These industries include retail, banking, law and healthcare – and chatbot developers are only just getting started.
Today, chatbots are opening doors to the way we search for, and acquire, information. With their ability to integrate with apps such as Facebook Messenger, Kik, WhatsApp and Slack, chatbots provide answers, advice and information without the user ever having to leave the app.
To help the advance of new technologies like chatbots, R&D (research and development) projects being undertaken can qualify for the UK government’s R&D tax credits incentive.
Under this, the staff costs, software, utilities and materials dedicated to the R&D of chatbots can be used to determine the value of the tax credit.
The type of work that could qualify for R&D tax credits include:
- Implementing chatbots into new industries.
- Developing algorithms and chatbot intelligence.
- Finding new ways to use existing chatbot technology.
- Improving chatbot user experience.
Worth up to 33p for every £1 spent, ForrestBrown helps companies performing R&D optimise their claims. These funds are highly valuable to SMEs, often helping them invest in further R&D of technologies like chatbots and AI.
What is a chatbot?
A chatbot is a computer program designed to talk to a person in a genuine, conversational way. A chatbot interacts with the user so realistically, they will feel like they are directly conversing with another human.
Some online chatbots such as Siri and Google Now take the form of a virtual assistant, making tasks simple and easy to achieve. This includes shortening the amount of time the user spends seeking answers to a question or finding a solution to a problem.
Chatbots function by using AI (Artificial Intelligence) and, specifically, NLP (Natural Language Processing). As an element of AI, NLP gives a bot the ability to understand human language through observing patterns in data. The bot can then recognise precisely what the user means, the context it is in, and provide human-like responses.
The ability to self-learn is part of the driving force behind the rapid growth and development of chatbots.
History of chatbots
The Turing Test, devised by Alan Turing in 1950, was established to determine the conditions of AI.
For a bot to pass the Turing Test, it must replicate the conversation of a human being and convince the user that they are speaking to another person.
Let’s take a look at the key milestones of chatbots:
1966 – ELIZA: An NLP bot built to demonstrate how man and machine interact with each other.
1972 – Parry: Created to simulate a person with schizophrenia.
1988 – Jabberwacky: The first real attempt at creating AI from human interaction.
1992 – Dr Sbaitso: An AI that used a digital voice.
2001 – Smarterchild: A pioneering bot used across SMS networks, it is hailed as the predecessor to Apple’s Siri and Samsung’s S Voice.
2010 – Siri: Integrated with Apple’s iOS, Siri answers questions and performs web search functions.
2012 – Google Now – Another AI bot, Google Now makes recommendations and performs web-based services.
2016 – Bots for Messenger: Facebook’s Messenger feature enables developers to build bots that directly interact with Messenger users.
Uses for chatbots
- Information updates
- Online shopping
- Customer service
One chatbot used to organise the daily life of its user is Google Assistant. And whilst this bot keeps track of events and calendar dates, it is also capable of sustaining conversation and giving tailored answers to specific questions.
By utilising AI, the more Google Assistant is used, the more it learns about the user, their personal preferences and interests. This means that over time it will provide increasingly accurate and reliable information.
But not all chatbots are focused on casual, human-like conversations. Bots on Facebook, Slack and WeChat are focused on providing solutions to questions and assisting with the search for information.
An example of this is Poncho. It works within apps such as Facebook Messenger, sending tailored weather forecast information, giving users real-time updates of the weather. This saves the user time, as they receive updates whilst in the app and do not have to go elsewhere to retrieve weather information.
Chatbots for helping learn language
Chatbots for learning languages are taking shape too. One example is Duolingo. This chatbot platform, powered by AI and machine learning, is the first bot that allows people to instantly use chatbots to learn languages. Having gained 150 million users since its inception, it provides users with 5 to 20 minutes of language training per day.
Duolingo works by observing how its users learn and what teaching methods they respond well to. Through learning the actions of its users, it provides tailored teaching methods. Therefore, learning a new language through Duolingo becomes easier the more it is used.
Chatbots for ecommerce
A company making strides in the development of chatbots for ecommerce is Inbenta, with their creation of the InbentaBot. This is a virtual chatbot that can multitask and perform searches and transactions – freeing up time and capacity for staff.
