CHATBOT MANAGEMENT PLATFORM
CMP
Conversational AI
CMP is a tool designed to transform the recruitment process. It functions as a personalized assistant for job seekers, seamlessly integrating with career websites to provide essential company and job details. Through customisable flows, users can tailor experiences for candidates, enhancing engagement levels. Leveraging LLM training, the AI evolves to emulate human interactions, fostering a more individualised approach.
Recruiters can effortlessly address candidate queries and share valuable questions with broader audiences. With options to customise bot interfaces according to brand identities, users have complete control over aesthetics and interactions. This versatile solution extends its capabilities beyond career websites to include platforms like WhatsApp and SMS, facilitating custom flows. Equipped with intuitive task management and workload optimisation features, the platform ensures a streamlined user experience, optimising recruitment processes.
TEAM
Machine Learning Engineer, Development Engineers, QA, Product Manager, UX Designer, Hybrid Designer (Me)
DURATION
24 Months
LOCATION
Onsite and Remote (Pandemic)
RESPONSIBILITIES
Competitor Analysis, Conceptualisation, Wireframing,
Visual Design, Testing
Our Happy Clients
These Clients were managed throughout the product lifecycle, addressing their requirements and implementing feature-specific updates and personalisation for the chatbot
And many more.
Michelle Hart
Lead Talent Strategist, Stanford Health Care
"Our leaders were blown away that so many people interact with the chatbot and get a lot of their responses that way which is why it's even more important that our responses are relevant and meeting candidates' needs."
Sarah Steinmann
Talent Acquisition Specialist, Southwest Airlines
“Our chatbot didn’t just have a random effect on our hiring processes, but it really directly impacted the candidates we saw and the future employees of Southwest Airlines."
Ashley Blckmore
Director, North America Talent Acquisition & Operations
"We have a 96% apply rate through the chatbot, which is amazing."
Brandon Prideaux
Director of Talent Acquisition, Community Medical Centers
“With the new chatbot, we’re able to engage with candidates at all hours and during weekends. Many of our job seekers are night nurses and other hard-to-reach clinical staff. This now serves that purpose."
Process
We designers followed an Agile UX process that is driven by providing feature-specific updates in an iterative environment by constantly communicating with development as well as customers
PLANNING & DISCOVERY
USER STORIES & SPRINT PLANNING
DESIGN
USABILITY TESTING
REWORK & IMPLEMENT
IMPROVEMENTS AND LAUNCH
Use Cases
A few use cases were defined based on product requirements from leadership and customer demands to establish clear goals in terms of deliverables
01.
One backend platform to manage and update all queries.
02.
Train bot to be more human like
03.
Customise Bot flows for every user scenario
04.
Realtime Career Site and Bot update and integration
05.
Bot interface personalisation
Tools
A few tools were used by the team's designers to establish an ecosystem of efficient and focus-driven task management
DESIGN
Sketch
USER TESTING
Fullstory
DOCUMENTATION
Evernote and Confluence
COMMUNICATION
Slack and Zoom
PROJECT MANAGEMENT
Jira
Information Architecture
A structure was established and finalised with the product manager to mitigate uncertainty in product design and development
Wireframes
After various iterations, a rough idea was prepared to comprehend user flows and navigation, while also considering the product's usability by different users
Knowledge Gaps
Reply to Candidate Query
Linking to FAQ
Bot Flows
Chatbot Profile
Style Guide
An entirely new library was designed to provide a fresh and appealing look, adhering to the company's brand guidelines, which are globally acceptable for the proposed revamp of the platform
Proposed a revamp for the existing platform running for 2 years based on the company's brand guidelines and aligning with a few internal applications. After multiple discussions, the management approved the request for CMP to get its own identity and then upon various iterations and visual audits the current design language was finalised. Few advantages were noticed as we revamped the entire platform which benefitted the team and the organisation.
1
Overall redesign of the application generated more engagement with clients and the product's demand
2
Many usability issues were tackled in the process and a few improvements in terms of navigation were thoroughly structured and planned in this phase
3
All UI components were presented and discussed with the development team beforehand and iterated to avoid potential challenges in building the system
Product Brief
An overview of the product and its features that were set as the requirement, followed by updates in the product lifecycle
KNOWLEDGE BASE
Unanswered Questions
Knowledge Gaps
Frequently Asked Questions
KNOWLEDGE BASE
A space to add and update information about the organisation and job details. All the information that the career site provides is expected to reflect on the chatbot. This space is primarily managed by annotators and recruiters.
