FOMO on AI? Let’s identify your low-hanging fruits @TOA Berlin 2016

Natural Interfaces and Intelligent Assistants @Latitude59 2016

Natural interfaces and Intelligent Assistants , panel at Latitude59 – 2016

Smarter Future: A.I. is here @TNW Europe 2016

TNW Europe 2016 | Smarter Future: AI is here

Deep Dive on Artificial Intelligence @Seedcamp

Podcast with Dr. Edward Challis, founder and CEO of Seedcamp-backed Re:Infer and Christoph Auer-Welsbach, IBM Watson Partner Innovation, joined Carlos to discuss Artificial Intelligence, Machine Learning, and whether or not we will have to worry about humanity’s fate in the future.


The Cognitive Era – How A.I. transforms products + services @TheFamily

Can our society cope with Artificial Intelligence? @StartupDay Stockholm 2016

Have Humans Gone Obsolete?

The question of human existence -whether as a species in general or just in the working world- has generated quite some concern around the world. Famous technologists and successful entrepreneurs such as Stephen Hawking, Elon Musk or Jaan Tallinn have responded with organisations like Future of Life Institute and OpenAI. They strive to protect human life from intelligent machines taking over the world.

Thanks to Paul O’Connell, founder and organiser of the Uprise Festival in Amsterdam, I had the opportunity to dive a bit deeper into this issue. As part of the Festival -already in its 3rd edition and coming to Dublin later this year- I was able to bring to the audience a panel of guest speakers with the goal of answering one bold question: are humans still relevant?

Together with David Vismans, chief product officer at the travel platform, Pieter Boon, partner at the data science company Xomnia, and Robert Verwaayen, partner at the venture capital firm Keen Venture Partners, we focused on the terminology used when talking about artificial intelligence, the relevance of technologies like machine learning in today’s businesses, and their impact on the future job landscape.


The Terminology

Artificial Intelligence is a term that was coined in the mid-1950s by John McCarthy describing the science and engineering of making machines intelligent. During the past 60+ years, thousands of scientists and engineers have been working on creating these intelligent machines. “Already in the 60s, a robot used image recognition to analyse the movement of an object to the left or right, but due to a lack of processing power it took up to 1 hour to calculate the next step. Today technology has caught up to make these machines powerful”, Verwaayen points out.

In the past years, we have seen a rising number of applications “using a very clear set of techniques to predict user behaviour and recommend users’ actions”, Vismans explains. “But today AI has become a loaded term, often misused for a variety of technical utilities to create intelligent systems.” In the engineering world, “data scientists still prefer to use the term machine learning”, says Boon.

At Keen Venture Partners, Verwaayen uses another term: distributed Intelligence. This takes a holistic approach combining all technologies around perception technologies, machine learning, big data analytics, and more. He classifies the various technologies in the following 3 areas:

  • IoT: the eyes and the ears of the Internet and machines,
  • Robots: the hands and feet to interact with our physical world, and
  • AI: interpreting the big sensor data available and acting in a smart way through the machines.

The term that seems most appropriate to me is machine intelligence, since I still doubt that a machine will ever replace human intelligence. A machine will always ‘think’ or ‘act’ differently, except when humans program it to do specific things in a human way.

The closest we can get to AI is with intelligent machines imitating human intelligence

Businesses of Today

With regards to the impact of AI on today’s businesses and product development, it seems that there is a pretty strong opinion that it is -or at least will be- huge. Vismans elaborates: “Machine learning has a fundamental impact on the quality of customer experience. In the near future product managers and engineers will need to be much more aware of where machine learning can be used to add product value. Over the next 5 years, people with this skill set will be very high in demand based on the enormous potential they can unlock for businesses.”

When we look at companies today, machine learning is typically part of a separate department, usually focused on big data & analytics. What we’ll see in the future is that the people working with big data & analytics will become an integral part of each business unit and project group, supplying their expertise to support each team directly with applying these techniques into the daily product development and customer service processes.

