Get access
Future of Work 5 min read

Starting an AI company: Why "AI-First" Products Succeed in the Market

When starting an AI or machine-learning company, it's crucial to build in a data-driven infrastructure right from the start. Failing to do so quickly leads to legacy systems, where adding intelligence as a “feature” on top is extremely difficult.
Ghaith Bilbeisi

The success or failure of artificial intelligence (in this context equivalent to “machine learning”) in a product often comes down to the two F's: Is it merely a feature or is it a fundamental element of the product. Companies that think of AI solely as a feature often come unstuck when they try to add it to an existing tool. That's partly because AI doesn't add something useful in every scenario and partly because gathering the necessary data to power AI and handling its outputs can require some major structural changes to legacy products.

The companies that use AI most successfully are those that use AI as their fundamental basis. They can bake AI in to their systems and structures from the beginning, meaning every aspect of their service harnesses its potential more effectively. That’s even more the case when companies use AI to achieve their core functionality rather than simply trying to think of a way to use AI for the sake of it.

The companies that use AI most successfully are those that use AI as their fundamental basis.

Krisp is a prime example of a startup business that's thrived by using AI as the fundament of its product. It developed an app that removes echo and background noise from videoconferencing calls so that all you hear is the speaker — an app whose time certainly came in 2020. The key to its success is using AI to develop rules for what to filter out and what to leave in, adjusting and updating those rules to fit a specific user’s surroundings. AI means the more customers Krisp has, the more its software learns and the better its service. That means the more it benefits from word of mouth, the more exponential growth is possible.

Sticking with videoconferencing, Zoom's main features harness intelligent and highly sophisticated software, with its main component being automatic adjustments to maintain video quality even with low bandwidth, but also including meeting transcripts or scheduling suggestions. The fact that its video quality is a lot superior to Google Meet's is the main reason why Zoom has been able to gain and maintain its market share, despite Google's omnipresence and integration with our everyday work tools.

Grammarly, as a simple and powerful example, is a spell-checker and grammar tool with a key advantage that differentiates it from the familiar underlining in Microsoft Word. AI helps it adjust on the fly to what somebody is writing, giving it enough context to make real-time suggestions for improving writing in a way that suits the tone and intent of the document. AI doesn’t simply make Grammarly better, it makes it different enough for potential users to give it a chance.

23andme’s entire business model relies on artificial intelligence, specifically machine learning. The way it analyzes DNA samples to trace ancestry is the perfect use of AI, in that it combines human-style logic and reasoning with the speed and capacity of computers. It’s also a great example of how a business with AI at its foundation can evolve into new services: The company has expanded its analysis to offer testing that can indicate the risks of specific genetic diseases.

🥇 Get early access to Reasonal Teams to explore how we have built an automated, smart, and collaborative file and content management tool based on machine learning.

AI-based data science in companies isn’t restricted to niche, single-purpose businesses, either: Almost everything Google does, from voice-based smart speakers to search results, incorporates AI. Indeed, it’s the reason you can carry out searches with natural language rather than having to type in a search term in a specific, stilted format.

AI-based data science in companies isn't restricted to niche, single-purpose businesses, either.

These successes raise the question of why so many business projects involving AI ultimately fail, with one study putting the flop rate at 85%. In many cases, it’s because the business didn’t have a clear goal or problem to solve, let alone one where AI in worktools was the most appropriate answer. But the most common reason seems to be that the business and its systems didn’t already have AI at their core, and it wasn’t willing to make the widespread organizational changes necessary to put it there.

The simple truth is that you can only realize AI’s full potential when it’s at the heart of what you do rather that something you try to patch on or plug in.

📬 If you found this blog interesting, subscribe to our newsletter, where you'll receive some exclusive content on productivity, future of work, coding, AI in worktools, and news about Reasonal.


Ghaith Bilbeisi

Ghaith Bilbeisi
Software Engineer



Enjoy the read?

Receive our monthly digest of future-of-work topics, coding insights, and Reasonal updates. Be the first to get early-access offers for our tool.

Unsubscribe anytime.

More from Reasonal

Sign up and stay in the loop.

Receive our monthly digest of future-of-work topics, coding insights, and Reasonal updates. Be the first to get early-access offers for our tool.
Unsubscribe anytime.