Chatbot Documentation - Vernon by

Chatbot Concept for internal L&D unit - Learning @ XYZ Co

alisonmartin41 — 2022-09-15 00:27:35 UTC

Internal XYZ Co (company pseudonym used for privacy) L&D unit specific Chatbot to be available 24/7 to assist workers/users to quickly and at a time that suits them, access frequently requested information (eg upcoming courses & dates running, contacting a learning advisor) and complete frequently requested tasks for rescheduling and cancellation.

User Persona

alisonmartin41 — 2022-09-15 00:27:35 UTC

Link to Miro Board - https://miro.com/app/board/uXjVPXPOQU0=/?share_link_id=896016357605

Who is the Audience?

alisonmartin41 — 2022-09-15 00:27:35 UTC

Users

End Users - See User Persona below - Internal 200+ grants assessing officers at state government dept responsible for assessing and paying grants to businesses
System/Admin Users - Internal 5 x L&D staff in internal corporate L&D unit - Learning @ XYZ Co

No special user needs are present currently. No requirement is present for use of voice or auditory methods - specific user requirement is for text based solution. This is open to review at any point for future iterations of required solutions.

What Problem Are You Solving?

alisonmartin41 — 2022-09-15 00:27:35 UTC

Internal L&D unit of XYZ Co conducts internal training
4 out of the 5 L&D unit staff conduct class facilitation 3-4 days a week & internal service tasks must be conducted outside of class times
Therefore L&D staff not available - on demand - to answer customer service type questions in person
Ongoing data collection with workers (12mths +) indicates:
- increased number of calls to L&D unit with basic FAQs about internal courses
- decreased satisfaction with L&D unit speed & efficiency of meeting needs however staff ratio = 5 L&D staff to 200+ workers means on-demand internal service by this unit is not feasible or possible
Most common information & tasks requests were for:
- information on upcoming courses & dates
- requesting reschedules & cancellations
- information on venue & logistics on the day of a course (eg times, location, what to bring etc)
- speaking to a learning advisor
Frequently asked questions are currently available via a static web page & information is provided at the time of course enrolment - however picking up the phone & calling appears to be a first preference for most internal workers & the volume is increasing
Learning @ XYZ Co are aiming to reduce call volume, provide self servicing for basic regular enquiries & requests & hence improve satisfaction with their internal service. A need was identified for a domain specific (i.e., learning unit information only) 24/7 chatbot - to assist workers/users to - quickly & at a time that suits them - access frequently requested information and complete frequently requested tasks. The chatbot is intended to replace the FAQ web page.
Please refer to the User Persona above which outlines users particular needs for the chatbot- which includes:
- Quick, efficient, effective - workers are time limited with high workloads
- User agency is enhanced through own ability to to obtain information they need quickly & at time of their choosing
- Specific user preferences - without which may negatively impact chatbot uptake success - was identified from data collection & results from previous IT chatbot deployment
- direct chatbot guidance required rather than extensive interaction by to-&-fro typing with the chatbot
- simple text based chatbot preferred - no unnecessary content or graphics
- professional, friendly approach & tone

FAQ chatbots - Web based, Text Only, Rules based

alisonmartin41 — 2022-09-15 00:27:35 UTC

Rules based FAQ bot - answers common questions - follows decision tree conversation structure - predefined rules design to solve specific problems or achieve particular goals - can be simple or complex - personalisation happens through making choices & information conditioning. Rules based chatbots match user input to a rule pattern & select predefined answers from set of responses with use of algorithms. Answers are automated, repeated and while may lack originality & spontaneity of human interaction - the key benefit is fast response times (Adamopoulou & Moussiades, 2020; Sethi, 2020). Our chatbot user requirements specify rules based text chatbot with high needs for fast & efficient responses - deeper interaction & examination of input text is not required. User needs also specify that extensive to-&-fro typing interaction with the chatbot is outright rejected.

More friendly & attractive to users than static content searches (e.g. FAQ lists) as offer efficient assistance, providing more engaging answers & respond directly to the users problems (Adamopoulou & Moussiades, 2020). Our chatbot user requirements specify efficiency and direct responses.

Productivity is reported as most important motivation for chatbot users (Adamopoulou & Moussiades, 2020). Our chatbot user requirements specify productivity as a key motivator when using chatbots.

Knowledge domain can be generic (i.e., operate in more than 1 domain) or specific (i.e., respond only to questions concerning a particular knowledge or domain)(Adamopoulou & Moussiades, 2020). Our chatbot user requirements are for specific knowledge and tasks to be actioned.