This means that machine learning will become a tool integrated at all levels, strategically as well as operationally

Jobs of Tomorrow

The intelligent machines of today can still be considered a weak form of AI. These are systems that are very good at one specific task. Although some of these seem quite smart (e.g. by winning against humans in complex games like GO), they are far from being strong AI or reaching the level of human intelligence. “We have no need to fear the replacement of jobs by a holistic, strong AI in the near future. For instance, we don’t even know how human creativity works”, Vismans points out.

Boon explains further: “Techniques such as image recognition are used for the purpose of behaviour detection to prevent attacks, analysis of technical issues such as cracks in aircrafts, or to prevent attacks or control a crowd”. All very specific tasks that assist human activity, but that cannot be performed without human supervision. Also, such tasks are limited by the way society handles liability and ethical issues.

In the mid- to long term, it seems undeniable that certain types of jobs will be replaced by intelligent machines

The American Truck Driver Association predicts that more than 8m jobs are at risk with the introduction of self-driving trucks. But also other types of jobs such as loan officers, information clerks, and assistants, as well as retail salespersons are at risk. Verwaayen’s opinion:“Repetitive, process-led jobs will be replaced since even having them outsourced in low cost countries will become too expensive and inefficient. This doesn’t mean massive unemployment, but it means humans will need to ‘skill up’ in the coming 5-10 years. Machine intelligence will move up to affect more complex jobs too, as we see happening already with robot advisors in the finance industry.”

So, are humans still relevant?

In my personal opinion, humans continue to be relevant, and will remain to be so also in the future. We have to adapt our society, economy and mindset in order to cope and -even more so- in order to benefit from this major leap in technology.

By taking the advancements one step at a time, our hands-on experience will help us in dealing with AI. The right guidance will come with the awareness of its effects on humanity. As Roy Bahat, head of the venture capital firm Bloomberg Beta, puts it:

Have more respect for the unknown

Welcome to the World of A.I.

Green lightbulb brain - Green concept

Artificial Intelligence is all too often associated only with futuristic technologies seen in movies or in the news. Yet what many people don’t realize is that technology disruptions have already been influencing our daily lives for more than a decade!

News on Artificial Intelligence and Cognitive Technologies now surrounds us on a daily basis – making these topics more accessible, and not just for the techies among us.

What is A.I. really about?

Artificial Intelligence is the theory and development of computer systems that normally require human intelligence. These days A.I. is also a buzz word that contains any technology achieving intelligent systems.  ‘Cognitive’ technologies – designed to simulate human thought – are organized into Cognitive Systems. They make use of Machine Learning and Natural Language Processing to enable humans to interact more naturally with machines, with the aim of enhancing and scaling human expertise.

The simulation of human thought processes has been implemented in a variety of consumer and business applications that millions – even billions – of people are using on a daily basis. Consider these examples:

  • Search engines like Google analyze user behavior to suggest potentially relevant information;
  • E-commerce stores like Amazon recommend potentially interesting products to customers;
  • Social networks like Facebook suggest friends we might want to add;
  • Apple’s Siri uses Natural Language Processing and Speech Recognition to convert speech into text and make sense of what the user is saying;
  • Even in the medical field, IBM Watson is already seen as one of the world’s best diagnosticians by using machine learning and computer vision, among others technologies.

These days, also startups are using a variety of cognitive technologies like machine learning or speech recognition to offer us, as their target audience, a new set of services. Magic and Operator are two startups that are able to deliver a new user experience by interacting with their users through a single messaging screen, giving users the feeling of engaging with a personal assistant rather than just a ‘regular’ app. Add: Just have a look at Watson’s Application Starter Kit for a Conversational Agent, or Facebook’s announcement on the launch of a bot platform for its Messenger.

