How to identify break-through ideas for AI
Co-Written by Adrian Locher and Finn Grotheer.
This article is part of an article series about Merantix. In this article, we will outline how our ideation process works — from hiring the right entrepreneurs, to identifying a second-to-none business opportunity, to validation with our network of advisors and industry partners. In interview rounds for our founder role, candidates often inquire where our company ideas come from or if they need to have an idea in mind already. So, we wanted to share some of our thoughts and best practices — that might also be interesting for a wider audience. After all, building a company that changes things for the better is in the interest of all. It warrants an open minded exchange with the wider community, to which this article might contribute even if ever so slightly.
Where our Ideas Come From: Ideas are Everywhere
The first and foremost well of ideas are the founders that join Merantix. We constantly are looking for the smartest entrepreneurs, equipped with stamina and charisma in order to lead a team and to overcome obstacles on the way. They distinguish themselves through unorthodox thinking and a deep-rooted ambition — to change the world, to transform the most important fields and industries, to invent their own job instead of taking up just another consulting opportunity. When founders join Merantix, they usually don’t have the company in mind that they want to build, but they may come with a particular experience that triggered their interest in a specific domain — ranging from a particular technological subfield, like Natural Language Processing, to domains, like healthcare or environmental topics. Other times, they have a more generalist background, accompanied by a curiosity for business models and go-to-market strategies. In essence, the variety of possible sources of company ideas is a manifold as the CVs of our founders are.
Once aboard, Merantix founders scoop ideas from multiple conversation partners and research topics. This includes friends and acquaintances from venture capital firms, our network in academia, advisors, industry contacts and other founders alike. At this point of ideation, it is important to think critically and combine learnings and thoughts from different layers of the process. Innovative ideas form through a complex interaction of suggested domains to explore, reported industry needs, and technological breakthroughs that open up new fields of application. Often, an approach from one industry could be remodeled for another industry; and some problems occur industry-agnostic.
Diving in: Top-down or Bottom-up
Generally, there are two natural starting points of exploration for founders. The first starting point is top-down. Here, the initial interest orbits around an entire industry or domain, like healthcare or cybersecurity. Often, the founder in question has a background in this particular field but that is by no means indispensable or necessary. For this approach, the founder structures the field along categories that are somewhat mutually exclusive. This can be done by looking at different stakeholders, different value chains or, plainly, through existing subdisciplines. Each and every category that emerges possesses possible cases of application: where is a lot of repetitive work done manually? At which point is precision key, but human labour imperfect? Is the work done with visual inputs, like pictures or satellite images, or with natural language? Could it be improved?
In contrast, bottom-up starts with a very specific idea that a founder strives to validate. In this case, a founder has stumbled upon a possible business opportunity by chance or design, on the job or at any other point during the week. But while the initial starting point is already a concrete idea in this bottom-up approach, it usually is not that easy. There are many necessary conditions that need to be fulfilled in order for an idea to fly: data availability and standardization, competition as well as timing are just a few of the most typical roadblocks. But a concrete idea zooms a founder very rapidly inwards. It establishes industry contacts and brings attention to the periphery of the original idea. A web of related issues opens up from which further points of exploration emerge.
Top-down and bottom-up are distinct approaches but they can complement one another in the ideation process. An initial domain interest will sooner or later breed a specific idea. But weighing up a specific idea, talking to experts in the field and trying to validate some of the underlying assumptions, in turn, inevitably leads to a wider understanding of the field and other problems and opportunities in close vicinity. As a founder, one of the main challenges is to relentlessly follow through on intuitions and hints while never losing sight of what is going on in the surroundings. Switching between top-down and bottom-up is tough but incredibly stimulating, too.
Navigating Multiple Ideas
We encourage our founders to look at several ideas at the same time, in different spaces and industries. Firstly, this reduces waiting periods — when waiting for feedback from external partners and experts, as it frequently happens during the ideation phase. After all, just like investors invest funds, entrepreneurs invest their time. And determined as well as educated entrepreneurs usually deal with considerable opportunity costs — making their time all the more valuable.
Secondly, building and growing a company takes years, sometimes a lifetime. While stumbling onto company ideas has been romanticized in pop culture, it is both unlikely and unwise. When one makes a decision that affects oneself for years, one should make sure to have done the appropriate research. So, looking at as many opportunities in as many fields as possible is the foundation for an informed decision about what a founder wants to spend the coming years on.
