AI Assistants
ScendCore includes 6 AI-powered workspace assistants that help sales reps with research, email drafting, meeting prep, prospecting, analytics, and coaching. Each assistant runs an agentic loop — a multi-turn cycle where the AI reasons, calls tools, reads results, and continues until it has a complete answer.
Accessing Assistants
Navigate to Workspace in the sidebar. You will see your assigned AI agent and a grid of available assistant tools.
The 6 Assistants
| Assistant | What It Does | Key Tools |
|---|---|---|
| Research | Looks up contacts, companies, and opportunities | Contact lookup, profile enrichment, account search |
| Meeting Prep | Prepares briefings before meetings | Contact history, conversation review, opportunity context |
| Draft Email | Drafts and sends emails on your behalf | Contact lookup, conversation history, email sending |
| Prospecting | Finds and qualifies leads | Contact search, profile analysis, account research |
| Analytics | Analyzes pipeline and performance data | Pipeline stats, activity metrics, conversion rates |
| Coaching | Reviews conversations and gives coaching tips | Conversation history, contact context, deal analysis |
Using an Assistant
- Click on an assistant card in the Workspace.
- The assistant panel opens with a chat interface.
- Type your request — for example:
- “Research Acme Corp and their recent funding rounds”
- “Draft a follow-up email to Sarah at TechCo about our demo”
- “Prep me for my meeting with John Smith tomorrow”
- The assistant streams its response in real time.
- If the assistant needs data, it automatically calls tools (database lookups, web searches, etc.) and shows the results.
How the Agentic Loop Works
Each assistant follows this process:
- Receives your message
- Decides which tools to call (if any)
- Executes the tools and reads the results
- Continues reasoning with the new data
- Repeats up to 5 iterations if needed
- Streams the final response back to you
Tool calls are visible in the chat so you can see exactly what data the assistant accessed.
LLM Provider Configuration
ScendCore supports multiple LLM providers:
| Provider | Default Model | Tier |
|---|---|---|
| Anthropic | Claude Sonnet | Premium |
| OpenAI | GPT-4o | Premium |
| Google Gemini | Gemini 2.5 Flash | Budget |
| Groq | Llama 3.3 70B | Budget |
Your platform admin can configure which provider each assistant uses. Tenants can also bring their own API keys (BYOK) if enabled.
To configure assistant LLM settings, go to Settings > AI Assistants.
Skills
Assistants can be extended with additional Skills like web search, LinkedIn lookup, document generation, and more. Skills are feature-flagged and can be enabled or disabled per tenant.
Streaming
Responses are streamed in real time using Server-Sent Events (SSE). You will see text appearing word by word, along with tool call results and structured cards as the assistant works through your request.
Tips for Better Results
- Be specific — “Draft an email to Sarah about the pricing discussion from our last call” works better than “write an email”
- Provide context — Mention the contact name, company, or deal so the assistant can look up relevant data
- Use the right assistant — Each assistant is optimized for its domain; use Research for lookups and Draft Email for writing
- Follow up — You can ask follow-up questions in the same conversation to refine results
Troubleshooting
Assistant not responding
- Check your internet connection — responses are streamed via SSE.
- Try refreshing the page if the connection drops.
Tools showing errors
- Some tools require external API keys (e.g., web search needs a Tavily API key).
- Check Settings > AI Assistants for provider configuration.
Slow responses
- Premium providers (Anthropic, OpenAI) typically respond faster than budget options.
- Complex requests with multiple tool calls take longer as each tool must complete before the next iteration.