ClientPoint's New DSR Feature: What is Meg AI Search?

Meg AI is an intelligent, conversational AI assistant embedded directly inside your ClientPoint Digital Sales Rooms. It allows your buyers, prospects, and internal team members to ask natural-language questions about the content in a proposal and get instant, accurate answers with citations pointing back to the exact source files and pages.

Instead of scrolling through dozens of pages across multiple documents to find a specific detail, recipients can simply ask Meg. The AI reads and understands every document in the DSR, then responds with precise, referenced answers.

Meg AI is powered by advanced Retrieval-Augmented Generation (RAG) technology, combining semantic understanding with keyword search to deliver the most relevant answers from your proposal content and custom knowledge bases.

Key Capabilities at a Glance 

Getting Started

Prerequisites

Before using Meg AI, ensure the following are enabled for your company:

1. Proposal Search must be active (this powers the content indexing pipeline)
2. Meg AI feature flag must be enabled by your ClientPoint administrator

Contact your ClientPoint account manager or support team to enable these features if they are not already active.

For Administrators: Setting Up Meg AI

Step 1 - Create Your First AI Agent

An "agent" is a configured AI persona that you add to your proposals. Each agent can have its own name, personality, greeting, and dedicated knowledge base.

1. Navigate to Administration > AI Agents
2. Click "Create New Agent"
3. Fill in the agent configuration:

4. Click Save

Tip: System Prompt Best Practices:

• Be specific about the agent's role: "You are a technical pre-sales engineer who explains complex features in simple terms."
• Define the tone: "Be conversational but professional. Avoid jargon unless the user uses it first."
• Set boundaries: "If asked about competitor comparisons, focus on our strengths rather than criticizing competitors."
• Add domain context: "Our company specializes in financial services software. Prioritize compliance and security topics."

Step 2 - Build the Agent's Knowledge Base (Optional)

The Knowledge Base lets you upload supplementary PDF documents that give your agent extra context beyond what's in the proposal itself. This is useful for:

• Product documentation and spec sheets
• Pricing guides and rate cards
• Case studies and testimonials
• FAQ documents
• Compliance and certification details
• Training materials

To upload Knowledge Base files:

1. In the AI Agents admin page, select your agent
2. In the Knowledge Base section, click "Upload File"
3. Select a PDF file (maximum 25 MB per file)
4. The file will show a "Pending" status badge while it's being processed
5. Processing involves extracting text, generating AI embeddings, and indexing the content
6. Once complete, the badge changes to "Completed" with a chunk count

Processing Status Guide:

To remove a Knowledge Base file:

• Click the delete icon next to the file. This removes it from both storage and the search index.

Important: Knowledge Base content is shared across all proposals where this agent is added. If you need proposal-specific context, the agent automatically uses the files within each individual proposal.

Step 3 - Create Multiple Agents for Different Use Cases

You can create as many agents as needed. Common configurations include:

For Proposal Builders: Adding Agents to Proposals

Adding an Agent to a Proposal

1. Open a proposal in Build mode
2. In the file toolbar, click the "Add Agent" button (or use the dropdown menu)
3. An overlay dialog appears showing all active agents for your company
4. Select the agent you want to add
5. The agent immediately appears in the proposal's file tree with a distinctive agent icon — no page refresh required

Key points:

• You can add multiple agents to a single proposal
• Each agent appears as a "file" in the proposal tree, just like PDFs or other documents
• The agent can be reordered, renamed, or removed like any other file
• Agents are visible to recipients in the viewer's Table of Contents

Controlling Which Files the AI Can Access

Not every file in a proposal should be referenced by the AI. You have granular control:

AI Searchable Toggle:

1. In Build mode, click the three-dot menu (⋯) on any file
2. Click "AI Searchable" to toggle the setting
3. When enabled (default): The AI can read and cite this file
4. When disabled: The AI ignores this file entirely

Internal Files and AI Behavior:

• When you mark a file as "Internal", it is automatically excluded from AI citations
• However, internal files still provide background context - the AI uses the information to give better answers, but will never reference or cite the internal file by name to external viewers
• This is useful for internal pricing sheets, competitive analysis, or deal strategy documents that inform the AI's responses without exposing them

