There’s absolutely no two opinions that whichever country or region masters Artificial Intelligence (AI) will dominate business and pretty much everything else for a long time to come.

That’s absolutely why China has been working so hard on AI. China Artifical Intelligence Plan is a hot topic through and through. That, despite the fact that AI has the potential to cause job losses (or rather, a radical shift in job profiles).  And job losses is something that China as the world’s most populous country can’t afford.

A lot has been said and analyzed about the way China has been trying to leverage AI. In fact, China’s strong social credit system is being built using significantly the tools that only AI can give.

And in all that din, even industry observers almost forgot Europe.

So what’s the status of Artificial Intelligence in Europe? What is the EU Artificial Intelligence strategy?

Today, we analyze Microsoft’s report on AI in Europe.

Background of the AI report

Some time back, Microsoft commissioned Ernst & Young (E&Y) to carry out a study on the status of AI in Europe with Sweden as a special focus. This report, formally titled “Artificial Intelligence in Europe How 277 Major Companies Benefit from AI Sweden Outlook for 2019 and Beyond”, is the summary of the findings of the survey.

The study was carried out in 15 countries of EU and 277 companies participated in the study. Most findings are classified in two parts: 15 European markets and Sweden.

It’s best to treat this report as a study that shows you the overall picture but isn’t highly accurate at the granular level. Also, it wouldn’t be a good idea to treat the findings as truly representational of the entire European Union (EU).

Note: There are other reports from Microsoft too that center around different countries of Europe. The one we are discussing is focused on Sweden.

How perfect is the report?

Like almost all surveys, this survey too suffers from biases.

What’s more, there are certain interpretations in the study that not everyone would agree with; for instance, it calls 44% a ‘majority’ whereas common sense requires you call something a ‘majority’ only when the percentage crosses the half-way mark, i.e. 50%

But that doesn’t mean the report is of no value – it provides great insights into the state of affairs of artificial intelligence in the countries surveyed.

It’s a wonderful peep into what goes inside the participating companies, what’s their stand as regards AI, how equipped they are, what all things are at stake, what level of preparedness the companies are, how these companies view the various challenges and opportunities ahead in the light of AI and much more.

The structure of report on Artificial Intelligence in Europe

The report is divided into five major sections excluding the preface.

Section 1

The first section begins by offering an executive summary of the findings. It lists out the participating companies and the research methodology. It moves on to defining what all technologies are included in the study and concludes with an overview of investments in the field of AI in Europe.

Section 2

The second section deals with the role of AI in European markets. It begins by showing at what level of the participating companies is the AI dialogue taking place.

Next, it explores the maturity and preparedness levels of AI within these companies based on what stage these companies are in their pursuit of building a competitive advantage through AI.

Finally, it concludes with laying out where AI is deployed across the companies.

Section 3

The third section starts off by spelling out the expectations the companies have from AI over the next 5 years and how closely those expectations are related to the core business of the companies today.

The next questions posed are key: what is a good framework to milk the benefits of AI and what are the sector-wise benefits from AI. The last part talks about the risks involved with AI.

Section 4

The fourth section defines exactly eight competencies companies would need to really leverage the true potential of AI. Each of the competencies were discuses on the basis of how the companies view their own readiness with respect to these competencies.

Section 5

The fifth and the final section is the shortest – it analyzes how the companies can take AI further.

Key findings of the Microsoft report on AI Europe

Here are the top 7 findings of the Microsoft’s report on AI in Europe:

Importance

1.Data: A total of 71% companies reported that Artificial Intelligence is an important topic at the executive management level (or C-level). Against this, 28% of the companies reported that the AI was an important topic at the non-managerial or employee level.

Interpretation: AI is still largely top-down rather than bottom-up. A great deal more people at the top than at the junior level believe AI is important.

Distribution

2. Data: The UK, France, and Germany have attracted 87% of investment in AI companies over the past decade.

Interpretation: The AI scene in Europe is nowhere close to even growth when you see 3 of the countries covered attracting nearly 9 out every 10 dollar invested.

Investments

3. Data: Private Equity and Venture Capitals (PE & VC) account for 75% of the total investment that has poured in over the past ten years.

Interpretation: One, AI is a high risk high return business, since more PE & VC firms than established corporates are investing in AI. Two, these established corporates, at this stage, don’t believe investing in AI must be their top-priority.

Status

4. Data: Of the companies that responded, 4% claimed (in “At a Glance”) they were at an advanced stage in AI, meaning that for these companies, Artificial Intelligence was “contributing to many processes and …. enabling quite advanced tasks”. The same number was 45% for Sweden.

Interpretation: Sweden seems to be doing a far better job at AI. That’s not surprising, since TechCrunch calls it the technological superstar of the North. The only question, however, is how consistent is the interpretation of the term ‘Advanced’ in different countries.

Deployment

5. Data: AI is deployed the most in the IT department (47%), while it’s least deployed in general management (4%) and HR (7%). Deployment in commercial activities (Sales, Marketing, Customer Services) is around 20%.

Interpretation: The relatively low deployment in Sales (19%), Marketing (22%) and Customer Services (24%) is a surprise, considering that chatbots have been a sort of rage all through 2018.

Impact

6. Data: Of the surveyed companies, 81% believe AI will have a high or significant impact on their industry over the next five years.