The big technology companies like IBM or Facebook offer infrastructure software and hardware as a service. Also startups like PredictionIO (acquired by Salesforce), Diffbot, Nervana, or Fuzzy enable others to make use of this game-changing technology to offer a completely new experience of products and services to users and customers.

The technology has sky rocketed in the past 2 years. Why?

The short answer to this question: humans are limited in the amount of information they can process. And there is a tremendous of information, or data, out there, from emails, photos and videos to posts in various social networks, documents, etc.

While data creation is nothing new, an impressive 90% of the world’s data has been created in the past 2 years alone! Domo, a BI and data visualization company, analyzed that every minute 350,000 tweets are sent, more than 4.2 million posts are liked and 300 hours of videos are uploaded. In this way, we create 2.5 quintillion bytes of data every single day. Today, companies like Facebook have already collected more than 1,000 terabyte of data. This is the equivalent of 1,048,576,000 MB or – to add a bit of nostalgia – 728,177,777 of those good, old Floppy Disks.

Not only does this generate massive amounts of data, but also 80% of this is unstructured and therefore practically unusable for humans. Examples of unstructured data include scientific data (atmospheric data), photos and videos (from traffic and surveillance cameras), company data (documents, emails and logs), and social media data (from platforms like Twitter, YouTube, Flickr).

In order to analyze and make predictions, information must be screened for patterns and anomalies. Cognitive Systems using technologies like Machine Learning and Natural Language Processing are highly capable of screening this information, analyzing it, and putting it into context. This requires computing systems that are not only able to simulate human thought processes, but also to learn independently. That way cognitive technologies can enhance and scale human expertise leading to greater advancements in human history than ever before.

The future is in our hands, isn’t it?

We are currently at the beginning of a revolution. Some call it the robot revolution – following the industrial revolution in the 19th century and the information revolution in the 21st century. Revolutions are typically a sign of big advancements in human history that offer exciting opportunities, but are also typically accompanied by anxiety and skepticism.

Apprehension is understandable when looking at the number of articles about robots and military applications that predict machines – often malicious to mankind – taking over the world. Movies like The Terminator (as an Austrian I am obliged to mention this one) or The Matrix show us a dark future, although always with a happy ending, thanks to Hollywood. In comparison, movies like Back to the Future look into a far brighter future, where technology is used to make life more fun and healthier, teaching us to take the outcome into our own hands. One great advantage of the revolution of today is the Internet enabling us to inform and educate ourselves, communicate with others instantly, and actively take part in such a disruptive time.

One thing that this revolution has in common with its predecessors is the fact that every revolution has an immense impact on our societal systems and ethical principles. If we are not aware of this technology’s impact and do not participate proactively in the coming years to develop it in the right direction, we won’t be able to ensure the kind of future we want for our children.

In the past months, famous technologists like Elon Musk, Stephan Hawking or Bill Gates have set up organizations and projects like the Future of Life Institute and OpenAI, making sure that these technology advancements will help shape our world for the better.

Finishing up with the words of Thomas Watson Jr.:

“Our machines should be nothing more than tools for extending the powers of the human beings who use them.”


Premature scaling is the #1 cause of failure among startups.

I love the enthusiasm founders have for their startups and I totally understand that each of them is eager to grow the hell out of it in order to become successful. But that more and more startups are talking about scaling as an all-encompassing success strategy is really alarming! Why? Since roughly 90% of startups aren’t ready. In most cases there simply isn’t anything there to scale. What’s more, most startups will never reach the stage where they can actually scale. Startup Genome found that 70% of all startups scale prematurely and that 74% of those scaling are failing! So why the hell is it that startups are always talking about scaling??

Let’s take a closer look at the problem of scaling with an example:

A mobile B2B startup raises EUR 500k in seed capital after hitting early traction, an MVP. Now in the venture game, the startup focuses on sales and therefore hires sales people to grow its funnel and spends money on PR campaigns and paid acquisition.