Automatic Sync Rules:

• Setting a file to Internal → AI Searchable automatically turns OFF
• Removing Internal status → AI Searchable automatically turns back ON

For DSR Recipients: Using Meg AI in a Proposal

Starting a Conversation

1. Open the Digital Sales Room (DSR) link shared with you
2. In the Table of Contents sidebar, look for the agent entry (marked with a distinctive AI icon)
3. Click on the agent name
4. The document viewer is replaced by the Meg AI chat interface
5. You'll see a welcome message and suggested questions to get started

Asking Questions

• Type freely: Ask anything in natural language - "What are the pricing tiers?", "How does the implementation timeline work?", "What security certifications do you have?"
• Use suggested questions: Click any of the pre-configured question chips for quick answers
• Follow up: Meg remembers your conversation context, so you can ask follow-up questions like "Can you elaborate on that?" or "What about for the enterprise tier?"
• Be specific: The more specific your question, the more precise the answer. "What is the SLA for tier 2 support?" will give a better answer than "Tell me about support."

Understanding Responses

Citations: Every response includes source citations formatted as clickable cards showing the file name and page number. For example:

Clicking a citation takes you directly to that file and page in the document viewer, so you can verify the information in its original context.

What Meg AI can do:

• Answer questions about any content in the proposal documents
• Reference and cite specific files and pages
• Summarize lengthy documents
• Compare information across multiple files
• Explain technical concepts in simpler terms

Navigating Between Chat and Documents

• Click any file in the Table of Contents to switch from the chat back to the document viewer
• Click the agent in the Table of Contents to return to the chat - your conversation is preserved
• Click a citation card in a chat response to jump directly to the referenced document and page
• Your conversation history is saved and available when you return, even across browser sessions (for up to 90 days)

Feature Deep Dive

Hybrid Search Technology

Meg AI uses a dual-search approach to find the most relevant content:

1. Semantic Search (k-NN): Understands the meaning behind your question. If you ask "What is the cost?", it also finds content about "pricing", "fees", "rates", and "investment" - even if those exact words don't appear in your question.
2. Keyword Search (BM25): Matches specific terms, names, numbers, and technical terminology exactly. Ensures precise results when you use specific product names, part numbers, or technical terms.

Results from both approaches are combined, deduplicated, and ranked to deliver the most relevant answer.

Conversation Memory

• Each chat session maintains up to 10 messages of context, allowing the AI to understand follow-up questions and maintain coherent multi-turn conversations
• Conversations are stored securely and automatically expire after 90 days
• Each proposal + agent combination maintains its own independent conversation thread
• Conversation state persists across browser sessions using secure local storage

Rate Limits

To ensure fair usage and system stability:

If you hit a rate limit, wait briefly before sending another message. Session limits reset when a new session begins.

Internal File Intelligence

One of Meg AI's most powerful features is its handling of internal files:

• Files marked as Internal in the Build screen are hidden from external recipients in the document viewer
• However, their content is still fed to the AI as background context
• The AI is instructed to never cite or reference internal files by name
• This means the AI can give more informed, accurate answers informed by your internal data - without exposing sensitive materials

Example use case: You have an internal competitive analysis document. A recipient asks "How does your solution compare to Competitor X?" The AI can draw on your competitive positioning from the internal document to craft a strong response, without revealing that the internal document exists.

Theming and Branding

The Meg AI chat interface automatically inherits your DSR's visual theme:

• Header background color matches your DSR header
• Text colors adapt for readability
• Primary action colors (send button, links) align with your brand palette
• No additional configuration is needed - theming is automatic

Best Practices for Maximum Impact

1. Craft Effective System Prompts

The system prompt is the single most impactful configuration for your agent's behavior. Invest time in writing a detailed, specific prompt.

Template for a strong system prompt:

You are [Agent Name], a [role description] for [Company Name].
 

