Akari Solutions help organisations to implement Microsoft technologies with a focus on change management, diversity and inclusion, and levelling the playing field for those with hidden or situational disabilities. This year we won Diversity & Inclusion Changemaker 2020 Microsoft Global Partner of the Year Award, as well as Finalist for Apps and Solution Microsoft Teams Award, and Teamwork Award. We have a wealth of experience in delivering change management engagements around Microsoft Teams, from small organisations, to large NHS trusts and enterprises, as well as building and deploying tailored Teams solutions for our customers. Promoting accessibility is at the heart of everything that we do, raising awareness of the importance of diversity and inclusion and ensuring each and every employee can contribute to the success of their organisation.
When we started our journey building a company with accessibility at the heart of it, the solutions and IP we planned to release had to be inclusive by design. Our mission is to provide opportunity to any individual, regardless of hidden disabilities, mental health, gender, or age.
We wanted to build a tool which could be used by anyone from students to employees within public sector or commercial organisations. One of the many barriers is cost of the applications and making the tool free and hosted within the UK will allow for educational organisations to provide ADI to all users to support with remote learning and understanding how to support themselves and their peers.
With around one in a hundred children in the UK with autism, and fewer than one in four school leavers with a hidden disability staying on further or higher education, there is a huge market for ADI to support those who are at risk of exclusion.
Accessibility is not only an internal consideration for public sector staff; A large proportion of staff are working with thousands of citizens, patients or customers who have specific requirements and organisations have an obligation to create better services for the end user, both internally and externally.
We've been fortunate to work with some amazing customers over the last 2 years, but our goal has always been to drive inclusion home without barriers. When we won Microsoft Global Partner of the Year award for Diversity and Inclusion Changemaker 2020, we decided it was time to give something back, and take everything we'd learned about apps and inclusion and build it into something that anyone can access without the barrier of cost.
What it does
ADI is a chatbot in Microsoft Teams which interacts with the user to answer questions and surface information on accessibility features such immersive reader, live captions, and message translation within Microsoft Teams, Windows 10 or Office 365. It can be used to support change management programmes and supplement training by pointing to bitesize videos, respond with conversational how-to answers, external resources on subjects such as diversity, inclusion and how to be a better ally.
ADI can assist with monitoring individual wellbeing, from starting a conversation about what troubles the them and what they are feeling to providing resources on/around mental health. The bot can also support employers to find resources and help understanding neurodiverse employees and how best to support them with their daily tasks. For a bit of fun and quirkiness ADI also tells jokes and provides fun daily affirmations. ADI can also assist students allowing them to ask question in their own time with a chat interface that they are familiar and comfortable with. ADI responds with useful facts and instructions around accessible tools including bitesize videos and links to external resources
We are keen for the knowledgebase to continue to grow and encourage users to submit topics or questions to a dedicated email address.
How we built it
The bot was a team effort, built over several months using the Microsoft Bot Framework SDK. Over the months we strategically planned and research into different questions and answers that we wanted to put in ADI's knowledgebase. We made use of Azure to host the app service for the bot and we deploy using GitHub Actions. We also made use of QnA Maker for the logic of ADI to allow us to build a large knowledgebase and tune this as we go on. The bot app service and QnA Maker app service are hosted on the same plan and we monitor the solution with App Insights. The bot code manages the user interactions and logging of questions and user queries.
Challenges we ran into
With the creation of the QnA maker we ran into some challenges with these being with the duplication of keywords within the QnA maker itself. For example if a user was asking how to support someone with dyslexia, they may have received the same response as other dyslexia questions/answers. We rebuilt the knowledgebase a number of times to get this right, and we believe that the managed QnA would allow us to tune this without developing dialogs. Additionally, we had to rethink our design to ensure that ADI wasn't just a QnA bot, this is something we are developing further.
Accomplishments that we're proud of
This is the first free bot which we have built using the latest bot framework methods, building ADI allowed us to deep-dive into building a simpler bot which would potentially be used at scale within one region. We have begun discussions with a number of schools and a university to see where ADI can be best applied and understand how we can use it, following into what we learned.
What we learned
From this experience we have learned and gained skills around the building and designing of a diversity & inclusion application. Alongside this, ADI is the first bot we have built with such a large knowledgebase, so we have taken learnings from this and are thinking on how we can apply this to other areas. Finally how to build scalable logging and message capture into the bot was a valuable learning exercise.
What's next for ADI (Akari Diversity & Inclusion)
We're hoping to build out a model to understand how people ask questions to ADI using a level of machine learning. To detect when questions are missed, how people are asking questions and 'speaking' to ADI. We also want to implement Language Understanding Intelligent Services (LUIS) to allow us to tune the QnA answers and potentially provide different QnA Maker knowledgebases per organisation. Finally we would make the switch to QnA Maker Managed when it becomes available within the UK. With ADI going through publication we are hoping to see it published into the store for installation anywhere.