The InbentaBot organises every product available in a company’s inventory into colours, sizes, prices, etc. By categorising the products, it can then present the most appropriate ones to the customer that match up with their search query.
When Franklin Planner, the producer of a paper-based time management tool, implemented the InbentaBot into their system, it improved their ecommerce search – resulting in a 20% conversion rate from search to cart and a 35% click to cart efficiency.
Chatbots for business are increasingly common, one survey suggested 80% of companies would like their own chatbot by 2020. Chatbots for retail companies are now being used too, such as high street clothing brand H&M, who are using bots on the messaging platform Kik to sell their products.
Customers can interact with H&M’s online chatbot and choose between outfits the bot presents them with. This allows the bot to acquire information about their clothing tastes, presenting them with increasingly suitable outfits.
Chatbot customer service
Chatbot customer service is becoming ever more present due to their ability to solve problems and provide useful tips. This is facilitating RTIM – Real-time Interaction Management.
Helpshift develops virtual chatbots that allow customers to serve themselves without having to be connected to a customer service representative. These online chatbots engage with customers directly, offering personalised support to their questions.
Courier service company Shyp implemented the Helpshift platform to their system – with a chatbot for customer support proving to be incredibly beneficial. Customer service had previously been a major cost to Shyp, but Helpshift cut these costs by 25%. Helpshift’s platform transformed the customer experience for Shyp’s customers, whilst also being a valuable money-saving tool to Shyp.
Barclays Africa is using chatbots to answer basic customer questions and provide immediate responses. This frees up staff to spend time on more complex enquiries. This marks a transformation in how AI can provide a seamless interactive experience and fully understand customers’ needs.
Chatbots in banking
An innovative use of AI in banking is being applied by Atom Bank. As the first UK bank to launch via a smartphone app, Atom Bank has developed a groundbreaking banking service by using voice and face biometrics for the customer to log in – by analysing a selfie or a voice clip of the user to prove their authenticity.
Moreover, Atom is offering customer support via a machine learning chatbot. This technology identifies the context of a customer’s query, answering appropriately and learning from experience. Atom is creating a more intuitive way for customers to interact with their bank and to manage their money in a stress-free way.
With its digital business model, Atom also has reduced overheads by not having physical branches, giving its customers better interest rates and lower costs. This is paving the way for how mainstream banks operate in the future and how they provide support and banking advice to their customers.
Today, chatbots can tailor a company’s products and services to their customers’ specific needs – all through machine learning and AI. Through collecting specific information on the user, marketing content can be delivered to consumers by a chatbot.
Bots collect customer information and tailor advertisements and marketing content to them, supporting them in their product search. This coming-together of technology and marketing is a sector of huge growth and opportunity.
Growthbot is a chatbot designed for marketing and sales professionals. Able to work within Facebook Messenger, Growthbot provides an ‘insight into the insights’, observing marketing systems such as Google Analytics and Hubspot, talking to the platforms that a company is already working with.
Growthbot works by its ability to answer questions relating to your target market. For example, if you sell software to SMEs and are seeking potential customers, you can ask Growthbot to “Show the SMEs in Bristol”. It will then display a list of these companies.
Integrating with Slack, and Twitter as well as Facebook Messenger, Growthbot has developed a way for market research data to be at the fingertips of the user.
Chatbots for legal support
When it comes to legal advice, a chatbot lawyer may sound like a peculiar form of sci-fi fantasy – but they are now being applied to real-world legal cases.
DoNotPay is hailed as the world’s first robot lawyer, with a chatbot conversational interface. It uses high-level AI to offer legal advice and its track record includes the overturning of 160,000 parking fines through giving free legal aid.
Able to work with Facebook Messenger, DoNotPay helps refugees in the US and Canada and helps those in the UK apply for asylum. This could be a nod to the future of legal support and how chatbots can offer low-cost, reliable advice.
Chatbots for friendship
Xiaoice, a Microsoft chatbot, is frequently used in the Chinese community. With over 20 million users in China, Xiaoice provides an emotional outlet for many due to its listening skills, sense of humour and compassion.
This is can help with mental health, with chatbots now becoming an emotional outlet for many – giving someone a person, or so it seems, to talk to.
