It quotes ranges based on the months behind and transaction volume the prospect describes - say, '4 months behind, low transaction volume' yields a different number than '12 months behind, high volume'. The actual scope is confirmed by a human after a quick QBO peek because catch-up always has surprises (uncategorised transactions, missing statements, opening balance issues), but the quote band the bot gives is anchored in your fee structure and is rarely off by much.

Yes. Add stack pages or use ACF fields to list which platforms you actively support (QBO, Xero, Bill.com, Gusto, Ramp, Brex) and which you do not. The bot uses that context to recommend the right onboarding flow, quote correctly, and decline politely when a prospect is on a stack you do not handle. That single behaviour saves the most painful conversation a bookkeeper can have.

Not directly. The chatbot logs every conversation in WordPress and you push the summary into Karbon, Anchor, Keeper, or your practice management tool via webhook or Zapier. The bot is the front-of-house intake layer; the practice management tool is back-of-house and stays where it is. Most firms map the bot's intake fields to their PM new-client fields once and the integration runs cleanly from there.

Yes. As long as those services live as pages on your site - what is included, what they cost, who they are for - the bot explains them and routes the lead to the right team member. Many firms structure their site as bookkeeping/CAS/fractional-CFO and use display conditions so each service page has its own slightly-different bot, which keeps the conversations focused on the offering the visitor is actually looking at.

It declines to give tax advice and redirects gracefully. The system prompt is configured to say something like 'we hand finished books to your CPA at year-end and can introduce you to the two CPAs we work with' rather than improvising a tax answer. That keeps the engagement scope clean and the prospect's expectations aligned with what bookkeeping firms actually do, which is the basis for a healthy ongoing relationship.

GPT-4o and Claude Sonnet both handle the technical accounting language well - reconciliations, accruals, AR/AP, classifications, retained earnings questions. Smaller or older models tend to fumble the bookkeeping-specific terminology in ways that erode prospect trust quickly. Bring your own key for whichever provider you prefer; switching the model on a bot is a settings change rather than a re-build.

Roughly, yes. The bot asks about monthly revenue, number of bank/card accounts, and a rough sense of how active the business is, and uses that to land in a tier band. The actual transaction count comes from peeking at the QBO file during the discovery call, but the chatbot's estimate is usually within a tier of the truth, which is enough to set expectations on price and avoid the worst pricing surprises later.

Yes. The bot recognises signals like 'three S-corps under one parent' or 'we operate in CAD and USD' and either quotes from your multi-entity tier (if you have one) or flags it as a complexity case that needs senior scoping. Multi-entity bookkeeping is a different shape of work than single-entity, and surfacing the complexity upfront protects both the prospect's expectations and the firm's pricing.