Conversational AI Summit schedule

Michael McTear is an Emeritus Professor at Ulster University with a special interest in spoken language technologies. He has been researching in the field of spoken dialogue systems for more than 20 years and is the author of several books, including Spoken Dialogue Technology: Toward The Conversational User Interface (Springer, 2004), Spoken Dialogue Systems (Morgan and Claypool, 2010), with Kristiina Jokinen, The Conversational Interface: Talking to Smart Devices (Springer, 2016), with Zoraida Callejas and David Griol, and Conversational AI ( Morgan & Claypool 2020). Michael has delivered keynote addresses and tutorials at many academic conferences and workshops, including SpeechTEK, Conversational Interaction, ProjectVoice, REWORK AI Assistant Summit, the European Chatbot Conferences, and the Altrusia Conversational AI and Customer Experience Summits. Currently Michael is involved in several research and development projects investigating the use of conversational agents in socially relevant projects such as mental health monitoring and home monitoring of older adults.

Conversational AI is a powerful technology and can deliver great results if used appropriately. In this talk we would discuss which types of interactions and projects are more suitable for Conversational AI and which are not. We will also cover some recipes to deliver a successful Conversational AI such as attention to conversational design, smart use of data and tracking the right metrics.

- The types of interactions suitable for conversational AI 
- Recipes to deliver a successful conversational AI 
- Smart use of data and tracking the right metrics 

Sid has been leading Intelligent Automation at BP since it’s inception and has architected several use cases by combining the power of AI and ML solutions with other automation tools to deliver end to end value. He is a great enthusiast about emerging technology and applying them to solve real world problems.

He is currently holding the role of Principal Automation architect at BP and instrumental in implementation of conversational AI solutions for a variety of use cases.

Sid holds a degree in chemistry and post graduate in HR and has worked on several technologies like SAP, Workday, Salesforce, Cloud native, AI ML etc.

With technologies becoming more advanced we had to build our chatbots to keep up, while also fulfilling the customers needs.

By hiring most of the team internally, this helped us focus on a more technical build as we brought in individuals who already understood Admiral’s goals and objectives, and the processes behind customer’s queries. This in turn helped streamline the design and build of our bots to the companies’ vision and customer’s needs.

It empowered the whole team to learn themselves how to re-build a chatbot better, sleeker and a design that can be re-used across multiple chatbots.

- Employing experienced staff members from within has allowed us to accelerate our conversation design and service capability within the bots 

- Allowed us to focus on learning the tech side to build better chatbots and scale up 

- “People who like what they do, do it better” - Admiral’s culture of enriching the staff to develop whilst also striving and maintaining excellent customer service has helped us make sure we are getting better all the time at building better chatbots 

Senior Chatbot Conversation Design working specifically with Dialogflow ES & CX and LivePerson's conversation builder. Worked in customer service for 12 years previously.

With technologies becoming more advanced we had to build our chatbots to keep up, while also fulfilling the customers needs.

By hiring most of the team internally, this helped us focus on a more technical build as we brought in individuals who already understood Admiral’s goals and objectives, and the processes behind customer’s queries. This in turn helped streamline the design and build of our bots to the companies’ vision and customer’s needs.

It empowered the whole team to learn themselves how to re-build a chatbot better, sleeker and a design that can be re-used across multiple chatbots.

- Employing experienced staff members from within has allowed us to accelerate our conversation design and service capability within the bots 

- Allowed us to focus on learning the tech side to build better chatbots and scale up 

- “People who like what they do, do it better” - Admiral’s culture of enriching the staff to develop whilst also striving and maintaining excellent customer service has helped us make sure we are getting better all the time at building better chatbots 

Enthusiastic about developing new skills to implement A.I. solutions which benefit both business and customer.