Interpretation: This is well corroborated by another data in the report: 21% of the companies believe AI is not important or only slightly important among their digital priorities.

Usage

7. Data: 74% of the companies surveyed expect to use AI to predict things about their business.  

Interpretation: As we earlier noted, less than 1 in 5 (19%) companies are deploying AI in Sales. So here you have a paradox: On the one hand, you have about 3 out of every 4 (74%) companies expecting to use AI to predict things. On the other, your deployment in predicting and finding more about future sales trends is at a low 19%. Simply put, companies are using AI predictions a lot more for other things than for predicting sales trends.  

Some challenges to the Microsoft report

The report, while carefully researched and put together, certainly has its set of drawbacks. Some data could have been presented better while there are some interpretations that you’d not fully agree with.

It’s important, therefore, to look at the report with a healthy pinch of criticism.

Here are the four major weaknesses of the report:

Update:

We reached out to E & Y as well as Microsoft regarding the following criticisms, using the email addresses provided at the end of the report.

Microsoft didn’t respond; E & Y did. We’ve included their side of the story wherever relevant.

1. The way companies are chosen for the survey is far from perfect.

A pie-chart on page 15 of the report shows the number of online surveyed companies per country. While most other countries have about 20 companies each in the survey, UK, France and Germany have a total of 15 companies.

Page 21 of the same report mentions these three countries have attracted 87% of investment in AI companies over the past decade.

Effectively, countries attracting 87% of the total investment get less than 5.6% of the total representation in the study.

Question unanswered: How can you explain why countries with an overwhelmingly high proportion of investment are so heavily under-represented?

E & Y wrote to us: “…the focus of these reports was on geographic western European countries that do not have recent reports. There has already been more focus on UK, France, and Germany, so, one of our goals was to see what is going on in the other countries.”

Our view: In that case, may be it would have been best to drop the top three performing countries. Including these three countries and keeping them under-represented doesn’t appear to contribute any additional value to the report.

2. The findings could have been worded better.

The bar-chart on page 21 says TMT (Technology, Media, Telecom) is most active, just behind PE & VC when it comes to investment.

However, a closer look reveals a significantly different picture.

The value per deal of PE & VC or TMT $7.2Mn/deal and $8.3Mn/deal is nowhere close to the top. Life Science, for instance, booked 12.09Mn/deal, nearly 50% more than that of TMT.         

At $30.6Mn/deal, the average investment per deal is the highest with Infrastructure.

Question: What are the grounds of labeling TMT most active, just after PE & VC, when the value per deal is much higher in Infrastructure, Industrial Products and Life Sciences?

E & Y wrote to us: “The number of deals is a measure of activity in the market, hence the ”most active” designator”

Our view: On one part, we stand corrected; earlier we had mentioned this pertained to Sweden. That’s wrong, it covers all the 15 countries.

We appreciate E&Y’s view that number of deals is a measure of activity in the market. On our part, we strongly believe that emerging technologies like AI are driven heavily by talent and capital (putting the mouth where the money is).

In that context, we prefer to believe it’s the value of deal rather than the number of deals that might be a better indicator.

Our view, however, does not refute that the fact number of deals is a strong measure of how widespread the overall technology push is.

3.   Since when did 44% become a majority?

On page 28 of the report, it says “The majority consider AI to be important” pointing to a small graph below. The number this caption points to is 44%. Traditionally, one uses the term majority only when the percentage is more than 50%.

Question: Any specific reasons the report calls 44% a “majority”?

E & Y wrote to us: “For the 15 European markets, 44% + 28% + 7% (= 79%) consider it, ”important”, ”between important and most important” and ”most important”, which is a majority. By putting the marker at ”important” it signifies that as the threshold above which there is a majority (79%)

Our view: By the same logic, 44% + 12% + 9% = 65% consider it “Not important”, between “Not important” and “important”, and “important”, which is a majority too. That could have been worded as 65% (or majority) believe AI is “important” to “not important”.

That said, we believe the 44% that believes AI is “important” is a good sign going forward.

4.  The interpretation comes close to being self-contradictory.

On page 61 of the report, it says, “A large proportion of companies consider themselves to have limited or no AI Leadership competency”.

Well, there’s another way of looking at the same data.

And the interpretation would come out exactly the opposite.

In the same graph a total of 64% (32% + 23% + 9%) of the companies rate themselves as Moderately (or higher) competent.

Question: What could explain the basis of saying a large number of companies consider themselves to have limited or no AI competency, when the converse is more accurate?

E & Y wrote to us: “In that same graph, 34% rate themselves as ”not competent” or ”between not competent and moderately competent”, which is a large portion of the distribution.

Our view: Agree – 34% can qualify as a large portion, especially when the other competency ratings are lower too, as E & Y further pointed out to us. And more in spirit than in letter.

That stand, however, isn’t consistent with the stand the report took in the earlier point (#3, right above this one). There, they chose to go with the upper-bound, or an optimistic POV while here they chose the lower-bound.

The summing up

The report is quite both exhaustive and educative. If the purpose was to take a general overview of the AI scene, it’d be accurate to say the report has achieved it.

We do concur with the general view that reports, especially on emerging technologies, are fraught with risks since there are too many grey areas.

In conclusion, we thank the E&Y team for having engaged in a dialogue following the publication of this post and sharing their side of the story.