Skip to main content

​
Understanding Assistants

Assistants are not Abbie. Assistants are bots you build and configure for a particular job. Abbie is the single AI coworker already built into QuivaWorks, with her own memory, backlog and settings.
Assistants are the intelligent core of QuivaWorks. They’re AI-powered specialists that understand context, reason through complex tasks, integrate with your systems, and continuously improve through feedback—all within boundaries you define.

​
What is an Assistant?

An assistant is an AI that can:
  • Understand — Parse natural language, context, and complex workflows
  • Reason — Think through multi-step problems with intelligent decision-making
  • Learn — Improve through feedback and user interactions automatically
  • Integrate — Connect to your tools, APIs, and systems seamlessly
  • Collaborate — Work alongside your team with clear communication and escalation
Unlike rigid automation or simple chatbots, assistants work like expert team members: you define their role, equip them with knowledge and tools, set clear boundaries—and then can continuously improve them by providing feedback on interactions.
Assistants are designed to handle complexity, variability, and judgment calls—exactly where human-like AI adds the most value.

​
Assistants vs. Automation vs. Chatbots vs. LLMs

Traditional Automation

“If this, then that”Follows exact rules. Breaks on exceptions. Requires programming every scenario.❌ Can’t handle variability
❌ Needs explicit programming
❌ Brittle with edge cases

Simple Chatbots

“Powerful text generation”Responds to keywords. Limited to chat. No real reasoning or actions.⚠️ Keyword-based only
⚠️ Can’t use tools
⚠️ Limited to conversation

Large Language Models

“Intelligent specialist”Advanced language models that generate high-quality text. Conversational but stateless—no memory or tool usage.✅ Excellent writing quality
✅ Fast responses
❌ No memory between chats
❌ Prone to hallucinations
❌ No decision-making

QuivaWorks Assistants

“Intelligent specialist”Reasons through problems. Learns from feedback. Integrates with systems. Handles complexity.✅ Intelligent reasoning
✅ System integration
✅ Feedback mechanism
✅ Works within boundaries

​
Core Assistant Capabilities

​
1. Intelligent Reasoning & Contextual Understanding

Assistants can:
  • Understand natural language and complex intent
  • Reason through multi-step problems with context awareness
  • Handle ambiguity, edge cases, and exceptions gracefully
  • Apply business logic flexibly based on situation
  • Remember conversation history and context automatically

Example: Documentation Generation

You: “Create documentation for our customer feature!”Your Documentation Assistant:
  1. Reads your description and has access to your documentation guidelines
  2. Generates clear, user-focused documentation based on your style
  3. You review and provide feedback: “Make this more beginner-friendly, add real examples”
  4. Assistant revises based on your input
  5. You vote on what worked well, and the Learning tab consolidates those votes into actionable insights
  6. You review those insights and refine this assistant’s instructions accordingly
You’re collaborating with an expert that understands your context, your systems, and your needs—not starting from scratch each time. You explicitly capture what works and apply it intentionally.

​
2. System Integration & Tool Access

Assistants integrate with:
  • Knowledge bases and documentation
  • CRM and customer data systems
  • Databases and data sources
  • APIs and external services
  • GitHub, email, and communication platforms
  • Custom business systems
Assistants decide which tools to use based on the task at hand, you just need to provide authorisation and access to the integration and the assistant will do the rest. You don’t program “if customer asks about orders, call Order API”—the assistant figures that out intelligently.
Integrations are configured once during setup, then your assistant automatically uses the right system at the right time.

​
3. Smart Configuration & Control

Assistants work within boundaries you define:
  • Instructions — Define role, personality, and specific capabilities
  • Knowledge — Provide training data, guidelines, and context
  • Output validation — Enforce required response formats
  • Integrations — Control which systems the assistant can access
  • Context limits — Set token budgets and memory size
  • Execution modes — Synchronous or background processing
Boundaries aren’t limitations—they’re guardrails that ensure your assistant works exactly as you need.

​
4. Continuous Learning & Self-Improvement

Assistants improve through:
  • Smart Context Management — Intelligently handles conversation memory without manual tuning
  • Learning System — Accumulates feedback from interactions and surfaces actionable insights
  • Performance Insights — Consolidates patterns from successful interactions so you can identify what works well
  • Intentional Refinement — You review insights and refine your assistant’s instructions and behaviour based on what you’ve learned
The Learning tab shows consolidated feedback, allowing you to make informed decisions about how to improve your assistant’s performance over time.

​
When to Use Assistants

​
Perfect For

  • Answer questions with full context and history
  • Troubleshoot issues using knowledge base
  • Apply policies with judgment (returns, refunds, exceptions)
  • Handle complex, multi-turn conversations
  • Escalate to humans when needed
  • Learn from each interaction to improve
  • Ask discovery questions dynamically
  • Research companies and prospects
  • Score leads based on ICP criteria
  • Enrich data from multiple sources
  • Personalise outreach at scale
  • Route qualified opportunities intelligently
  • Create personalised campaigns
  • Generate and adapt messaging by audience
  • Maintain brand voice across channels
  • Repurpose content efficiently
  • A/B test messaging variants
  • Extract information from documents
  • Validate data against business rules
  • Make contextual decisions on exceptions
  • Cross-reference multiple systems
  • Flag issues for human review
  • Learn from manual corrections
  • Triage and manage issues
  • Automate routine technical tasks
  • Generate documentation and reports
  • Coordinate between teams
  • Learn best practices from interactions

​
Not Ideal For

When to Use Automation Instead

  • Simple, predictable tasks — If it’s always the same steps, use a condition or rule
  • High-volume, low-variability — Assistants have per-run costs; save them for complexity
  • Pure data transformation — Use Rules or Functions for straightforward data work
  • Time-critical micro-operations — Assistants add latency; use functions for speed
  • Deterministic calculations — Use Rules for exact math and logic
Rule of thumb: If you can write “if X then Y” rules that cover every case, use automation. If there’s judgment, context, learning, or exceptions, use an assistant.