The main goal of this talk is to explore how NLP helped us to design solutions for our customer care. Our problems range from assigning the emails in customer care to the right agent, to transcribe Swiss German voice messages to German for further processing and a chatbot for automatically answering frequent customer questions. In this talk I will dive deeper, how we apply natural language processing to classify various products or processes in mails or intents in chats, how we use that information to automate certain tasks and therefore manage the workload of our agents. Furthermore I will show how to  integrate such a solution in on environment where our agents, can retrain them on demand and react to data drifts

- How to apply NLP in customer care applications

- How to design a managebale software that can react to changes

- How to integrate it into a CI/CD pipeline

With the background of a Master's degree in Computer Science with the focus of modeling
and scaling up AI models I am currently working as a Senior AI Engineer at Swiss Post.

My passion is to focus on delivering business value in customer care by operationalisation of use cases in production. Personally I am a networker and an open-minded, motivated person with the mission of transforming the Swiss Post to an AI driven organisation.

The main goal of this talk is to explore how NLP helped us to design solutions for our customer care. Our problems range from assigning the emails in customer care to the right agent, to transcribe Swiss German voice messages to German for further processing and a chatbot for automatically answering frequent customer questions. In this talk I will dive deeper, how we apply natural language processing to classify various products or processes in mails or intents in chats, how we use that information to automate certain tasks and therefore manage the workload of our agents. Furthermore I will show how to  integrate such a solution in on environment where our agents, can retrain them on demand and react to data drifts

- How to apply NLP in customer care applications

- How to design a managebale software that can react to changes

- How to integrate it into a CI/CD pipeline

During my MSc in Computer Science at ETHZ I discovered my deep fascination for machine learning and it’s capabilities for automation.

This fascination lead me to the Swiss Post where I’m developing, enhancing and implementing the machine learning backend for various applications. Examples include the chatbot or AI generated suggestions for customer support agents.

I particularly enjoy getting requests from support staff to enhance our software and cover more cases since it shows me that not only are we making processes more efficent, we also help people in their day to day lifes.

- It is possible to innovate and grow a new channel even within the confines of a larger organisation 

- Be agile… but take your stakeholders on the journey with you! 

- Innovation in the early stages can be fast with lots of quick wins… once its more mature then the stakeholders you’ll have brought on the journey with you will start to be impacted so keep them engaged and track the benefits. 

- Why is the technology so popular and what’s causing the shift towards voice? 

- Voice-first technology is enhancing the customer experience by adding convenience to communication 

- Utilising speech recognition technology to meet the shifting consumer demands and unlock business advantages 

Kamal is a technology leader obsessed with customer experience and have expertise in leading world class products including AI / Machine Learning, Cloud Computing, in a Software as a Service (SaaS) model. Currently he is working as Principal Data Engineering Manager for British Telecom (BT), London, UK.

Kamal has a rich global experience across the UK, US, Spain, South Korea, UAE, and India.

Kamal is a speaker at multiple tech events and voluntarily mentor young engineers across different start-ups, universities, global platforms and also publish articles in various international journals in his efforts to contribute back to society.

Michael is the Chief Science Officer at Prudential plc and the founder and head of Prudential’s Centre of Excellence for Artificial Intelligence (AI CoE).  He joined Prudential in 2016 from Silicon Valley based Pivotal Labs where he built and led the Data Science team. His experience lies in the application of artificial intelligence methods to large-scale, multi-structured data sets, in particular neural network based deep learning techniques. Michael previously founded and sold a London-based machine learning startup and prior to that was a partner at a major consulting firm. Michael holds an MBA, a PhD in theoretical physics from the University of Cambridge and is a Fellow of the Royal Statistical Society.

Andreas graduated with a computing master's degree in 1990 and has since worked in London, firstly as a software engineer before co-founding Biomni in 1999 and becoming its CTO. Over the years Andreas has continually improved quality software development practices, built and retained talented technology teams, and been instrumental in the establishment of successful global software products. Andreas and his team are now bringing their experience and skills to bear on designing and building conversational AI and knowledge networking features within Biomni’s latest product, Tenjin.