PERSONALITY:

- [Tone guidance - e.g., Professional but approachable]

- [Communication style - e.g., Concise, use bullet points for lists]

EXPERTISE:

- [Domain focus - e.g., Enterprise software, financial services]

- [Key topics to emphasize - e.g., ROI, security, scalability]

GUIDELINES:

- [Specific behavior rules - e.g., Always mention our 99.9% uptime SLA when discussing reliability]

- [Topics to handle carefully - e.g., Redirect pricing questions to the included rate card]

- [Things to avoid - e.g., Don't make promises about custom development timelines]

2. Curate Your Knowledge Base

• Upload high-quality, well-structured PDFs - clear headings and organized content produce better search results
• Keep knowledge base files focused and up to date  remove outdated materials
• Use the knowledge base for persistent information that applies across proposals (product docs, case studies)
• Let proposal-specific content come from the files within each proposal itself

3. Write Good Suggested Questions

Suggested questions serve as conversation starters and set user expectations. Make them:

• Specific: "What is the expected ROI in year one?" not "Tell me about ROI"
• Relevant: Align with what recipients commonly ask about
• Varied: Cover different aspects - pricing, implementation, support, features
• Action-oriented: "How do I get started with onboarding?" drives engagement

4. Use AI Searchable Strategically

• Disable AI Searchable for cover pages, table of contents, and boilerplate legal pages that add noise without value
• Enable AI Searchable for substantive content - proposals, spec sheets, case studies, pricing
• Mark supplementary context as Internal when you want the AI to know the information but not expose the source

5. Use Multiple Agents for Complex Proposals

For large, multi-stakeholder deals, consider adding multiple agents:

• A General Assistant for business decision-makers
• A Technical Expert for IT and engineering evaluators
• A Procurement Guide for legal and procurement teams

Each agent can have its own personality, instructions, and knowledge base tailored to its audience.

Frequently Asked Questions

Q: Can recipients see my system prompt or knowledge base files? A: No. The system prompt and knowledge base files are never exposed to recipients. They only see the AI's responses.

Q: What file types can the AI read? A: Meg AI can read and index PDF, DOCX, PPTX, and HTML files added to proposals. Knowledge base uploads are limited to PDF format.

Q: Can I use Meg AI with password-protected or PIN-secured DSRs? A: Yes. Recipients who have been verified via PIN or external authorization can use Meg AI. The authentication is handled seamlessly.

Q: How quickly are new files indexed? A: Proposal files are indexed asynchronously when added to a proposal. Most files are searchable within a few minutes. Knowledge base PDFs follow the same timeline - watch the status badge for confirmation.

Q: Is conversation data secure? A: Yes. All conversations are encrypted in transit and at rest. Chat history is stored in isolated, per-proposal sessions and automatically expires after 90 days. No conversation data is used to train AI models.

Q: What happens if the AI doesn't know the answer? A: Meg AI is designed to be honest. If the answer isn't in the provided documents, it will say so rather than guessing or fabricating information.

Q: Can I delete or reset a conversation? A: Conversations are tied to browser sessions. Starting a new session (clearing browser data or using a new device) begins a fresh conversation. Administrators cannot view or modify individual conversations.

Q: Does Meg AI work on mobile devices? A: Yes. The chat interface is responsive and works on mobile browsers when accessing a DSR link.

Q: Can multiple people use the same agent at the same time? A: Yes. Each user or guest gets their own independent conversation thread. Multiple people can interact with the same agent simultaneously without seeing each other's conversations.

Q: Is there a limit to how many agents I can create? A: There is no hard limit on the number of agents. Create as many as your team needs for different use cases and audiences.

Troubleshooting

Summary

Meg AI transforms your Digital Sales Rooms from static document repositories into interactive, intelligent experiences. By adding AI agents to your proposals, you give every recipient a personal guide who can instantly answer questions, surface relevant information, and navigate them to the right content, all while respecting your access controls and branding.

To get started:

1. Ask your administrator to enable Meg AI for your company
2. Create your first agent in Administration > AI Agents
3. Upload knowledge base documents for extra context
4. Add the agent to a proposal in Build mode
5. Share the DSR - recipients can start chatting immediately

For additional support, contact your ClientPoint account manager or visit the ClientPoint Help Center.