Ranges for automation that saves real work: WhatsApp assistants, internal copilots, document flows, and guarded LLM integrations tied to your data and ops tools.
AI project cost is driven less by model hype and more by guardrails, observability, and what happens when the model is wrong. We price for production behaviour: escalation paths, logging, and human oversight where it matters.
What moves the price
Use these drivers when you compare quotes. Two proposals with the same page count can differ sharply once integrations, roles, and post-launch load are included.
Channel and volume
WhatsApp, web chat, email, or internal tools, plus peak message and token load.
Knowledge and data access
RAG (Retrieval-Augmented Generation), document stores, and permissions on sensitive data.
Tooling and actions
Whether the agent only answers questions or also updates CRM, tickets, or orders.
Evaluation and safety
Test sets, refusal rules, PII handling, and review workflows before wide rollout.
Hosting and model costs
Inference spend, caching, and South African latency and connectivity constraints.
Human-in-the-loop
Admin review queues, handoff to staff, and audit logs for regulated or high-trust use cases.
Typical AI automation engagements
Project type
Typical range (ZAR)
Timeline
Pilot assistant
One channel, bounded knowledge, manual review, limited actions.
R80k – R200k
4–8 weeks
Production WhatsApp or ops bot
Live channel, CRM or ticket hooks, monitoring, admin oversight.
R200k – R500k
2–4 months
Multi-workflow AI platform
Several agents, document pipelines, analytics, strict reliability targets.
R500k – R1.2m+
4–8 months
Ranges are planning guides for South African delivery in 2026. Final quotes follow discovery. Third-party fees (stores, APIs, messaging) are excluded unless stated in scope.
Retainer options after launch
Models, APIs, and business rules change. AI systems need ongoing evaluation, prompt and workflow updates, and incident response like any production service.
Send your brief on WhatsApp or run the estimate tool for a ballpark on greenfield builds. Retainer fit is agreed after we review your live stack and roadmap.