Job Search, when compared to most other online activities, has a vastly higher potential of being life altering and is usually in equal measure a lonely activity. The use of conversational AI, especially the inclusion of Large Language Models ( for instance GPT-3/ChatGPT models, Sparrow, Blender Bots etc), allows for the automated job search conversations to be far more engaging, supportive, and personalized compared to traditional interfaces. Large language models, however present their own challenges as in terms of fine tuning with enterprise specific data, hallucinations hindering reliability and cost. Some practical work arounds exit, but for now only certain use cases will see inclusion in Enterprise grade solutions. However, conversational interfaces in JobTech domain are bound to see significant progress in the future.

- Why conversational AI is best suited for JobTech given the unique nature of job search which requires multiple engagements and has softer aspects inherently entrenched. 

- Contemporary Large Language Models, a quick comparison of the same and top use cases from a Job Tech perspective – with insights from our implementations

- Key challenges faced when building Enterprise grade solutions using Large Language Models and potential work arounds explored by us.

Somnath Biswas has been working on enterprise and retail digital products for the past 20+ years, working with Tier 1 consultancies and multiple startups including his own in the enterprise conversational AI space. His products have spanned across Telecommunications, Financial services, Healthcare, Regulatory & Compliance and JobTech domains. For the past 6-7 years he has been building core ML and NLP products for the likes of NatWest, IQVIA, Amazon and most recently with the TotalJobs Group, where he heads the conversational products domain.

As Conversational AI continues to evolve and become more prevalent in our daily lives, it is important to examine the different approaches that are being taken to design and develop these systems. In this talk we will review two different strands of conversational AI: traditional conversation design and generative AI. The first part of the talk will outline the strengths and limitations of each approach and discuss their potential risks and challenges. In the second part we will explore some of the exciting possibilities of generative AI, including its potential to revolutionize application areas such as customer service, healthcare support, and content creation. The talk will conclude with a brief look at how traditional approaches based on best-practice guidelines are being combined with approaches based on generative AI in a large R&D project in which I am currently involved.

- There are two distinct strands in conversational AI, one employing traditional methods of conversation design and development based on best practices, and the other based on recent advances in generative AI in which large language models are trained to generate conversations automatically 

- Both approaches have strengths as well as limitations and it is important to consider these issues when developing a conversational assistant 

- Generative AI is an exciting new development in conversational AI that has the potential to revolutionize our approach to developing more powerful conversational assistant

Michael McTear is an Emeritus Professor at Ulster University with a special interest in spoken language technologies. He has been researching in the field of spoken dialogue systems for more than 20 years and is the author of several books, including Spoken Dialogue Technology: Toward The Conversational User Interface (Springer, 2004), Spoken Dialogue Systems (Morgan and Claypool, 2010), with Kristiina Jokinen, The Conversational Interface: Talking to Smart Devices (Springer, 2016), with Zoraida Callejas and David Griol, and Conversational AI ( Morgan & Claypool 2020). Michael has delivered keynote addresses and tutorials at many academic conferences and workshops, including SpeechTEK, Conversational Interaction, ProjectVoice, REWORK AI Assistant Summit, the European Chatbot Conferences, and the Altrusia Conversational AI and Customer Experience Summits. Currently Michael is involved in several research and development projects investigating the use of conversational agents in socially relevant projects such as mental health monitoring and home monitoring of older adults.

With technical advancement in conversational AI moving at a dizzying pace, it can get overwhelming for businesses to make the right bets and receive the best possible return on their investments. In this presentation, I look forward to sharing some insights on how to approach this by selecting the right outcomes and tracking the right metrics. New AI advancements often create a lot of excitement in tech enthusiasts and early adopters but may also make a lot of conservative business leaders uncomfortable at the thought of investing in such technologies that have a level of uncertainty associated with it. Does the recent innovations in the space of Conversations AI like OpenAI's ChatGPT improve the success rate of conversational AI technology?

- Removing the overwhelming feeling of new technical advancements in conversational AI 
- Selecting the right outcomes and tracking the right metrics 
- Does the recent innovations in the space of conversations AI improve the success rate of conversational AI technology? 