Frequently Asked Questions
The primary hub of the entire platform where all the information expected by the user is stored with multiple variations to train the model.
BOT FLOWS
Career Site
SMS
Knowledge Gaps
Records all the instances where the users weren't satisfied with the answer provided by the bot. This is often triggered by the feedback feature provided after each answer to observe discrepancies in providing the desired information. These are resolved by either linking them to an existing FAQ or adding an entirely new FAQ.
Chatbot Profile
Small Talk
Fall Back
SETTINGS
Users:
Annotator
Recruiter
Hiring Manager
Admin
Unanswered Questions
The area where the bot is not capable enough to understand the question yet which already exists in the database but is asked in a different manner or an entirely new question that wasn't considered or has no information in the system. This is handled by either linking the question to an existing question if the answer already exists or providing a personalised answer to the candidate via email.
BOT FLOW
Customise bot conversation path based on candidate type on various mediums. The customisation of flow is not limited to the career site but extends to Whatsapp and SMS for frequent edits based on organisational updates and candidate requirements. This provides the candidates more personalised experience based on their queries and search results. Assigned users have full access to various categories of bot flows and a reset option to always go back to the default flow.
SETTINGS
This consists of all the configurations such as interface customisation, permissions, and channel utilisation. Settings also provide a section to make the conversation more human-like by providing small talk that provides the bot with the ability to respond to common phrases which aren’t related to company information. Also when the bot does not understand what a candidate has said it provides a fallback i.e. completely customisable, which is further handled in Knowledge Base.
Design Overview and Iterations
The idea generation process, along with iterative considerations for a few pages and features in the design timeline, continued until the result was achieved
LAYOUT DESIGN
1
2
Menu Design
Begun the revamp by going through the IA and restructuring the flows to solve the hierarchy issues and make improvements accordingly.
1
In the initial versions before the revamp the menu was quite simple and Knowledge Base accommodated all the flows in one screen with a differentiation of tab navigation. But as the product became more complex we decided to move them under the primary navigation to avoid clutter providing page reload.
2
As Knowledge Gaps was introduced we thought of creating a subtle urgency in handling the candidate queries by the users hence we came up with showing the number of unresolved queries for both Knowledge Gaps and unanswered queries.
A few Menu designs were iterated to experiment and make better judgements on what goes well in the platform's layout.
In the beginning, we tried to stick with the internal product's menu system and incorporate a fixed extension with clear labels but over the period of creating versions we went from collapsible menus to section-based hover menus, eventually, we made the menu fixed with clear labels of all the items visible all the times.
Type of Bot
The chatbot is used in 2 platforms. The primary usage is in career sites and secondary in the company's internal job portal making similar candidate usage available to internal employee's job movement. A filter at the header beside the client name was introduced when the feature rolled out for the users to update information by changing the type of bot.
1
2
3
Search
The search was only capable of generating results on queries in Knowledge Base. But as we want to improve the search on other factors involved to make it more powerful we introduced a complex search mechanism after discussing with the development team about its implementation possibilities.
1
A Recent Search was provided based on what the user has searched to give quick glance of previous activity. A text of what kind of search was provided differentiating from Query to Topic as the search shown in the image is from the FAQs page. This varied in other pages based on the priority of search results.
2
To increase more usage of filters rather than manually scrolling through results or sorting the status we introduced Status filters in search as a beta feature for the user to quickly access high-priority queries which are either pending approval or unpublished.
2
A clear differentiation by Topic and FAQ was provided upon typing by adding a title in order to avoid confusion on the type of search result.
Filters
We wanted to provide users with a more efficient way of searching queries. One was to introduce a powerful search but it was specific to one query. The second feature that could enable task management and monitoring was adding filters as per the list of queries. We discussed amongst the team the filters that would be useful for our users and came up with a design based on the new style guide that helped us reimagine the usability.
1
On click of the filter beside the search bar, the user can filter queries based on 4 criteria. Filtering by 'Assigned to' allowed management to identify the progress of work and avoid potential roadblocks by connecting with the respective owner for the task.
2
Bulk action was introduced on the selection of multiple queries from the list which provided options to optimise time in performing respective actions for example selecting multiple queries upon filtering by status and assigning owner, publish, unpublish, export or delete.
v1
v2
v3
1
2
Pagination
Until the revamp, the queries in FAQs, Knowledge Gaps, and Unanswered Questions were stacked one below another in the list which is accessible only by a lengthy scroll. After observing our client's usage, we came across a few who had more than 1000 queries handled in the Knowledge Base. An issue that was addressed by a few of the clients was about the page load time. After a certain threshold of queries, the page load was compromised. To tackle these issues and to promote the usage of the new search and filters we introduced pagination in the table. Users can see up to 10 queries at once and have to click 'next' to access the next 10 queries. This reduced page load and increased the speed of the page making it an ideal solution.