Chatbot classification can be informational, conversational, and task-based. (Adamopoulou & Moussiades, 2020). Our chatbot user requirements are for informational and task-based aims for the chatbot.

Text based chatbots should be designed with the end user in mind - helping to ensure optimum user experience - and suggestions for design include:

Being human friendly (e.g., use of an avatar, professional attitude & tone, use natural language in the responses)
Keep conversation simple with chatbot easy to use (e.g., direct responses, no unnecessary jumping or branching)
Fewer typing, more clicking - typing a lot generally leads to lower completion rates - structured click & choose & structured responses can help smooth integration processes (Sethi, 2020).

Text rules based FAQ Chatbots can be simple, friendly & conversational - saving on employee costs & avoiding delays & human errors (Sethi, 2020)

Chatbot deployment benefits include improved efficiency, delay free and always available service, support for end users conducting practical simple tasks & making their life easier (Koivunen et al., 2022). Our chatbot user requirements specify these specifically as problems to be solved. Web based chatbot is specifically required as it is replacing web based FAQ page on a web page.

Rule based Chatbots are particularly effective in removing repetitive customer service tasks for workers such as answering common questions, assisting users, which when removed generates more times for workers to perform other tasks. Technology can be used to save time & improve productivity (Sandoval, 2018). Our chatbot user requirements include a required reduction in call volume & subsequent non productive time use i.e., time spent answering repetitive common questions or repetitive change requests.

User Flow

alisonmartin41 — 2022-09-15 00:27:35 UTC

A user flow was drafted to visualise elements
Miro link
https://miro.com/app/board/uXjVPXMiDk4=/?share_link_id=747859673762

Technology Used

alisonmartin41 — 2022-09-15 00:27:35 UTC

Landbot is a no-code, quick & easy to use, chatbot builder platform. It has a number of features that while not required for the current chatbot, could be utilised and expanded on in future iterations of this current initial chatbot eg. predesigned templates, human takeover, custom CSS & JS, additional capabilities through easy integration with other apps such as Google Dialogflow for NLP and machine learning, etc.

It has an intuitive interface, multiple integrations available & the ability to scale unlimited.

NLP Chatbot - Landbot Integration with Diaglogflow

alisonmartin41 — 2022-09-15 00:27:35 UTC

Other Chatbot Platforms

alisonmartin41 — 2022-09-15 00:27:35 UTC

landbot.io
chatbot.com
drift.com

Technical & Creative Challenges & How Solved

alisonmartin41 — 2022-09-15 00:27:35 UTC

Adamopoulou, E & Moussiades, L 2020, ‘Chatbots: History, technology, and applications’, Machine Learning with Applications, 2(100006-).

alisonmartin41 — 2022-09-15 08:10:50 UTC

Sethi, F 2020, 'FAQ (Frequently Asked Questions) ChatBot for Conversation' International journal of computer sciences and engineering, 8, pp. 7-10.

alisonmartin41 — 2022-09-15 08:25:56 UTC

Koivunen, S Ala-Luopa, S Olsson, T 2022, ‘The March of Chatbots into Recruitment: Recruiters’ Experiences, Expectations, and Design Opportunities’, Computer Supported Cooperative Work (CSCW): The Journal of Collaborative Computing and Work Practices, 31, pp. 487-516.

alisonmartin41 — 2022-09-15 08:45:41 UTC

Sandoval, ZV 2018, ‘Design and Implementation of a Chatbot in Online Higher Education Settings’, Issues in Information Systems, 19(4), pp. 44–52.

alisonmartin41 — 2022-09-15 10:36:23 UTC

Landbot 2022, Website Chatbot, Landbot, viewed 22 August 2022, .

GetApp comparison

alisonmartin41 — 2022-09-15 10:50:47 UTC

Landbot is slightly less expensive and has a free version compared to Chatbot. Drift would not provide pricing so a complete comparison of the 3 could not be done.

https://www.getapp.com.au/compare/108699/108899/chatbot/vs/drift?vs[]=71854

Other Technology Used

alisonmartin41 — 2022-09-15 11:06:54 UTC

Canva - Graphics
Miro - Online Whiteboard - Brainstorming, Flowcharts, User Persona

https://www.canva.com/en_au/
https://miro.com/about/

Website Chatbot Example - Scout - Suncorp

alisonmartin41 — 2022-09-15 11:14:47 UTC

"Scout", the Suncorp virtual assistant (i.e., chatbot), provides 24/7 support for customers to ask simple things like how much can I spend from my savings account, or when my car insurance expires.
https://www.suncorp.com.au/