With around 10 years of experience in tech Industry and most of those years focused on Natural Language Processing (NLP) ranging from but not limited to semantic search, conversational AI, knowledge graphs, ontology development, recommendation engines, Jyoti has worked with companies ranging from very early stage startups to big multinational organisations. Jyoti is a T shaped professional and has worn different hats while solving NLP problems ranging from User Research, Product Management, Roadmap Development, Key Metrics Identification, Data Governance, Data Quality and Integrity, Data Security to implementing latest research papers in NLP domain, productizing cutting edge NLP solutions, working on backend engineering to implement NLP models, DevOps, MLOps, Data Engineering and ML/NLP Engineering. Jyoti has enabled non technical teams and stakeholders to make informed decisions throughout NLP product lifecycle to make efficient use of resources. She has also worked along side Legal, Privacy, Security, Marketing, Competitive Intelligence teams to enhance user privacy and also attain a leading position in the market.

In this workshop, you will learn how to leverage Large Language Models (LLMs) to develop and deploy intelligent virtual assistants in a faster and smarter way. With the advent of LLMs, such as GPT-4, the potential of virtual assistants has increased significantly. In this workshop, you will learn how to leverage LLMs for virtual assistant development, including designing and building conversational flows, model fine-tuning, and deployment. The key takeaways of this workshop include:

- Understanding the benefits of LLMs in developing and deploying virtual assistants

- Gaining practical knowledge of how to leverage LLMs in the Kore.ai XO Platform to transform your virtual assistant strategy

9 years in the systems automation, process orchestration and conversational AI space delivering extraordinary experiences and business value with cutting-edge tech.

Tessa specialises in the ethical governance of algorithmic and data intensive systems, considering dimensions such as fairness, accountability, transparency and explainability. She has delivered Responsible AI programs covering governance, training and resource development in scientific publication and health domains, and is part of an international research group which explores how trust is built in digital environments.

Conversational AI has enabled interesting human computer interaction scenarios. In this talk we discuss one such unique solution space where we leverage conversational AI to personalize to the users and deliver unique advertising interfaces to them. The current applications of conversational AI focuses on applications where humans initiate a conversation with a specific query in mind. We extend conversational AI to scenarios where computers can initiate a conversation with humans and deliver advertising messages back to them. We focus on unique challenges that confront such a system and how we can leverage signals from user interactions to model latent intent of the users and orient them towards task completion.

- How to model task completion through conversations 
- How to model latent intents and conversation starters 
- Impact of modeling relevance of conversation along with objectives around task completion 

Rahul Agrawal is a senior director AI at Sharechat where he leads a team of 40+ machine learning engineers and scientists to build the computational advertising platform. Prior to Sharechat, he has worked at Meta, Microsoft Bing, Yahoo! Labs, and Veveo. He has 18+ years of experience in building large scale recommendation systems, natural language understanding/generation, computational advertising, and large scale ML on graphs.

At Sanofi, we have implemented chatbots to improve the way our employees communicate and get work done. Our chatbots are helping employees to quickly and easily access HR information, request time off, and submit expenses, freeing them up to focus on more important tasks. Additionally, our chatbots are enabling cross-functional and cross-border collaboration, making it easier for teams to work together and share information regardless of location. In this presentation, we will share case studies and best practices for implementing chatbots in the workplace and discuss the benefits we have seen at Sanofi. Attendees will learn how chatbots can help to improve efficiency, increase productivity, and create a better work experience for employees.

- Implementing chatbots can help to improve efficiency, increase productivity, and create a better work experience for employees.
- To be successful, chatbots must be continuously monitored and updated to ensure that they are delivering value and meeting the evolving needs of users.
- Chatbots have the potential to transform industries and the way we work, and it is important for organizations to stay informed and embrace new technologies as they emerge.

