01Where a chatbot earns its keep
The honest case for a chatbot is repetition, not novelty. A salon answering the same booking and price questions forty times a week, a restaurant confirming the same opening hours every night, a clinic fielding the same three intake questions before every appointment: these are exactly the situations where a scripted, tested answer saves real hours without losing anything a human would have added. The bar is not 'can it sound smart', it is 'does the same question arrive often enough to justify building an answer for it once instead of typing it out every time'.
- Same question, high frequency, low judgment: the ideal chatbot case
- A booking or price question does not need a human's discretion
- The saved time is measured in repeated typing, not in cleverness
02Where it quietly makes things worse
A chatbot dropped onto a business where every case is different does not remove work, it adds a layer a frustrated customer has to get through before reaching a person. If the product line changes weekly, if pricing depends on a site visit, or if the wrong answer could mean a missed allergy or a legal claim, a scripted bot will eventually give a confident wrong answer, and confident wrong answers cost more trust than a slow honest one. This is the case PROOF48 turns down rather than sell into, because a bot that damages trust is worse than no bot at all.
- Every-case-is-different businesses are the worst fit for a script
- A confident wrong answer costs more trust than a slow right one
- PROOF48 turns down this build rather than sell into it
A five-question quiz instead of an open chatbot
Korean Cosmetics did not get an open-ended AI chatbot for its skin quiz, it got a five-question flow that matches a visitor to the right product line, because the answer space was narrow enough to script completely and regulated enough that an open chatbot improvising about ingredients was the wrong tool for the job. The lesson generalises: sometimes the honest answer to 'do we need a chatbot' is a simpler, fully scripted flow that never has to admit it does not know something.
See the build: Korean CosmeticsQuestions, answered
What is the actual cost of getting the chatbot wrong?
Not the chatbot's price, the trust it burns. A customer who gets a confidently wrong answer once tends to stop asking the bot anything, which pushes them back to the phone call the bot was supposed to remove.
Does PROOF48 build chatbots for every AUTOMATE client?
No. AUTOMATE includes a website chatbot, but we only turn it on where the question set is stable enough to script well. If it is not, we say so before building it.
Where PROOF48 fits
An AI chatbot is worth it exactly where the same narrow set of questions repeats often and a wrong answer is low stakes. Outside that, it adds a layer between the customer and a person who could actually help. PROOF48's AUTOMATE plan builds it only for the first case.
Cosmin Tiu · Founder, PROOF48
Cosmin has worked in online marketing and software development since 2018, and has built with AI every day since January 2023. He still writes code from scratch daily, in Python, React, Rust and C#. PROOF48 ships business websites in 48 hours and invoices afterwards, so every opinion on this blog was paid for in real builds, not in theory.