Description
Summary We are an Australian service-based business using **Sophiie AI** as an AI phone receptionist. A substantial portion of the initial setup has already been completed. We have created a number of scenarios, responses and triggers, but the receptionist is still making avoidable mistakes and regularly failing to follow the intended conversation flow. We are looking for someone with practical experience configuring AI voice agents, conversational AI systems or AI phone receptionists to review the existing setup and improve it. This is not a general AI automation project involving basic Zapier workflows, lead forms or generic chatbot setup. The work is specifically focused on improving how an AI receptionist understands callers, follows instructions, triggers the correct scenarios and handles phone conversations. ### Current issues Some of the problems we are experiencing include: * Scenarios not triggering when expected * The AI ignoring or inconsistently following configured instructions * Calls ending before enough information has been collected * The receptionist giving a basic response such as “I’ll pass that on” without properly handling the enquiry * Incorrect pronunciation of certain business terms, names or words * Difficulty accounting for different ways callers may phrase the same request * Conversation flows working during some tests but failing during others * The AI reverting to unhelpful default responses * Gaps between the scenarios we have created and how real callers actually speak For example, the AI may ask what the caller needs, receive an explanation and then simply say that the message will be passed on before ending the call. It may not collect the caller’s name, phone number, email address or sufficient details about the enquiry. ### Main objective We need the receptionist to handle common enquiries more reliably and follow a clear fallback process whenever a specific scenario cannot be matched. At a minimum, the fallback process should be able to: * Acknowledge the caller’s enquiry appropriately * Collect their name * Collect or confirm their phone number * Collect their email address where relevant * Record a useful summary of what they require * Direct them to the correct service email address where appropriate * Confirm what will happen next * Avoid ending the call before the required information has been collected The aim is not necessarily to make the AI capable of resolving every possible enquiry. The priority is ensuring that it behaves professionally, gathers useful information and follows an acceptable process even when it does not understand the request completely. ### Scope of work The work may include: * Reviewing our current Sophiie scenarios and instructions * Identifying conflicts, gaps or unclear instructions * Improving scenario triggers and alternative phrasings * Creating more reliable fallback logic * Structuring conversation flows so required details are collected before the call ends * Improving pronunciation of specific words, names and industry terminology * Testing the receptionist using a broad range of realistic call scenarios * Accounting for interruptions, unclear answers and callers providing information out of order * Ensuring the AI confirms important details instead of making assumptions * Recommending where the setup should be simplified rather than adding unnecessary scenarios * Documenting the final structure so we can make smaller updates ourselves later ### Experience required We are primarily looking for someone with experience in areas such as: * AI voice agents * Conversational AI * AI phone receptionists * Prompt and instruction design for voice models * Intent recognition and scenario triggering * Conversation flow design * Fallback and escalation logic * Speech recognition and pronunciation issues * Testing AI agents against realistic conversations Direct experience with Sophiie would be ideal, but it is not essential if you have strong experience configuring similar AI voice platforms and can quickly understand how its scenario system works. Experience with Australian service-based businesses would also be useful. The receptionist needs to understand fluent conversational English, Australian terminology and the different ways local customers may naturally describe an issue. This is a targeted setup and troubleshooting project. We are not looking for a broad AI consultancy package or a proposal to rebuild our entire business around AI automation. Please include details of any AI voice agents or phone receptionists you have configured, including the platform used and the type of conversation flows you worked on.