Having looked at a variety of fields and ideas is also a means for quality control. If your idea has beaten 50 other ideas on close inspection, you can be confident that it is indeed a good idea. This approach pulls up the portfolio approach of VCs from behind, putting it in front of the actual investment. Instead of spreading the bets and weeding out the losers, a long and intense ideation phase can identify the winners upfront.
What Is a Good Idea?
It is hard to tell what features qualify a business opportunity as a valuable company idea. After all, it is a constitutive part of disruptive innovation that no one else saw it coming, and that no one-fits-all formula can lead you there. Nonetheless, there are some features that we try to look for. Most obviously, as an AI Venture Studio, Merantix is looking for cases where AI can make a difference.
Some ideas are truly new and revolutionary. They offer a solution to a previously unresolved problem or open up new market opportunities, untapped until today. In these cases, only hand-on research and contact to potential users and customers can clarify if there is an interest for this kind of product in the market and, importantly, a willingness-to-pay.
Other ideas aim to improve existing processes. We tend to think that the main drivers behind AI innovation are accuracy and automation. Ideally, an idea in this realm incorporates both elements: automating a fairly manual task while achieving gains in precision. When exploring a business idea it can be valuable to think about both parts independently: what are the most repetitive steps in the value chain? At which point would accuracy make an impactful difference?
Further points to take note of are the market size, the level of industry agnosticism, the relative relevance of the problem resolved through an AI solution, as well as a good justification for the right moment to move. First movers face information challenges and technological skepticism, while last movers encounter a crowded market. When an idea’s time has come, there is already some sense for the opportunities that AI can bring to a company or industry, but at the same time there has not yet been found an appealing go-to-market strategy that integrates well into a legacy process, albeit capturing a sufficient part of the value chain.
Validation: Challenge Approach, Check Feasibility, Confirm Industry Interest
What follows is an iterative, collaborative and creative process of validation. There is no set recipe — and there can’t be. Once a rough idea stands, it is necessary to confirm all the hypotheses that underpin it. On the technical side, founders need to confirm that their idea can be implemented with the tools and technology at hand — and what kind of data quality and structure that would necessitate.
Importantly, potential first users from industry need to confirm that there is an interest in the product proposed and that the technical needs can be met. This commercial validation phase is strikingly important — and routinely overseen by deep tech teams and companies. Given the technology’s complexity, teams often focus on getting a system running without checking if there is anyone out there willing to buy it. And whereas consumer products have relatively compact developing phases, cheaper prototypes, and shorter sales cycles, ML systems need more time and money until a prototype is viable. As a result, it is harder to receive user feedback early on and improve an application iteratively. So, we see a lot of solid systems failing to fly because teams were unable to confirm sufficient demand for their product or missed an important feature that would have been crucial.
Through our industrial vehicle, Merantix Labs, we have been able to observe many real-world industry projects and have identified commercial validation as a major chokepoint. Together with our network of partner companies, we run pilot projects in order to receive honest and trusting user feedback and build minimum-viable-products collaboratively. This strengthens our industry knowledge and de-risks the incubation process for founders, while helping companies on AI knowledge-buildup and solving specific problems.
Technical validation, commercial validation, building a prototype: all of this happens in parallel and might require multiple rounds of iteration. It is a little bit messy, requires high frustration tolerance and sales skills. But it is also incredibly rewarding — a high frequency of challenging and smart encounters.
About the Authors
Dr. Rasmus Rothe is co-founder and CTO of Berlin-based Merantix as well as a founding board member of the German Association of AI Companies (KI Bundesverband e.V.). He has published over 15 peer-reviewed papers while attending Oxford, Princeton, and ETH Zurich, where he received his Ph.D. in computer vision and deep learning.
Adrian Locher is co-founder and CEO of Merantix. He has been a serial entrepreneur and investor for the past 20 years, founding more than 10 companies both in Europe and the US, in digital healthcare, e-commerce and AI.
Finn Grotheer is a public affairs fellow at Merantix. He is a graduate student of International Affairs at the Hertie School of Governance and a fellow of the German Academic Scholarship Foundation. Before joining Merantix, he gained work experience at the Boston Consulting Group and Hering Schuppener Consulting.