Earlier this month, it was announced that Dutch companies leading the way in applying artificial intelligence compared to other European companies. Indeed, many companies in our country appear to be practical-minded, cooperate a lot with universities and are focused on making work easier by using technology to support the employee. Even more than a third of the nearly three hundred companies surveyed are said to be working on, or about to start, AI projects, while three-quarters are already machine learning (a form of AI) commitment.

This coverage reminded me of a quote by Thomas Watson Jr. (former CEO of IBM and son of the company's founder) who once said, “Our machines should be nothing more than tools for extending the powers of the human beings who use them.” In my opinion, this statement still holds true today. It is a vision that you should be able to plot over and over again for the future. You see the core message popping up everywhere now that AI is a hot topic. You also see it in IBM's AI strategy (a forerunner in this field), because the ultimate goal is still to provide tools so that people can make better choices.

A piece of intelligent software

Will AI replace the web analyst?

I heard from someone the other day that the answer to this type of journalistic headline is always invariably ‘no. They also call this the Betteridge's law of headlines (named after an English technology journalist with this last name). There is, of course, a reason why these, I'll call it ‘what-if questions,’ are asked. Not only does it attract attention, but it also, at least very briefly, prompts thought. So? Did I manage to do that?

Have you ever wondered if the work you do as a Web analyst could also be done by a piece of intelligent software? Perhaps it would be better to put it like this: ‘Have you also ever wondered if it wouldn't be better if you could be supported by software in performing certain analysis tasks?’ Hopefully the answer is now ‘Yes’.

An example of such a piece of intelligence is the Insights option within Google Analytics, which is also called Analytics Intelligence (called AI for short!). Instead of creating the report yourself, the interface lets you ask simple questions by entering them - for now only - as text. The text-string appears to be a kind of query that Google Analytics eventually uses to look up the result and share that with you. These can be questions such as: “Why is my conversion rate down?”. But actually, you might want to ask Google the question, “What advice do you give me to improve the conversion rate of my non-paid channels?”. We don't always have to line the Google coffers by purchasing AdWords, of course. Oh well, I'm sure you get my point. These kinds of questions are probably much harder to answer. Developments on this may soon follow, but at the moment this is not available in Google Analytics.

The virtual analyst

That other big player in this market, Adobe, announced late last month that it has spent three years developing a virtual analyst whose goal is to bring qualitative insights to the user across Adobe's clouds (Creative, Experience, Marketing, among others). This virtual analyst, which listens to the name AI Assist, uses Adobe Sensei, Adobe's AI and ML framework.

So with this, the virtual assistant is also coming to Adobe Analytics soon. A virtual assistant for the web analyst. So it can be done? This announcement not only indicates that Adobe continues to actively work on rapid development of the analytics platform, but also embraces rapid developments in the market and is also guided by the efficiency that AI can bring to further support companies, where it is really needed.

The new virtual analyst was developed based on Adobe Sensei and a set of several solutions previously introduced within Adobe Analytics (Anomaly Detection from 2013, Contribution Analysis from 2015 and Intelligent Alerts from 2016).

For the introduction of AI Assist, Adobe states in the press release that the new virtual analyst has the following three functionalities in it:

• Making the unknown visible: The virtual analyst constantly searches and analyzes the company's data and prioritizes interesting changes. Deep learning models give the virtual analyst access to every data point in every customer interaction, ranging from how long a consumer visits a website, to switching between app and website. These unknown patterns not only offer new insights, but also reveal to companies what they have not looked for before.
• Maximum relevance: The virtual analyst proactively prioritizes data analytics based on both business and user contexts (both using real-time activity and historical data) without the user having to launch the system itself. Machine learning algorithms also ensure that recommendations are bundled. For example, a spike in sales, orders and conversions all related to the same event are aggregated into one meaningful insight.
• Adaptive learning: Gradually, the virtual analyst takes into account users‘ preferences and consumption patterns to provide more intuitive and relevant insights. It also analyzes the behavior of all other users within the organization to find corresponding individuals and then use this to make recommendations for further personalization. The system offers the ability to use ’like‘ or ’dislike" recommendations that reinforce the machine learning model and make the virtual analyst increasingly intelligent over time.

The above three points are all additions with which Adobe actually also unerringly exposes the underlying problem that many companies are currently struggling with. Not to get bogged down in terms like ‘big data,’ it is of course a fact that the mountain of (web) data to be analyzed is getting bigger and bigger and there still seems to be too few qualified personnel to analyze all that data and actually extract useful insights from it.

“We no longer have to be embarrassed about getting help with our daily analysis work.”

Therefore, applying AI in analytics software seems like a good step to me. It sounds like a perfect practical example of how the fourth industrial revolution is really starting to take shape. We no longer have to be embarrassed about getting help doing our daily analytics work. After all, we no longer think about how many electronic devices we use in our daily lives, do we?

Voice and A/B testing

Another development from Adobe that appealed to me was in the area of voice. With an increase in the use of smart voice assistants such as Amazon's Alexa and Google Home, its real breakthrough is probably not long in coming. Adobe is capitalizing on this by introducing analytics and AI applications within Adobe Analytics tools to help companies extract more value from their voice-based data. It is also considering experimenting with different voices, for example.

In the area of A/B testing, there has also been some development. Following the Auto-Target announced last year, Adobe will integrate AI there, too, to ensure that a user - for example, a marketer - can choose the best algorithm to offer the visitor a tailored and totally personalized experience, based on an individual profile and the behavior of other visitors with similar profiles. By the way, not only does that experience need to improve, but of course those conversions need to go up as well.

AI within web analytics and CRO

Artificial Intelligence within web analytics and CRO (Conversion Ratio Optimization) is a fairly new but very exciting development to follow. It's about our future, isn't it? Where there are often a lot of question marks about AI in general and AI applications in certain areas in particular, I still think it is a welcome development within our field.

Do you have any doubts about AI yourself? I personally always trust the following statement, “If the machine can beat humans, the human+machine couple will always beat any machine.”