onWebChat

Website chatbots have moved from a nice-to-have widget to a measurable line item in agency retainers, because buyers now expect 24/7 answers before they ever fill out a form. The strategic stake is not the chat tool itself but the recurring analytics, content training, and escalation work that keeps it accurate, which is where agencies earn margin. Agencies that sell chat as a one-time install watch it decay in 90 days; those that sell it as a monitored retainer compound client value.

Widget Exit Cost is the framework for pricing a chatbot engagement by how much of the client's conversational asset survives a vendor change. Three layers accumulate: the widget embed, the trained knowledge base, and the conversation history with lead records. A full white-label platform such as SiteGPT or Havlo lets an agency keep the client-facing brand and, in Havlo's case, run several client sites from one dashboard, which raises the agency's switching cost rather than the client's. A no-code builder like Typebot gives the agency portable flows but no resale layer, so the retainer stays tied to build hours. Before quoting, map which layer the client owns and which the vendor holds. If transcripts and lead data cannot be exported in a usable format, the agency is renting a client relationship it cannot move, and every renewal conversation starts from a weaker position.

Answer Ownership Split is the practice of deciding, question by question, whether the chatbot or a human owns the reply, then wiring the handoff before launch. Agencies that skip this step ship widgets that either over-answer (inventing policy on pricing, timelines, or refunds) or under-answer (bouncing every visitor to a form). The split matters because retainer renewals hinge on deflection rate, and deflection without a clean escalation path just moves the queue into a client's inbox. A workable split assigns FAQ, hours, and location questions to the bot, and routes anything with money, legal exposure, or an angry tone to a named human within a stated response window. Brevn's approach is instructive: it crawls site pages to ground answers, captures contact details when buying intent appears, and emails the business so a person can follow up. That is ownership split in production, not a settings toggle.

Escalation Debt Ratio is the share of chatbot conversations that must hand off to a human, and it is the number that decides whether a website chatbot is a margin asset or a hidden retainer cost. A widget that deflects 70% of inquiries looks strong in a demo, but if the remaining 30% arrive without context, an agency absorbs the re-qualification time inside a fixed monthly fee. The ratio matters because it moves in both directions: better grounding lowers it, while new client offerings raise it. Havlo and SiteGPT both push full conversation context to the human agent at handoff, which keeps the debt from compounding. Track the ratio monthly per client, price the human tier against it, and treat any climb above the contracted threshold as a scope conversation rather than an absorbed cost.

Before quoting a chatbot upsell, confirm the client can export conversation data and re-point the widget without rebuilding the knowledge base, and price the migration path into the proposal.

Audit and consolidate the client's answerable content before installing any widget, because a chatbot trained on stale or contradictory pages will confidently misquote the client to their own customers.

IF you already hold web or retainer clients and want chatbot revenue that survives a client leaving, THEN resell a white-label platform you control and bill it as a monthly line item. IF chatbot work is one-off, scope is unclear, or the client's own support team will own the widget, THEN install a single lightweight tool per site and hand over the keys.

The choice hinges less on feature counts than on who owns the client relationship after launch: white-label platforms let an agency bill chat as its own line item, while no-code builders trade that branding for faster iteration on complex flows. Pick the routing-first option only when the client's real need is a faster inbox, not automation. Whichever you choose, confirm that conversation analytics and integrations can be exported, because the lock-in risk in this category sits in the reporting layer, not the widget.

A productized engagement that installs a rebranded AI chat widget on a client site, trains it on their content, and hands over a resale-ready support and lead capture layer the agency can bill monthly.