Rodolf is responsible for the architecture of chatbots at Sanofi, having gained four years of experience in the field. He has previously worked at Wizz Air, where he was part of the team who created the chatbot Amelia, and at Artive, where he contributed to the development of the SCAI chatbot platform. Rodolf has also partnered with various top-tier AI applications, such as synthesia.io and IBM Watson Assistant, and has expertise in natural language processing, machine learning, and artificial intelligence. Currently, he is working at Sanofi on a mission to create and develop a standalone multi-intent chatbot to provide the best possible experience for Sanofi employees.

In this talk, I will explore the significance of understanding human behaviour by adopting a theory-based approach in the design of conversational AI, using Phyllis as an example. Phyllis is an evidence-based AI chatbot which is being designed to provide personalised conversational support to children and young people (CYP) in overcoming barriers to physical activity (PA) and addressing the global inactivity crisis. The chatbot aims to transform lives at scale by offering evidence-based support to educate and motivate children and young people to be more physically active. Advances in AI mean that organisations can go ‘beyond a service’ and provide a wellbeing centric approach to support people creating mutual benefit for both people and organisations.   

- Understanding Behaviour and Barriers: The talk emphasises the importance of comprehending human behaviour, particularly the barriers individuals face in accessing services. By identifying and understanding these barriers, businesses and organisations can develop effective strategies to address them, leading to improved outcomes and user satisfaction.

- Theory-Based Approach: By employing a theory-based approach, businesses and organisations can leverage existing knowledge and research in behavioural sciences to enhance their services. Phyllis demonstrates how theory-driven interventions can be applied to design comprehensive solutions that align with user motivations and effectively promote behaviour change.

- Evidence-Based Design and Evaluation: Phyllis highlights the significance of evidence-based design in conversational AI. By basing interventions and strategies on solid empirical findings, businesses and organisations can increase the effectiveness of their services and minimise potential risks. Additionally, the talk emphasises how a theory-based approach can enhance evaluation to measure the impact and success of conversational AI solutions.

Richard is a Senior Research Fellow whose background is in applied research, leading evaluations of national sport and physical activity programmes in the UK, for clients including the Premier League, Rugby Football Union and Department of Health and Social Care. He now applies this knowledge to develop novel, evidence based Conversational AI solutions which support young people and adults to improve their health and wellbeing and leads the Conversational AI Research Cluster at Sheffield Hallam University, working on projects for a range of organisations including the NHS (i.e., DAFNE). He’s also the founder of Phyllis (Phyllis.chat), a chatbot which aims to support adolescents to overcome barriers to physical activity and is a mentor on the Advanced Wellbeing Accelerator Programme (AWRC) which support start-ups (i.e., Therapha) in health and wellbeing.

Joshua is a software engineer, technology architect, and entrepreneur specialising in machine learning, automation, and AI. Currently, Joshua leads the product development and innovation team at Tovie AI, a UK-based company that provides conversational AI solutions to help businesses grow by enhancing customer experiences, improving internal processes, and creating monetisation opportunities.

Before founding Tovie AI, Joshua spent 10 years leading conversational AI business functions for several European software consultancies, including Infosys Consulting Europe, where he delivered bespoke and third-party AI solutions to the European enterprise market.

Johannes studied Business Computer Science and Data Engineering in Potsdam, Germany, at the HPI. During his studies, he founded an AI consultancy focusing on unstructured data, which eventually turned into Kern AI. With Kern AI, he's following an open-source approach to help developers work on reliable AI, e.g. in conversational AI. Further, Johannes enjoys to teach about AI in online courses and has already reached more than 24,000 participants in 4-week-courses to explain AI in simple terms.

Question answering technology has the potential to improve customer service by providing instant and accurate answers to inquiries. In this talk, we will explore the latest developments in question answering technology, including the use of large language models and neural networks, and how we build German question answering services for our voice platform. In this talk, we will discuss the benefits of using question answering, as well as the challenges and limitations of current technology.

