The system prompt encodes the criteria you actually use in sales calls today: company size, vertical, current stack, budget range, evaluation timeline, and any deal-blocker you regularly hit. The bot asks each one conversationally and assigns a score per the rules you defined. Only leads above your threshold post to the CRM as opportunities; below-threshold leads route to a nurture flow or a self-serve onboarding link.

Via standard webhook or, for Salesforce, the existing WordPress connector if you have one. The bot POSTs the qualification fields and the chat transcript to the CRM's inbound API, including a routing hint for territory or vertical. There's no required custom integration; CRMs that accept structured webhook payloads cover the standard cases, which includes the three you named.

Two paths: route to a nurture inbox with the transcript so marketing can decide whether to drip, or route to a self-serve onboarding page with a clear scope statement ("if you're under 5 seats, our self-serve plan is the right fit"). Nothing is dropped. The point of qualification is matching the lead to the right path, not throwing the unqualified ones away.

Only if you write the prompt badly. The right pattern is to answer two or three of the visitor's questions first and only qualify when the visitor signals real intent ("can someone show me how it works", "can I get a quote"). Asking for company size before the visitor has expressed any interest is the form-style mistake; the bot's job is to qualify in the natural rhythm of the conversation, not interrogate.

Yes. Display conditions per URL pattern, UTM, or referrer plus Multibot mean a paid-search landing page can run a tighter qualifier than the home page, and a vertical-specific campaign can ground against case studies from that vertical. One install, multiple tuned qualifiers, each routing to the right CRM segment or sales pod.

Conversation logs live in your WordPress database under your retention policy. The model provider call goes to whichever provider you configured (OpenAI, Anthropic, Google, OpenRouter) so the residency question maps to that provider's terms; this is the same posture as any chatbot using BYO key. The bot can require explicit consent before email capture and route a privacy-policy link automatically when contact data is collected.

Drift and Qualified are full SDR-assistance suites with their own pricing per seat and their own contact database. SleekAI is a WordPress plugin: flat license, your model provider key, your existing WordPress content as the grounding source. It handles the qualification conversation and the structured handoff; the CRM, inbox, and pipeline reporting stay where you already keep them. If your stack centers on WordPress, the operational shape is cleaner.

Two levers move: SDR hours saved on disqualified leads, and faster response time on qualified ones. A medium-velocity inbound funnel typically sees 20 to 40 percent of leads disqualified on the first call. Removing those calls is real time. The qualified leads that remain reach the SDR faster, with structured context attached, which improves the first-call show rate. Compared to model cost (well under a cent per conversation on a mid-tier model), the ROI is heavily SDR-time-dominated.