Context

Kare (then called Gluru) started as an AI powered to do list app for busy professionals. Upon realising few business opportunities, in 2017, we decided to pivot the business into the customer service area. We already had the foundation of ML technologies from the AI-powered To-Do list app and our aim was to apply these existing technologies to build a cognitive automation platform for  for businesses, that enables them to efficiently communicate with their customers through self service while helping them to keep their knowledge base up to date.

Kare MIND is the only Cognitive Automation Platform that integrates and manages entire company knowledge through the real customer insights, while focusing on providing conversational Customer Experience.

Metrics
0 to 5

paying customer

90%

deflection rate

*industry average 25%

2537

response created

1823

improvement cards approved

Request the case study
Role

Duration: April 2016 - February 2020
Team: 10+ Developers & 1 UI Designer

I joined Kare a year before the pivoting as the Lead UX designer, and undertook the Product Manager and Head of Product roles in until February 2020 as the company grew.
As the Head of Product, I drove the product vision and I was responsible for the product strategy, prioritisation and end to end product design cycle. I defined the roadmap and made sure that the product mets its key milestones, converted customer and sales lead requirements into features and incorporated in the roadmap; while communicating it with all stakeholders.

Accomplishments

• Owned the product strategy and contributed to the business strategy

• Brought the product from pivoting to scale-up that unlocks a healthy business with 5 live customers and an additional £600k+ projected ARR 

• Designed the UX architecture, created wireframes tackling complex problems in Human-AI Interaction 

• Contributed sales pitches, conducted stakeholder and client workshops, collected requirements and use cases (with the likes of GSK, Aegon, British Gas)

• Conducted user research and usability tests with customers and end users

• Collaborated with product development team for product consistency, and R&D team for synchronising on ML requirements

• Contributed to 3 patents in AI/ML, written a research paper on cognitive modelling and system architecture

Awards

Top 10 Customer Experience Companies
2019 - CIO Review

Next Gen Innovation for Customer Service

2018 - CIO Review

Best Innovation in Deep Learning
2017 - AI Summit

Best AI Start-up
2016 - AI Summit

Patents & Papers

• Method and system for a user-specific cognitive unit that enhances generic recommendation systems [patent]

• Semantic definition language for action discovery in cloud services and smart devices [patent]

• Application of unsupervised question clustering in conjunction with supervised answer labelling to discover and create missing nodes and missing edges in questions-answers knowledge graphs powering [provisional patent]

• Towards predicting cognitive behaviours by learning personal situational biases [paper]

Process overview

Customer and
Market Research

Researched with more than 15 companies
Run usability testing sessions with managers and users
Conducted workshops with customers and leads
Collected requirements and architected solutions

Challenge

Companies are usually blind when it comes to knowledge creation and enabling automation to their customers. This process occurs in isolation, mostly in advance, even before any actual customer traffic; followed by a cumbersome process to manage and update the knowledge. Loads of valuable data is getting lost in the process, which causes companies to miss impactful automation opportunities, resulting with increased inbound tickets and cost.

Strategy

Our main value proposition is to provide a data driven, intuitive way for businesses to create and manage customer facing knowledge. We achieved that by applying our NLP technologies and other Neural Networks in a cognitive architecture; so that it can answer customer questions while analysing them to identify knowledge gaps and potential automations.
This provided a consistent way of knowledge creation and management through insights from real customer activity, while providing self service to customers to reduce contacts.

Product Design

Overall, the platform initially consists of 2 main components:
• Customer interface
• Admin console

Customer interface is a widget where end customers can interact and get information from.
Dialogs are powered by the flexible dialog structure instilled in the knowledge graph, that enables the customer to communicate with the business knowledge base intuitively in a conversational way.
User Flow for dialog engine protocol
Conducted 15+ user testing sessions with WeWork members, with a prototype that’s mimicking the WeWork help center with the Kare widget on it. Participants are described some scenarios, and asked to resolve their issues using the prorotypes help center.
All steps on the conversation flow are individual responses with their own queries and specifications, and represented as nodes on the knowledge graph. So the system enables the end user to start from any step according to the amount of information provided in the query.
Admin Console
Main ux architecture of the admin console
Improve page is the A-HA moment of Kare product where the system communicates with the knowledge manager to take semantic actions and improve the knowledge.
Responses are the conversational elements which consist of building blocks such as actions, videos, documents etc. They can also be linked together to build flexible dialog flows. Responses represented as nodes on the knowledge graph and trained with customer queries by the knowledge manager.
Analytics dashboard is designed after first couple live customers, with their requirements and field research take aways in mind.
Dashboard for customer activity and knowledge performance
(scroll on the screen to see more)
See more UI on Behance