Starter Bot
$150-300
A clean support assistant for smaller sites and first deployments.
- FAQ grounding
- Brand voice intake
- Visible human handoff
Services
The client offer is simple on purpose.
Buyers should understand the scope quickly: one clean chatbot path, one stronger business layer, and one deeper agent option for teams that really need it.
Business Bot
$500-900
A stronger support or lead-flow assistant with retrieval and routing logic.
- Custom RAG when justified
- Escalation rules
- Model and cost guidance
Custom AI Agent
$1500-3000
A deeper system for businesses that need tools, structured actions, and auditability.
- Custom integrations
- Reviewable fallback logic
- Operational workflows
Model and scope guidance
Choose the cleanest system that solves the real support problem.
The site stays away from prestige-model theater and keeps the decision logic grounded.
Default model path
Most SMB deployments should start with a balanced, cost-aware model choice instead of a prestige model.
Retrieval only when earned
Heavier RAG is quoted separately when the knowledge problem is actually messy enough to need it.
Clients own usage
API billing belongs to the client by default so there is no hidden pass-through complexity later.
Looking for something else?
Wrong page? Two doors over.
Deep research
The reasoning behind the bots, in writing.
These are the internal reports Fabi AI Labs uses when scoping a build. Shared here so clients can see the thinking without having to ask for it.
Research
AI model selection guide
How Fabi AI Labs picks between OpenAI, Anthropic, and Google models based on volume, tone, and cost. Pricing snapshot, Arena ELO, TTFT, honest tradeoffs.
Read guide →Process
Delivery process
From signed order to live chatbot in seven focused days. Day-by-day timeline, review checkpoints, and the full delivery package.
See the timeline →Principles
Safeguards playbook
The four guarantees baked into every bot — human handoff, uncertainty gate, no gaslighting, full conversation transparency.
Read the playbook →Architecture
Multi-stage RAG architecture
The reference architecture behind the precision RAG tier: query decomposition, retrieval fan-out, reranking, and the compaction path that keeps responses grounded.
Open PDF →Methodology
Retrieval benchmark methodology
How retrieval quality is measured before a bot ships. The actual test harness, the failure categories, the bar a build has to clear.
Open PDF →