There is little focus on product development or on the business model, so the revenue it should be generating doesn’t flow in. The startup doesn’t fully understand the needs of its customers and has found a distribution, but not one that can actually be scaled.

So, the customer is not satisfied with the product, development cycles increase, the hired sales people are focusing now on customer service and there is no capacity left to handle the requests coming in from the PR campaign.

As a result, the startup shifts back to product development, tries heavily to find a business model since their road map indicates that they should be preparing for their series A round. In addition, their existing investors are putting their foot down, raising concerns about the chosen strategy and requesting to focus on delivering the numbers they want to see.

Sound familiar? Then it’s high time to face the music.


Lesson 1: Scaling isn’t even possible for most startups.

The bulk of knowledge shared these days on this topic – from great articles to whimsical posts – all too often lead founders to believe that they must scale in order to grow their business. Or even worse, lead them to believe that they can scale their business at all.

Whether you’ve already begun scaling or are just thinking about it, lesson 1 is a crucial checkpoint. Scaling may not be possible in the way that your business is handled at the moment. It may not even be desirable for your business.

Scaling is only right for your business when you’ve developed a product or service offer that truly scales, which means that it can generate revenues by consistently serving many customers. A startup must have found a scalable business model: a model that can be easily expanded, repeated or upgraded on demand. This doesn’t require all business processes to be efficient or automated. However, if your business is not able to consistently deliver revenue while catering to many more customers, then you have no business scaling!

A common example is the e-book that can be paid for and downloaded directly from a website. A sudden peak in demand won’t cause a bottleneck in production. Suddenly producing more hardcopy versions, however, is limited by factors like printer availability, resources for packaging and posting, delivery time, etc.


Lesson 2: Scaling is a RESULT of growth, NOT A DRIVER of it.

What if I told you that scaling does not lead to growth? It doesn’t. It’s not even a growth strategy. Does this surprise you?

The scalable business model presents a way of organizing business operations so that the startup can deliver its product or service quickly and effectively, regardless of whether demand is high or low. A business grows when it can create and cater to increasing demand consistently at every step of operations: development, distribution, marketing, production/service and generating a return for the value delivered. This can happen with or without a scalable business model. A scalable model, once in place, simply means that business operations will expand (or scale) more easily as the business grows.


Lesson 3: Time it right, or die trying.

Nathan Furr analyzed that ‘most startups are dying and they are dying because they are doing good things but doing them out of order’. In the startup scene, this failure phenomenon is called premature scaling. And since every startup is trying desperately to scale, it is no surprise that premature scaling is the No. 1 cause of startup death.

So once you’ve established that your business model is in fact scalable, then you’ll want to know when to scale. Usually, the time to scale is just after a startup has hit its MVP and is then working towards product/market fit and beyond. Startups, in the B2B sector typically go from serving 100 customers to 1,000, generating from EUR 10,000 of early revenue to EUR100,000 monthly. In the B2C sector, startups have usually found their first 10,000 users by this point and are generating EUR 1,000 in monthly revenue from their core of active users.

Another sign that the time is right for scaling is when startups experience increasing active user growth and retention rate during a period of 6 months. It is highly likely that these startups have figured out a customer acquisition channel that works for them, enabling them to deliver the value to an increasing number of users.


Lesson 4: Scaling isn’t a strategy; it’s a way of life.

There are certainly more signs and metrics that can signal to founders when it is time to scale their business operations. But most importantly, as a general rule: Don’t set your focus to scale; otherwise you’ll fail!

If you scale one area only, then your startup becomes imbalanced and inefficient. So, for example, when startups attempt to scale only by increasing the marketing budget – but they neglect to factor scaling into the production process or distribution channels – then it’s no wonder that they fail.

Much like meditation on the path to mindfulness, if you only focus your mind from the comfort of your sofa, then you can’t expect to be feeling zen at the office right before the deadline. As a founder or founding team you need to focus on ‘becoming zen’ in all areas of your business – customer, product, team, financials, business model – with equal attention and commitment.