Mitsuku – the winner of the distinguished 2013 and 2016 Loebner Prize – is a virtual chatbot that learns by experience. Similar friendship chatbots that use AI and machine learning are Cleverbot and Eviebot. The more these chatbots are interacted with, the more intelligent and humanlike they will become.
Chatbots for healthcare
Pushing the boundaries in chatbot healthcare is Your.MD – a bot that answers patient’s questions by giving personal and trustworthy medical advice.
Harnessing AI, it has the world’s largest Map of Conditions, and is able to identify over 1.4 million medical conditions.
Your.MD works by the patient entering information about their condition. The bot then asks a series of questions, finally suggesting conditions based on the symptoms described by the patient.
This innovative use of AI could be a source of healthcare in the future, saving medical professionals valuable time and healthcare providers money.
Chatbots use a range of technologies to function – and with their AI and ability to assist users, their ascension makes perfect sense. Their quick responses and progressively humanlike features indicate just advanced they are becoming.
AI and chatbots
AI is an integral part of chatbots, giving them the ability to not just interact with people, but have engaging, genuine conversations.
Machine learning and chatbots
Machine learning algorithms enable computers to learn through interaction and pick up traits by finding patterns in data and instructions.
After one conversation with a human, a machine learning bot picks up patterns in the interaction, understanding the language used, rather than simple commands.
Pandorabots is a web service that facilitates the construction of bots and their application to other platforms. So far, over 285,000 chatbots have been created through Pandorabots.
Chatbot developers use an API (Application Programming Interface) to build and develop bots. An API is a set of functions that allow software programs to communicate with each other, with the API acting as an interface, managing how these pieces of software interact.
Major APIs used by chatbot developers include API.ai, Wit.ai, MS Bot Framework and Motion AI. There are also online communities dedicated to the development of chatbots – such as those building a slack chatbot.
Some problems with chatbots are based on their rushed production, with developers skipping user-testing phases. This has left the market littered with bots that don’t perform to their full potential – they are clunky and rigid, with pre-programmed answers.
A major issue with Facebook Messenger chatbots is that it is often unclear how to get them started. In order to overcome this obstacle, chatbot developers have been developing a menu that allows multiple items, giving users a new way to interact with bots. This new menu displays all the bot’s capabilities on an interface, meaning easier access to its capabilities.
Chatbots have also been known to go haywire and stir up controversy – a notable case being Microsoft’s Tay.
Tay was designed to interact with people via Twitter to improve its conversational skills through machine learning. But within a few hours, Twitter users were bombarding Tay with misogynistic, hateful and racist tweets. And because Tay was a machine learning bot, it absorbed these statements and begun spouting obscenities. In less than 24 hours of being online, Tay was deactivated from Twitter.
Another uncertainty presented by chatbots is if (and how) they store a user’s personal information. With chatbots now being in the domain of private communications, the Electronic Frontier Foundation has developed the Secure Messaging Scorecard. This evaluates apps and tools based on specific criteria to check they are secure. Companies such as DoNotPay are also addressing these issues head on by destroying data from its servers within ten minutes of using the bot.
One of the biggest technical challenges that chatbots pose is how they decipher ambiguous questions. These are often mistyped, with poor grammar and punctuation. Inbenta has overcome this challenge however, by taking vague enquiries to the next level. It has developed the InbentaBot to understand the context of the questions being asked – all through a highly-sophisticated spelling algorithm.
Inbenta has its own database of English words and can detect the most likely word combinations. E.g. it can detect if the word “well” is mistyped because the question it is in does not make sense. This enables the bot to find the right answers to incorrectly typed sentences, a significant step forward in chatbots’ ability to detect human error.
Chatbots and R&D
The chatbot and AI industry is a hotbed for R&D, with groundbreaking technologies being used to overcome challenges and present new solutions to existing problems.
At ForrestBrown, our tech-savvy tax advisers specialise in identifying the areas of R&D to fully optimise your claim, enabling you to make the most of this incentive. It’s an award-winning approach that saw us judged ‘best independent consultancy firm’ at the 2016 Taxation Awards.
If you are a company using chatbots, developing them, or working on a project to overcome technical uncertainties within the world of AI, contact ForrestBrown on 0117 926 9022. Speak to a member of our tax team and we can help you see if you qualify for R&D tax credits.