat toolbar.
2

Invite by Email

Enter the email address of the person you want to invite and choose their role: Viewer (can see and chat) or Editor (can see, chat, and edit).
3

Track Invitations

See the status of each collaborator — Accepted (green), Invited (amber), or Declined (red). Remove collaborators with the X button.

Custom Instructions#

Set persistent AI instructions for a notebook that apply to every chat message. Click the Custom Instructions button (gear icon with chevron) in the Notebook chat toolbar to open the inline editor.

1

Open the Editor

Click the Custom Instructions button in the Notebook chat toolbar. A collapsible panel opens below the toolbar.
2

Write Your Instructions

Type instructions that the AI should follow for all chats in this notebook — for example, "Always use APA citations" or "Focus on FDA 510(k) pathways."
3

Save

Click Save. Your instructions take effect on the next message you send. A success notification confirms the save.

Subscription Tiers#

Intelligence Orchestration features are available across different subscription tiers:

CapabilityFreeProMax
AI Memory
Auto-Routing
Custom Instructions
Model Council
Shared Notebooks

Features that require a paid subscription show clear messaging. The Council toggle displays "Upgrade to Pro to use Model Council" for free-tier users. The collaborator invite form is replaced with an upgrade message.

Frequently Asked Questions#

What is Model Council and why should I use it?

Model Council runs your question against three frontier AI models simultaneously. If three independent models reach the same conclusion, you can be much more confident in the answer. The agreement score (shown as a percentage) tells you exactly how aligned the models are.

How does AI Memory work? Do I need to set it up?

AI Memory works automatically. As you use Search and Notebooks, the system extracts and remembers key facts. You can also add memories manually. Visit AI Memory in the sidebar to manage your memories. There is nothing to set up.

Does Model Council cost more than a regular query?

Yes — Council queries three models instead of one, so the cost is approximately 3x a standard query. Each model's individual cost is displayed in the response view.

Can collaborators edit Custom Instructions?

No. Only the notebook owner can view and edit Custom Instructions. Collaborators benefit from the instructions when chatting but cannot modify them.

What is the difference between AI Memory and Custom Instructions?

AI Memory is global and automatic — it remembers your preferences across all conversations. Custom Instructions are notebook-specific and manual — you write them, and they only apply to one notebook.