FAQs
Sidepanel
On click of an FAQ from the list, the user can access all the information related to the FAQ in a sidepanel. The actions that the user can perform are to change the question and answer, assign or modify the owner who will address the FAQ, add a topic that the FAQ falls under, and add similar questions to increase its confidence score.
1
An organisation having a career page with multilingual support needs a chatbot with the same language capabilities as selected by their candidates hence we wanted to provide editing an FAQ in multiple languages as a feature on click of a tab.
2
Status visibility after opening an FAQ on the query as the list behind hides the status column at the same time creating an urgency in handling an FAQ.
3
A character limit was provided as a recommendation based on the bot height and content displayed in it for better readability. This isn't a limitation as the user can type more than the limit if required but a suggestion from the product owner for a better experience.
1
2
3
4
4
Many times the users (owners) want to save the FAQ after adding relevant content to get approval or reiterate as they see fit before publishing it on the career site bot. 'Save' adds the FAQ in the database of the platform and on click of 'Save & Publish' the FAQ is pushed to reflect on the bot.
Modal
Adding an FAQ is kept to a minimum by only providing the Topic and FAQ that is necessary for information to be saved in the database. Adding a topic creates a pool of FAQs segregated to fall under a specific topic. By clicking on 'Add FAQ' the FAQ is added to the list on top which can then be further edited using the side panel if required. This has been kept with limited actions to provide bulk addition of FAQs as noticed commonly based on user behaviour.
Bot Trainer
'Confidence' is a feature solely added for training the bot to understand and handle similar FAQs which are paraphrased. An FAQ can be asked differently by different candidates. We wanted to make the bot smart enough to understand all the different variations of a query. A threshold was set hence the indicator displaying 3 levels. The more queries there are for an FAQ, the less chances of error and for the bot to be accurate. This was suggested by a machine learning engineer which was later translated into the current design.
Onboarding
Usually, our clients are given a walkthrough of the application but based on the user behaviour observed in Fullstory, we noticed that many first-time users face a few challenges while performing certain tasks or aren't aware of it. We wanted to provide basic instructions on what the platform is capable of by providing a few key information to our first-time users. This is to improve the overall experience of the application and reduce the overwhelming feeling of using a complex application.
SETTINGS
1
2
5
3
4
Bot Personalisation
A lot of clients had requests for customisation in colour, text, height and styling of the bot that they want to display on the career site. These requests are often taken as ad-hoc tasks by designers and developers. Handling individual client requests is a daunting task. At the same time, many clients are unable to make a clear judgement about the appearance of the bot. More often after delivering the customisations, they would seek extended requests to fix it. We wanted to provide a tool to provide customisation. We noted the criteria that create an impact, standardised the UI elements, tested with edge case scenarios and provided access to personalising bot as needed to the clients as a tool within the platform.
1
A set of bot icons were offered in this package with an option to upload their own icon. Upload feature was provided as few clients requested their company's avatar or logo as an icon.
2
Information provided to see where the customisation will impact and what changes can be made.
3
Easy access to create brand identity in the chatbot.
4
Most often requested customisation was the bot height but we thought of providing more after testing the aesthetics of push notification, icon plaement, and button styling.
5
A preview of all the customisation can be viewed on the page itself after making necessary changes to check how the bot looks before pushing it to live on the career site.
These design considerations and iterations are a few of the lot taken care of in the product design lifecycle which played a significant role in shaping the product until the end of my term in the organisation.
Final Design
The result of the design lifecycle in the form of user interfaces ready to be developed that meet the company's and customer's expectations
FREQUENTLY ASKED QUESTIONS
KNOWLEDGE GAPS
UNANSWERED QUESTIONS
BOT FLOWS
SETTINGS
OTHERS
Product Outcome
Goals achieved by the designer that set product benchmarks and demand in terms of usage from the developed version of the platform
Improved task management by providing an advanced search and filtering mechanism for easy info access.
Increased product demand by revamping the entire platform and restructuring the user flows by 30%.
Provided fully customisable Bot flow management for all channels and mediums.
Reduced employee dependency for ad-hoc tasks in terms of bot personalisation from clients by 80%.