2024-07-12 Ocelot platform review

Owen Knight, Alex Pritchard, Mike Goodman, Scott Zitzmann, Shawn Abbott, Duane Pitre, Mike Pilman, Colette, Denise, Katie McBride, Catie Woessner (rep)

chatbot = virtual assistant

Ocelot support: support.ocelotbot.com

Live webinars 3 days/week.

Admin portal

admin.ocelotbot.com

Test My Bot

• Answers questions, operates just how the bot will when live (without pulling in testing data).

• Languages: English, Spanish, Chinese, Vietnamese

Look & Feel

• Where chatbot is positioned and what its messaging is (globally or by office). Should it look the same across all units or should it look different based on unity specification.

• Knowledge base is shared across all units once embedded on website.

• If on Advising and asks a Fin Aid question, it will still respond to the Fin Aid question.

• Position usually bottom right.

• Simple vs call-out bubble. Callout a little larger, more engagement. Simple is just a button.

• Roaming - allows bot to scroll up the page to prevent covering information.

• Callout Heading and message. Can keep consistent and at office level, can interchange Fin Aid, Advising, etc.

  • Can be used to do announcements, too.

• URL Matching - matched home pages for us. The URL is the parent page, and all child pages will have the same look & feel settings.

• Office overrides global settings.

Behavior settings

Office >

• Operation mode - Bot first vs live chat first

  • As soon as you start using live chat and student indicates they want to speak with someone, it would assign them to the live chat agent directly.

  • During busy periods, may want to do bot first and then escalate to live chat.

  • Can choose to require authentication prior to assigned to live agent.

• IDK settings

  • When asked a question the chatbot can’t support, it will notify the student and ask if they want to contact the office.

  • Can give custom response to this or walk them through a contact form. Chat transcript and contact form info would be emailed to the responsible office.

• User Feedback

  • Turn on at school level.

  • Any time a response is presented it will ask the student if their question got asked (thumbs up or thumbs down). Can leave comment as to why.

• Inquiry Form

  • Used to collect personal information from students from the moment they open the chatbot.

  • Before you get started, can you answer these questions?

  • Choose from any of 30 options.

  • Good for prospective students.

  • Can have digest or immediate email notification upon submission.

Global >

• Automatic Content Generation

  • Spiders for each unit.

  • Tells AI to scrape unit home page and subpages.

  • Can display site map to show what’s included in spider.

  • Can exclude pages. Maybe it’s an old URL that doesn’t exist anymore. Only trying to scrape applicable web pages.

  • Spidered information is used

  • Spidered information supports students students in 3 ways:

    • Explore bar link to pull up the URLs on your website to navigate to.

    • Responses: Content generator. Tell it to only use Spidered content. Can still edit/modify responses.

      • Option to use spider or GPT or both spider and GPT.

    • Automatic content generation

  • Content generation hierarchy: Knowledge base for 1:1 response, look for suggestions > no answer but content available > responds with that available content

  • Users and departments notified which responses are coming from knowledge base vs website (generative AI).

  • Turned off by default which means it only pull from knowledge base/custom responses.

  • May be configurable in the future by office.

Variables

• General - office contact info, tied libraries to offices (content lives in general library).

  • Asterisk variables are required.

  • Custom variables - we will want variables for respective portals (Slate)

Going live

• Launch can go live all at the same time or each can roll out at different times.

• Embedding > copy code

• All unit pages/subpages

How the AI works

• Office set-up form collected any contact related info for each unit. (How can I contact you? Shows all supported areas. Student selects. Displays email, phone, office hours, and ask if they need help contacting someone)

• Turned on entire libraries of content to support each unit. Tied to office permissions (Advising cannot edit Fin Aid questions).

• To change contact info, contact CSM to adjust office information.

• When questions are asked: Looks at intents and then entities to provide consistent, accurate responses to the different ways a question is asked (misspellings, alternate phrasing)

• Suggestions for related questions, links to applicable pages on our website for more information about their question.

Next meeting 7/24 3pm

Catie will ensure that all on this call are included on the calendar event.

Questions

SA: We have 5 units onboarding WaveBot - will there be 5 instances or will there be 1 instance with each of the 5 units?

CR: Sending a user to office information. If we have ticketing system, can we include that to get our customer response?

CR: Slate integration?

DP: How often do spiders run?

SA: How should we handle multiple portals? Slate, for example.

DP: Can embed code be inserted in body? We don’t have access to the head tag.

Catie’s follow-up

Here are the next steps and resources for each of the office settings we covered:

• Review the matched URLs to ensure we captured the correct departmental home pages.

• Add or adjust any URLs included in your spider.

• Complete, review, and adjust your required variables as needed.

Resources:

• Platform Review Recording - here (password: V8#^A^B6)

• Admin Portal URL - https://admin.ocelotbot.com; your username is your work email address

• User Roles and Permissions

• Welcome Message Examples:

  • Boston University

  • Empire State University

  • Denver University

  • Boston College

Topics