Introduction to the new era of intelligent automation – Capgemini
The event launched with Jaakko Lehtinen, Capgemini’s Head of Nordic Intelligent Automation, briefing the audience about automation trends in 2019. According to him, customers in the Nordics are increasingly looking to move automation platforms and operations in cloud, and many of them also moving all the RPA related operations work to suppliers.
At least automation visionaries are starting to see the real business value of automation, and business cases are not only concentrating on working time savings: according to Jaakko, intelligent automation does not only refer to a wisely selected set of automation technologies. It also refers to a right kind of attitude and courage to do and see automation opportunities in a new way. Jaakko also told that artificial intelligence is raising its head with an increasing amount of real production deployments taking place at the customers.
Real-life chatbots – what is possible and where to start? – AlphaBlues
The amount of conversations conducted by AI is increasing rapidly. Indrek Vainu, the CEO of Estonian start-up AlphaBlues building AI-based chatbots, revealed that the number of conversations their chatbots have on a monthly basis is 250 000+. And there is no reason why each company shouldn’t start thinking of ways how automate their communication because the number of possibilities is huge. With the help of data, AI chatbots can be programmed to identify certain parameters such as gender, location, device, and combine these pieces of information to perform different activities according to these parameters. For example, a chatbot can be used for increasing sales by recommending men to buy something nice for the ladies when the Women’s Day is approaching and wish a nice and relaxing Women’s Day for the female visitors.
Indrek also answered to the burning question of how difficult it is to build a chatbot for Finnish language. He agreed that the language is super difficult, but when it comes to building a chatbot in Finnish, or any other language, the complexity of the language isn’t the factor that makes it complicated. According to Indrek, the complicated part is to understand human behaviour and learn the different ways how your clients are behaving and asking questions from the bot. For any intent, you can have a lot of different ways of writing it, so the best thing is to start with one language, let the AI learn from the various communication styles, and then gradually add other languages.
Tips for launching a chatbot project in your organisation:
- Pick a use case together with the business and define what you want to achieve with the chatbot
- Learn what tools are available and how these suits your needs – can you use existing AI or do you need to build one yourself
- Decide where you want to launch your chatbot. Although we use social media a lot, website is still the place where people go when they have a problem.
- Pick a language (if your users use Finnish, then there’s no reason to start building a chatbot in English). First learn to understand the customers and train the bot (silent live). According to Indrek, “bot is kind of like wine, it gets better on time”.
- Go for the low hanging fruits and start from the 20% of questions that bring 80% of the volume.
- Have a human fallback option to have a complete customer experience. It is also always good to mention it to your clients if they are having a discussion with a bot, e.g. “Hi, I am Nemo your virtual assistant.”
Handling unstructured data for enterprise transformation – WorkFusion
There is a huge number of transactional data that companies process daily and the number of sources has great variety too. Typical sources of data are emails, customer channels, documents, websites, data feeds and databases. In addition to source, the format of data also varies a lot – it could be anything from a paper document to electronic file or picture. Processing and combining this data require a lot of work and can be very manual. Or in other cases, partially automated.
Craig Sumner, Senior Solutions Consultant from WorkFusion, described their Everyday AI approach as compared to a Deep AI. According to him, usually 80% of the work with an AI project is collecting and preparing data. He was also referring to a Google study according to which “[o]nly a small fraction of real-world [machine learning] systems [are] composed of [machine learning] code [and] the required surrounding infrastructure is vast and complex”.
In Everyday AI approach, the required machine learning algorithms for identifying unstructured source data are not being developed for each customer separately by Data Scientists, but they are readily available in the platform that is automating work. The classification and structuring of source data are done by these machine learning algorithms that automatically learn by observing provided samples, or by observing manual work examples. Different algorithms are also competing against each other’s, and the best performing model will be utilized in production. Human quality assurance, and human input when automation fails will provide new data for the self-learning algorithms and upkeep the machine learning models’ performance.
Towards an AI first organisation: how can machine learning enhance your business? – Capgemini
As the fourth speaker of the event, Capgemini’s Diego Lopez Gonzalez talked about AI first organisations and what kind of opportunities for growth machine learning brings. According to Diego, now it is a really good time to invest in machine learning. The early adopters have started building value with AI already 5-10 years ago. During this time, we have already seen some successful cases but there are still a lot of companies that are not utilising AI as much as they could. The biggest implementers of AI projects at scale are companies in telecom and retail industries, but for example in the manufacturing and industry sector there is still a lot to be gained.
According to Capgemini’s survey on the companies that have already implemented AI reported to have over 10% boost in sales, operations, and customer satisfaction and engagement. They also reported that they were receiving better analysis and new insight on their business.
How to decide where to make AI investments? A good way to measure and identify potential investment targets is to compare the complexity of an AI project against benefits. According to Diego, the best place to start is in a high benefit and low complexity items. These low hanging fruits are often neglected when deciding where to start with AI although these are the easiest and at the same time most visible ways to start an AI journey.
What defines an AI first organisation then?
- AI is not considered as a tool, but as an enabler.
- AI is a long-term goal and considered as a base for development.
- AI is an integral part of the value creation chain.
- Can easily scale up in complexity and size, relying on AI to keep operational.
Video analytics: what artificial intelligence can do with your videos – even in real time? – Valossa
As a final presentation, Mika Rautiainen, CEO of Valossa presented us the opportunities of AI for video content. As a conclusion, AI can already be made to understand video pretty much the same way as humans and it has a lot of potential for many industries.
If you look at general video productions, AI can help you recognize and analyse the content (e.g. people, objects, scenes and sentiments), it can help you shorten the videos, make teasers for movies, or collect the highlights of a certain episode. It can also generate a description of the episode or pinpoint known brands or people from live-stream videos.
During his presentation Mika showed us two impressive demos how Valossa tool had been used; the other one was a full analysis of a TV series episode and the other one was fully AI constructed trailer from a complete movie – both of these examples seemed like they were crafted by a human instead of an automatic AI tool.
Entertainment is not the only industry or area where the use cases of video analytics are increasing. Some examples are:
- Automotive: driving behaviour and safety related analysis and automation
- Consumer research: understanding where consumers’ eyes are pointing and what they are interested in stores
- Safety and security: identifying misbehaviour in CCTV feeds
- Emotional feedback: analysing attendee expressions to understand how they react to a presentation, piece of information or other perceived things
- Industrial knowledge:
All in all, the Digital Dawn event brought a lot of new insight on what is already possible with today’s technology readily available for customers. These should help us all in finding the courage to try new things and move along in the journey of becoming an AI first organisation.