- Question answering system for a voice platform
- Is multilingual better than mono-lingual models for NLP ?
- Challenges and limitation of current technologies

As a Senior Data Scientist at Deutsche Telekom, Fang Xu specializes in AI technologies for Natural Language Processing (NLP). Having completed his Ph.D in the field, he brings extensive project experience to the table, having worked on a wide range of topics such as question answering, text ranking, and chatbots development. Most recently, he worked on question answering systems for Telekom's Magenta voice speaker and platform. He is not only involved in ideation and research, but also in bringing these ideas to life, taking them from early prototypes to scalable production.

The proliferation of conversational AI technology is transforming the way we interact with machines and access information. However, as its usage becomes more prevalent, it is imperative that we consider the implications on user's safety and privacy. This session will cover the necessary facets of safeguarding and duty of care with regards to conversational models. The importance of privacy and data protection, the need for transparency in AI systems, the potential for bias and discrimination, the requirement for explainability, the need for safety, the role of human oversight, and the importance of continuous monitoring - by understanding these key aspects, we can ensure that systems are used responsibly and ethically.

- The ethical considerations
- Practical steps that should be taken
- The core principles to apply that will ensure that conversational interfaces are employed in a safe and responsible manner.

Jon is an Executive Product Manager for BBC Children’s and Education with a responsibility for bringing the benefits of new technologies to young audiences. He has worked with AI, intelligent systems and digital products for over 15 years. Jon was a member of the core team that authored the BBC Machine Learning Engine Principles, as well as delivering many successful innovative awarding-winning products.

- Building a framework of ethics 

- Ethical concerns in conversational AI differ depending on the application domain, target user group and goals of the agent 

- Encouraging contextual and plural approaches over a set of abstract principles to develop ethical and responsible AI 

Rahul Agrawal is a senior director AI at Sharechat where he leads a team of 40+ machine learning engineers and scientists to build the computational advertising platform. Prior to Sharechat, he has worked at Meta, Microsoft Bing, Yahoo! Labs, and Veveo. He has 18+ years of experience in building large scale recommendation systems, natural language understanding/generation, computational advertising, and large scale ML on graphs.

I run BT's AI & Data Science research programme with a team of 25 researchers at BT and 15 scientists from our global network of universities and research collaborations. The programme looks at wide spectrum of AI technologies, like NLP, Autonomics, Federated Learning, Ethical AI, AI Safety & Governance, Bias & Fairness Metrics, Anomaly Detection amongst others.

I am a computer scientist by training with a background in AI, Machine Learning and Data Science and 30+ years of experience in the field.

I have a strong interest in Responsible AI and establishing best practice in Data Science and Machine Learning. My aim is to develop Data Science and Machine Learning into a rigorous field of engineering where we create beneficial solutions, understand the impact of our work, and take responsibility for what we build.

I lead on Responsible Data and AI for the BBC, ensuring that the BBC’s use of data and AI align with our values and legal and regulatory obligations. I’m responsible for governance of internal development, procurement, and editorial use of AI/ML through our Machine Learning Engine Principles and checklist; and for turning principles into practice by providing teams with tools and support, and building capacity and culture throughout the BBC. Previously I worked in the BBC’s conversational AI team, including responsibility for ethics and editorial standards in conversational AI.

Julian is a product designer at Meta, an ex startup founder and a design ethics advocate. Before joining Meta in 2020, and as part of his work for Camelot, he redesigned the new digital Play experience for all of the UK National Lottery online games (Lotto, EuroMillions, etc.) leading to record sales. He was also in charge of designing several RCS business chatbots for a well know telecom operator, in both English and Arabic. Over the years he has grown to embrace a more society-centric approach to design, firmly believing in designing solutions that serve business objectives without compromising ethics and inclusivity. In his free time, Julian is an AI and smart home enthusiast.

Tessa specialises in the ethical governance of algorithmic and data intensive systems, considering dimensions such as fairness, accountability, transparency and explainability. She has delivered Responsible AI programs covering governance, training and resource development in scientific publication and health domains, and is part of an international research group which explores how trust is built in digital environments.