A.I, Blockchain: buzzwords aren’t (always) products.

In2009, three French entrepreneurs, frustrated by the painful process of purchasing train tickets online, decided to tackle the issue and build an improved version of a ticket-purchasing platform.

The product was simple. In fact, “simple” was the whole point of it. Their UX was flawless, while the UI was sleek and appealing to the eye. Buying tickets could now be done in under a minute, including the most convoluted journeys that were previously huge hassles for travelers.

Spreading like wildfire, the success of their service eventually reached the shores of England, where the already popular ticket-purchasing platform “Trainline” decided to acquire them in 2016 for nearly $200 million.

But what was the product they created? Was it UX? No. Captain Train was a train ticket selling platform [Product] which main’s value [Unique selling point] was an improved User Experience, that allowed for quicker and simpler purchasing of tickets, especially for complicated routes[Feature] — and that’s probably how they pitched it to carry out their successful exit.

Their USP was the reason the Product was so popular, and what enabled them to produce superior features than their competitors.

So why do so many people, entrepreneurs and investors alike, mistake USPs and features for Products?


An element of answer, indisputably, lies in the trending “buzzword-mania”, for which social networks (and the infamous hashtag) are partly responsible. Some technical terms, because of their excessive media coverage and the complex underlying principles they convey, hold a heavy “sexy factor” while reducing the exposition to questions that matter, such as… “what’s your plan?”

As such, entrepreneurs often opt for the shortcut that makes them look smart and expert. Why would one go through the whole process of designing features, finding an initial niche to iterate, A/B test it during months of coffee-infused nights to reach their Product/Market fit in the hopes of raising funds, when a single magic word on a powerpoint (or white paper) can get them piles of cash wired to their accounts from undiscerning VCs or amateur Business Angels?

We can’t blame either side: entrepreneurs are simple biological, bipedal machines thriving on survival instincts; investors, on the other hand, tend to lack the free time to dig deep into brand new and very complex subjects (even though in this case, it is usually better to fold and play the next hand).

This phenomenon stands particularly true for the two buzzwords en vogue: Blockchain and Artificial Intelligence.

There are incredible projects out there leveraging each of these techs, often carried by A+ tech-oriented founders able to execute on their plans, but unfortunately they’re the exception rather than the rule.

Blockchain, the infamous.

Blockchain is a complicated topic because of how technical it seems to be, safeguarded by experts that prefer keeping the veil of mystery rather than having its purpose democratized for everyone to join the party.

Overall, Blockchain is to be thought as a USP or feature. Don’t mistake this argument for a debate on whether it is, or not, a game changer— that’s for another day. But blockchain is usually to be used as a feature for an existing project.

One could argue that in the case of Ethereum, EOS, NEO (etc.) their product is a blockchain. They’ve designed their own blockchain with a specific set of features (pricing, speed, unbreakability), allowing other projects to get built quicker and simpler. In the traditional tech world, they would either be a framework or an operating system, depending on their usageThey can, for instance, replace cloud-hosting solutions with added complexity in the tech stack, as the infrastructure where data is stored, secured and delivered.

However, those represent less than 1% of projects marketed as blockchains; in fact, the whole industry can be really divided into 3 types of products: Blockchain as a product (works for the examples above along with cryptocurrencies); Products that are improved by the addition of Blockchain (especially relevant in cybersecurity); and products that could build similar features without a Blockchain (90+% of projects).

3 types of Blockchain usage, from most to least valuable

The latter is the reason why the industry got suspicious in the first place. In the wake of the “ICO furor” of 2017, opportunists started advertising their “Blockchain project” applied to a product or service, when it should really be the other way around. If I could use your product perfectly before Blockchain became popular, why are you asking for $50mln to make it your new main selling point?

“I’m building a Blockchain project” can hence only be used for the first kind of projects outlined, as Blockchain is not a product in the vast majority of cases, and founders should be aware of this.

So how should you pitch your project to level-headed investors, bringing up the blockchain without falling into the buzzword fashion? Here’s an example:

“We’ve created a a platform that offers a new way to [secure complex data]. Because of the flaws in traditional systems, we’ve opted to rely on a blockchainfor optimal security, which allows us to offer [insert feature].” Clear, simple, and straightforward.

Now, to our other superstar.

Artificial intelligence, on the road to singularity?

The tricky part about this tech is that there are still disagreements on what is and isn’t considered A.I. A big chunk of projects marketed as A.I today are basic algorithms, formerly promoted as “our engine”, henceforth sold as “a cutting-edge A.I”. Breakthroughs wearing this label usually stem from the intersection of the ever growing availability of computational power and the exponential growth of data points rather than algorithmic improvements.

As such, the meaning of “building an A.I startup” is unclear — which opens a huge opportunity for malicious founders. In a broad sense, A.I is generally defined as an engine able to predict outcomes and improve from observation as does the human brain, specifically the input/output screening performed by our neurones. This reaches far beyond the traditional iterations rules of software algorithms.

Given that the best researchers in the world gladly admit they’re far from reaching this description for their prototypes, there is only room for intermediary steps that tend towards a true A.I. But while climbing this steep ladder, a project must have a value-proposition that can be immediately of use to their prospects, which usually demands harsh tradeoffs on the roadmap. So if you don’t have access to the same brainpower or quantity of data than the GAFAs or BATXs, you are probably not doing A.I yet.

An engine can self-improve in several ways, the most popular being Machine Learning. Its main perk is that it is simple to deploy thanks in part to open-source projects, generally made available by large tech companies (ironically, hackers now use their public resources to attack them — talk about backfiring!). Machine learning is hence widely used, despite a major drawback: it requires immense amounts of data to be efficient. The implications for early projects are simple: if they’re unable to gather massive quantities of public cohorts for their desired usage, their product can’t become competitive. This is a negative loop: the slower they grow, the smaller their product’s value is — they might be unable to deliver performance to their first users. That’s why in the end, the Silicon Valley often wins. A striking example of this rule is the recent launch of Waymo’s self-driving taxis, that outpaced their competitors because they had simply stacked more driving data.

When discussing with founders that claim to develop an A.I startup, they’ll sometimes discreetly acknowledge with a bashful smile that “it is rather a marketing term”, or what they call A.I is “only on the long term roadmap”. Reducing a startup to its main long-term goal, the A.I as a product, exposes entrepreneurs to the risk of appearing to deliberately mislead investors, or worse, being naive. Prefer a straightforward approach, defending your project’s early value while outlining your plan to reach the full-throttle AI-hybrid engine you’re aiming for. And remove the term from your one-liner pitch.

Just as for Blockchain, relying on “A.I is far from mandatory to deliver performing tech products, and more often than not founders should be able to demonstrate their product /market fit plus iteration capabilities before a more advanced tech becomes relevant.

How then should your pitch your project?

“We’re building a platform that predicts [something]. To improve our engine, we’re looking at using machine learning models as we collect data to ultimately outperform competitors. In the long run, we intend to have an autonomous and self-improving engine.” That’s your AI right there.


Fortunately, as the hype is cooling off and investors have had sufficient time to benchmark those technologies, it‘s getting increasingly hard for the serial “buzzword marketers” to close funding rounds without having proved their worth beforehand.

But that doesn’t mean we shouldn’t stay alert. In recent memory, several projects lackluster of any of the usual conditions to close rounds have managed to get funded, sometimes above their expectations. Will they fail ? Maybe not. Is it wise to finance founders that — deliberately or not — promote their USPs in place of their product, before they even demonstrate any tangible proof that they’re on the right track? Definitely not.

0 Comments

Leave a Comment