Website Chatbot Example - Nomi - NRMA

alisonmartin41 — 2022-09-15 11:22:56 UTC

NRMA webchat bot "Nomi" provides assistance with general insurance
https://www.nrma.com.au/

Website Chatbot Example - Alex - ATO

alisonmartin41 — 2022-09-15 11:27:41 UTC

ATO Virtual Assistant (chatbot) - Alex - provides assistance to users on the ATO website

https://www.ato.gov.au/

Challenge #1: Development Process

alisonmartin41 — 2022-09-15 11:35:15 UTC

Challenge #1: A number of stakeholders from across the organisation would need to work together to develop the chatbot. What approach would be best to ensure rapid development, stakeholder input & deployment of the product?
Solution: An iterative agile framework was used as the approach and guiding principles for the project. This approach is one focused on speed, flexibility, teamwork and collaboration. Practices are merged into “sprints” where elements or “chunks” of the product are delivered for evaluation. This enables rapid prototyping where changes and shifts in content and directions from stakeholders can be easily accommodated. Collaboration with stakeholders to evaluate each product iteration is available at the end of each “sprint”. An added benefit of transparency is evident through the frequent evaluations allowing addressing of issues as they arise rather than at the very end of the product design process.

(Mokhtar & Khayyat, 2022)

Challenge #2: What will the Chatbot contain?

alisonmartin41 — 2022-09-15 11:35:25 UTC

Solution:

See Idea/Concept Section & Above Agile Methodology used
Plan, Design, Develop, Test - Quantitative & qualitative analysis of stakeholder needs resulted in development of items described in Idea/Concept section - User Persona, User Flow, Problem being Solved. Content(eg Q&As text) for the Chatbot was developed in consultation with stakeholders, then programmed in Landbot & initial testing with a small number of stakeholders.
Items arose & were resolved successfully in this part of the process before final testing, review & launch included:
- Minor changes to design of the Avatar - Vernon
- Minor changes to wording of Q&As
- Minor changes to layout from user feedback eg scrolling, frame, etc
UX guidelines for Designing Chatbots (Budiu, 2018) was also used in development including:
- Be upfront about using a bot and not a human
- Clearly tell people what tasks the bot can do. Make sure you don’t create false expectations
- Don’t be overly ambitious: create bots for simple tasks. Complexity is not well handled in the limited bot interface.
- Allow sorting and filtering to let people narrow down through results.
- Save information from one task to the next.
- Program some flexibility into the bot - allow people to go back in the linear flow.
- Offer an escape hatch in the form of a phone number

Laws of UX (Yablonski, 2020) were incorporated including:

Aesthetic-Usability Effect - a simple but aesthetically pleasing design was used to enhance usability
Doherty Threshold - users can interact at scale & pace - neither has to wait on others - enhancing productivity
Zeigarnik Effect - users remember uncompleted or interrupted tasks better than completed tasks - ensuring no false endings enhances user experience
Hicks Law - number & complexity of decision choices was kept to those necessary so that time to make a decision did not overly increase

alisonmartin41 — 2022-09-15 12:16:02 UTC

Mokhtar, R & Khayyat, M 2022, ‘A Comparative Case Study of Waterfall and Agile Management’, SAR Journal, 5(1), pp. 52–62.

alisonmartin41 — 2022-09-15 12:42:06 UTC

Budiu, R 2018, 'The user experience of chatbots', Neilsen, 25 November 2018 [Blog], viewed 22 August 2022, .

Challenge #3 - Deploy Pilot, Review, Launch

alisonmartin41 — 2022-09-15 12:54:36 UTC

Solution: After multiple iterations, testing & tweaking, the product was deployed for an initial 3 month pilot to the workgroup on 14/9/22 and analytics have commenced being tracked for review. Final changes will be made and if the pilot is successful, launch of the product into the wild. Ongoing monitor, review & evaluation will occur & continual improvement ongoing.

Vernon can be accessed at the following link:
https://sites.google.com/view/xyzcolearningfaqs

Flow Analytics

alisonmartin41 — 2022-09-15 13:15:13 UTC

So far:

Nearly all users (96%) are personalising experience by entering their name
13% of users enquiring about upcoming courses
29% of users requesting reschedule (majority are outside of 6 days where fee would apply)
11% of users requesting cancellation
37% of users enquiring about the venue & logistics
2% of users enquiring to speak to someone personally

alisonmartin41 — 2022-09-15 13:35:12 UTC

Yablonski, J 2020, Laws of UX: using psychology to design better products & services, CA, O'Reilly Media.