Public Funding for Action Artificial Intelligence Limited

Whether you are a small business or large enterprise, a Telephone Virtual Assistant (TVA) for automated customer services can be a huge asset to your business. These 'robotic concierges' are on hand and available to work 24/7 and will tirelessly attend to customer queries to ensure they are answered, actioned, and resolved. The goal of this project responds to the difficulties organisations have in identifying and accessing specialist expertise necessary to create a sophisticated TVA, and responds to the weaknesses of developer tools on the market commonly available to build such assistants, such as Google's Dialogflow. The outcome of this project will be a new software development toolset that allows businesses to easily and cheaply create powerful automated customer service TVAs that have human-like conversations with customers. A TVA created by a business will be tailored to suit the exact needs of customers, and importantly, will not require specialist expertise in-house. This project extends work undertaken in highly successful Innovate UK projects that resulted in building industry-specific and multilingual TVAs that are sophisticated in understanding users' language. These TVAs far outperform those created using conventional tools on the market.

The result of this research and development project will be a set of new software development tools that enable businesses to create their own bespoke, artificially intelligent Conversational User Interfaces (CUIs). CUIs operate on channels such as Amazon Alexa or Facebook Messenger and allow users to perform a range of tasks using natural language via voice or text. The planned work in this project follows work from project 105036 which demonstrated that, currently, commercial strength CUIs can only be developed with deep expertise in CUI design, computational linguistics, and machine learning. Existing development tools are not sufficiently sophisticated and result in CUIs that are oversimplified and not strong enough for commercial use. The CUI development environment that is proposed addresses this core barrier to entry in the CUI market. It will support businesses through all development lifecycle stages of CUI design, and, where beneficial, seamlessly enable crowdsourced assistance directly within this lifecycle when required. The crowdsourcing element of the tool seamlessly tasks: (i) external people working remotely who can collectively carry out low skilled, repetitive microtasks, such as creating and labelling large amounts of training data; (ii) external experts who can work remotely to carry out specialist macrotasks in the CUI development lifecycle; such expertise is typically unavailable within organisations wishing to create CUIs. The new tool will be developed by action.ai. Work is grounded on prior research and understanding of how reusable, commercial-strength CUIs can be created for specific business verticals.

COVID-19 has placed tremendous strain on many SMEs' finances. Falling demand has threatened commercial viability, especially for businesses with low cash reserves or unstable cash flows. 80% of SME failures are a result of cash flow problems. Deloitte's report 'COVID-19: Managing Cash Flow' (2020) highlighted the centrality of effective cash flow management to businesses' chances of surviving the pandemic. This project aims to provide a tool that makes it easy and quick for non-experts to understand and better manage their company's cash flow. The technology proposed makes cash flow management simpler, quicker, and more accessible for all SMEs. This will be achieved by building an artificially intelligent Virtual Assistant (VA) that is permanently on call for users via text or voice interfaces. Business owners can access the VA to interrogate their cash flow positions using normal, conversational language through a range of channels including instant messenger, Webchat, or via the phone using automatic speech recognition. Open banking integration will help pull information into one place. Market distribution will be via partnerships with existing suppliers of services to SMEs. The objective is to save thousands of UK businesses over a five-year period; this will be achieved by reducing the failure rates of the SMEs using the proposed service. The VA bypasses the need to use financial terminology because users can express themselves naturally. For example, a business owner may say: "I reckon a third of my clients are slow paying. Who are they and how long are they taking on average, and can you tell me what I'm owed?". Responses will be immediate, providing up-to-date information from both bank accounts and accounting packages - all presented back to users in everyday natural language. Business owners and managers can then take action or request more detail.

COVID-19 has placed unprecedented strain on many peoples' finances, whilst at the same time creating new pressures on financial institutions. The number of people in debt is likely to increase significantly leaving many in financial distress and placing significant burdens on financial institutions. The resultant social, health, and economic costs threaten to be severe. This work programme provides a conversational interface for people in financial distress to enable them to engage with their financial institutions in a user-friendly way via frictionless, asynchronous, and automated technology. The benefits include providing people in financial distress with a supportive and sympathetic process that they can use in their own time. For the financial institutions, they benefit from greatly reduced costs of service. Commercial-strength Conversational Interfaces are simple to use because users can talk about their requirements and actions in everyday language. The ability to use text and voice, and supporting vernacular, slang, and poor grammar and spelling, makes the conversational interface non-discriminatory and accessible regardless of gender, age, or ethnicity, whether stationary or on the move. All segments of society are targeted as beneficiaries. The extension to this successful project is focused on building the commercial model and the go-to-market strategies and plan for rapid market entry. The new solution can now be demonstrated and will be tested with users, showcased to potential commercial partners, and input will be sought on the commercial model, including pricing. Given the complexity of the market, commercial partners will be identified that can bring the solution to the millions of people in debt in the UK. At least one initial commercial partner will be secured during the extension phase. In addition, market dynamics will be investigated in international markets, including in the US, as there are indications that the handling of people in debt has resonance with approaches in the UK.

"An Artificially Intelligent Conversational Interface (CI) is automated technology permanently on call for users via text or voice interfaces - accessed typically via Instant Messenger channels such as Facebook Messenger and voice channels like Alexa. CIs promise 24/7 automated, frictionless communication between users and businesses to enable quick, convenient task completion. Significant barriers exist for businesses wishing to benefit from commercial strength CIs. Software tools on the market are inadequate for developing commercial-strength CIs. There are also significant challenges collecting CI training data. These difficulties render it practically impossible to build CIs beyond prototype quality. This project's expert consortium aims to make it quick and inexpensive for enterprise development teams without expertise in AI to create sophisticated commercial-strength CIs. The delivered solution removes barriers to market entry, constituting game-changing models that empower enterprises of all sizes internationally. The consortium has deep expertise in machine-learning and commercial messaging integration across a broad business spectrum, and is positioned to design innovative technology that addresses CI creation challenges. Core elements of a Language Classifier Marketplace will be delivered in this project, enabling companies to use prebuilt, instantly configurable machine-learning language classifier templates to create high-quality CIs without machine-learning expertise."

Action Artificial Intelligence has built an artificially intelligent shopping assistant that works within messaging platforms, for example, Facebook Messenger, Whatsapp or via SMS. We call this Conversational Commerce. The solution enables people to order products simply by texting in their natural language what they would like to do – the process is 10x faster than anything in the market. The ordering process is intuitive with no apps to download and no need to learn any keywords or complex systems – the user simply says what they’d like and can place an order or seek information in seconds. The approach is perfect whether ordering a single item, adding an item to a shopping basket, or re-ordering a whole basket of goods.