How should we think about AI cost and return?
Model fees are the tip; design, integration, and control are the real investment.
Cost buckets usually include discovery, integrations (WhatsApp, CRM, site), model/API usage, knowledge prep, testing, monitoring, and ongoing care. “We opened ChatGPT” is not a production system.
Return is easiest to discuss with numbers: hours saved × labour cost, value of recovered leads, faster response, fewer errors. “It would be cool” is not a business case. A pilot with a kill criterion is healthier than an endless vague subscription story.
Risk has a cost too—wrong answers, privacy issues, brand tone failures. Logging, approvals for sensitive actions, and staged rollout belong in the budget.
We outline options for your volume and channels in a proposal rather than promising a universal ROI number here.