​
Building Your First Assistant

1

Define the Role

What is this assistant responsible for? Customer service? Sales support? Technical operations? Be specific about the domain and focus.
2

Write Instructions

Describe the assistant’s role, personality, communication style, and specific capabilities. This is like writing a detailed job description.
3

Add Knowledge

Provide training data, guidelines, policies, and context. Upload documents, paste knowledge, or link URLs. This is how your assistant learns your business.
4

Choose a Model

All assistants default to a fast Claude model, included in every plan. Accounts with bring-your-own-key enabled can connect their own API key to use a different model.
5

Connect Integrations

Add the systems your assistant needs: GitHub, CRM, knowledge bases, APIs. Start with essentials, add more as needed.
6

Set Boundaries

Define what the assistant can and cannot do. Set output schemas, validation rules, escalation criteria.
7

Test & Learn

Use the Chat interface to test with real scenarios. Review the Learning tab for insights on what’s working and what needs refinement.
8

Deploy & Monitor

Activate in flows, review performance regularly, and refine based on feedback and learning insights.

Create Your First Assistant

Follow our step-by-step guide to deploy your first assistant in minutes

​
Configuration & Smart Features

Every assistant has intelligent configuration areas:

​
Instructions Configuration

Define your assistant’s identity and behaviour:
  • Role & Personality — Who is this assistant? What’s their expertise?
  • Specific Capabilities — What tasks should they handle?
  • Communication Style — How should they respond? Formal? Friendly?
  • Limitations — What should they NOT do? When should they escalate?
Instructions are like a job description—be specific and clear about expectations. Learn more about Instructions →

​
Knowledge Management

Train your assistant with relevant information:
  • Manual Entry — Paste guidelines, policies, or context directly
  • File Upload — Import documents, PDFs, or training materials
  • URL Import — Link to web pages, documentation, or knowledge bases
  • Knowledge Relevance — The assistant automatically uses relevant knowledge when needed
Knowledge is what makes your assistant domain-expert smart. Learn more about Knowledge →

​
Integration Configuration

Connect your business systems:
  • GitHub Issues & PRs — Manage repositories and workflows
  • CRM Systems — Access customer data and history
  • APIs & Webhooks — Connect custom systems
  • Knowledge Bases — Link documentation and guidelines
  • Email & Communication — Send messages and updates
Your assistant intelligently decides which system to access for each task. Learn more about Integrations →

​
Context Variables

A free-form set of key/value pairs, configured on the Instructions tab, that gets injected verbatim into the assistant’s system prompt — useful for things like a project name, a default region, or other values you want the assistant to always have on hand.
Values are not sanitised and can enable prompt injection — use Context Variables only with trusted input. They are not a place to store API keys or credentials; those belong in Secrets or an integration’s own auth configuration.
Learn more about Context Variables →

​
Provider Settings

Choose the AI model:
  • Default Model — a fast Claude model, included in all plans
  • Bring Your Own Keys — connect your own Anthropic API key on accounts with bring-your-own-key enabled
  • Output Token Limits — Set maximum response size
  • Configuration — Fine-tune behaviour settings
Different models excel at different tasks. Test to find what works best for your use case. Learn more about Provider Settings →

​
Learning System

Enable continuous improvement:
  • Feedback Accumulation — Collect insights from interactions
  • Performance Insights — See what’s working and what needs improvement
  • Consolidate Learnings — Turn feedback into actionable guidance
  • Intentional Refinement — Review insights and decide when and how to update your assistant
The Learning tab shows you exactly how your assistant is performing and where to focus refinement efforts. Learn more about Learning →

​
Assistant Costs

Assistants are billed through a credit system based on your plan: How Credits Work:
  • Each interaction consumes credits based on model usage and complexity
  • The included allowance refreshes monthly and is shared across the whole account, not allocated per user
  • On Pro and Team, purchase additional credits in packs—they never expire and roll over
  • Different models draw down credits at different rates, so cost varies by which model an assistant uses
Credit Allocation by Plan:
  • Free: 1,500 credits per account per month; top-up packs aren’t available on Free
  • Pro: 20,000 credits per account per month, plus optional top-up packs
  • Team: 90,000 credits per account per month, plus optional top-up packs
Pro and Team share the same top-up ladder, priced per pack rather than a flat per-1,000 rate—larger packs cost less per credit. Default Model: All new assistants use a fast Claude model by default—cost-effective and optimised for most business workflows. Cost Optimisation Tips:
  • Use the default model for most tasks—a good balance of performance and cost
  • Start with shorter conversations and context to reduce credit consumption
  • Consolidate feedback through the Learning system to refine instructions and reduce re-work
  • Monitor assistant performance in the analytics dashboard to identify cost-saving opportunities
See Plans & Pricing for detailed billing information and plan comparisons.

​
Next Steps

Create Your First Assistant

Step-by-step guide to building and deploying an assistant

Instructions Configuration

Define role, personality, and capabilities

Knowledge Management

Train your assistant with domain knowledge

Integrations

Connect GitHub, CRM, APIs, and more

Context Settings

Tune memory, context limits and output schema

Provider Settings

Choose AI models and fine-tune behaviour

Learning System

Enable continuous improvement through feedback

Use in Flows

Deploy assistants in your workflows
Ready to deploy your first intelligent assistant? Start simple—define a clear role, add essential knowledge and integrations, then watch your assistant improve with every interaction. The Learning system will guide your refinements.

​
Get Help

Community

Share assistants and learn from others

Support

Get help from our team

Marketplace

Browse assistant templates and examples