b'2. Faster Response Times then and there without having to speak with anyone. This FOGHORN FOCUS Before the chatbot, response times depended on staff avail- kind of feedback reassured us that the chatbot was meeting ability.Now,answersaredeliveredinstantly,whichhasand even exceeding expectations. significantly improved customer satisfaction. LESSONS LEARNED3. Improved Team Efficiency Theroadtoimplementingachatbotwasntwithoutits With repetitive inquiries taken care of, our team can nowbumps. Here are a few takeaways from the process:dedicate more time to personalized services, such as plan-ning group visits or assisting customers with unique needs.1. Start Small and RefineWe began by programming the chatbot to handle a lim-ited set of FAQs, ensuring accuracy before expanding its capabilities. This gradual rollout allowed us to monitor its The success of the chatbotperformance and make necessary tweaks. We also closely monitoredwhenthebotwasstumpedandwhatitwas has inspired us to thinkstumped by to ensure it had the correct answer for the next time. Over time, as it continues to learn, this happens less about how else we can and less frequently as the bot is prepared to answer more integrate AI into our and more questions on its own.2. Regular Updates Matteroperations. For example,As we added new tours and promotions, we updated the chatbots knowledge base to reflect these changes. Keeping were exploring options toit current was an ongoing effort, but it was essential for maintaining trust and reliability. Going forward, the new use the chatbot for bot will be able to pull new information directly from our site and adapt its answers without us having to provide ad-upselling and guiding ditional input or training. customers directly to the3. Monitor and AdaptAnalyzing chatbot performance metrics, like its hand-chatbot from their off rate and customer feedback, has been important for making improvements. For example, in the beginning, pre-arrival SMS message.wenoticeditwashavingtroubleansweringquestions about when the best time was to see whales when the guest gave what it considered a vague timeline, such as the last week in September. To combat this, we adjust-THE CUSTOMER PERSPECTIVE ed the chatbot to recognize this type of question and Initially,weworriedaboutwhethercustomerswouldrespond correctly. warm to interacting with a chatbot. Would they see it as impersonal? The feedback, however, was overwhelminglyLOOKING AHEADpositive. Many customers appreciated the instant answersThe success of the chatbot has inspired us to think about PHOTOS: MEGAN CLOUSE/CAPN FISHS and ease of use, and they particularly seemed to like thehow else we can integrate AI into our operations. For ex-abilitytotextourstaffviaSMSwhentheyhadmoreample, were exploring options to use the chatbot for up-complex questions or arrangements beyond what the botselling and guiding customers directly to the chatbot from could handle.their pre-arrival SMS message. Were also now utilizing AI to generate a first draft of Google and Tripadvisor review One customer shared: I wasnt expecting much when Iresponses, which continue to get more and more personal-clicked on the chat but ended up booking my trip rightized and save our team time. FOGHORN 36'