From MVP to product/market fit – why definitions matter

Over the past few months I’ve had discussions with various investors, entrepreneurs and others active in startup ecosystems about startup stages, terminology and definitions frequently used in our industry – terms like MVP, product/market fit, and more. After spending nearly two years in Silicon Valley investing and mentoring tech startups and developing tech ecosystems and startup communities, my experience suggests that people active in that space are frequently using these terms because they need some way to explain what stage they are at, but they are actually still getting it wrong.

Just this morning I came across an article by Lean Startup consultant Dan Olsen titled “A Playbook for Achieving Product.Market Fit.” This article made me curious because it claims to offer a guideline to achieve product/market fit, which is a big deal these days. However, as it turns out the article is actually an aggregation of wrongly applied terms leading up to advice which mixes up things to really get to product/market fit.

So let’s start at the beginning by explaining some terms:

The first iteration of your product or service may be a prototype, which is developed in order to test hypotheses and gain invaluable feedback, but represents a non-viable product from a customer’s perspective. Startups and entrepreneurs may use the prototype to test for the existence of a problem, people often also call it a ‘Vanity MVP’.

A Minimum Viable Product (MVP) is not a product startups can decide to launch; rather, the market tells you when you’ve reached your MVP. This is the most common problem when it comes to terminology, because entrepreneurs as well as investors are often talking about ‘launching or creating a MVP’, which is simply wrong. MVP is the stage of product development entrepreneurs could get to through the build-measure-learn cycle of the Lean Startup Methodology. In contrast to a prototype, a MVP needs to be a viable product or service to customers, so must include enough to get people using it and extracting value from it.

Product/Market Fit is achieved if the startup has found a repeatable, scalable model that drives demand. It is the stage after you’ve hit the MVP, and can be described as tuning the engine of your business by developing, selling and marketing your product, as well as building out the team and operations. The goal of the MVP is to create customers, the difference with product/market fit is that it involves measurement, learning and fine tuning, including actionable metrics which are not available before a MVP has been reached.

Some people also distinguish between before product/market fit and after, thus having a product focus which then turns into a distribution focus. Ash Maurya describes the various stages as following:

  • Prototype -> the type of product or service used for customer discovery
MVP -> the type of product or service used for customer validation
P/M fit -> the type of product or service used for customer creation
Growth/Scaling -> the time to actually build your business

Finally, what does it feel like to achieve product/market fit, or how can you notice it?

In general, achieving product/market fit can be noticed by the time the buying behaviour of your customers is switching from push to pull. Instead of startups pushing every potential customer through the door in order to use and buy their product or service, startups will receive inbound requests through various channels like paid ads, developed viral loops, improved SEO, etc.

Andrew Chen also provides some hints based on his experience investing and working with tech startups:

Product/Market fit is the time when

  • customers are buying the product or service as fast as the startup can make it
  • usage is growing as fast as you can scale up your technical infrastructure
  • money through sales is piling up in your bank account;
  • lots of sales and customer support staff are being hired;
  • reporters and news publications are writing about your value proposition and growth;
  • investors are interested to follow up or get in touch;
  • the competition starts popping up all over the place (different regions, verticals).

In order to achieve product/market fit, the time when founders go crazy because “Many people use it, and they pay!”, you need to have hit the MVP first. How to get to the MVP was best explained by Ash Maurya at his talk at SXSW. To get to product/market fit though, startups need to scale customer adoption, test marketing campaigns and customer acquisition channels, discover their revenue/unit economics to find profitable customer segments and streamline their business model. From a business perspective, startups need to assess the market size, focus on hiring key team members, set up operational processes like sales operations, and improve your product while testing additional features.

From MVP to product/market fit, definitions matter. Without understanding where your startup is going it’s going to be very difficult to get there.