How To Create An Artificial Intelligence Startup
AI (artificial intelligence) is definitely the “in” thing right now in the tech world. It seems like there is a new startup spinning up every day. And yes, many existing companies are re-branding themselves as AI operators.
The irony is that this technology has been around for decades but it is only recently that it has gotten traction. Then again, there has been a convergence of various technologies that has made AI a reality.
“There has been more progress in speech recognition technology in the last 30 months than in the first 30 years,” said Jamie Sutherland, who is the CEO and co-founder of Sonix (which is an AI service for transcription). “It’s not just the massive amounts of data that are being collected, it’s the fact that this data can be mined at amazing speeds. Computing power is increasing at an exponential rate. This opens up a whole new world of opportunity for novel applications to be developed that wouldn’t have been possible only a few years ago.”
Yet the AI market is getting increasingly crowded, with the noise level hitting fever pitch. So then, how can you set yourself apart? What are the strategies to consider?
Well, to get some answers, I reached out to various CEOs and founders of AI companies:
Dan O’Connell, chief strategy officer and GM of VoiceAI at Dialpad (formerly the CEO at TalkIQ):
Founders often forget about three important elements of building an AI company, which we learned first-hand while building TalkIQ and are now applying as we continue our journey at Dialpad (which acquired TalkIQ in May). The first is hiring a team with a diverse combination of academic experience and product development expertise – this mix is important in building models, designing features, and bringing them to market. The second is setting realistic expectations on which problems your product can and can’t solve — otherwise users are highly likely to experience disappointment because AI as a whole is still in the first innings. Finally, it’s crucial for your engineers to interact directly with customers and the features they’re building. If you’re not excited about what you’re building and it doesn’t work in the desired way when you test it, you can’t expect someone else to pay for it.
Roy Raanani, the CEO and founder of Chorus:
Plan a product roadmap that gets a minimum viable product into customer’s hands as quickly as possible, even with no AI implemented. This will dial you in to your customers, how your solution fits into their existing days and workflow, and allows you to start collecting real-world data you can evaluate AI algorithms on.
Peter Wang, co-founder and CTO of Anaconda:
AI startups face both strategic and tactical challenges. Tactically speaking, getting and maintaining access to high-quality data is typically the lynchpin of any prediction system. Oftentimes, the corporate customers most likely to understand the value of a sophisticated AI product will also probably have massive amounts of internal, messy proprietary data to integrate with before they will purchase. On the flip side, potential buyers with lesser data integration challenges may also have less understanding of the unique value of AI over traditional business analytics and statistics.
Related to this, AI startups must make a fairly strategic decision early on in their lifecycle about whether they primarily want to be a “prediction-as-a-service”, API-style offering, or if they want to build a full-featured polished app that faces the business end user. The former lets them focus on core differentiators, at the risk of commoditization by other larger platform vendors. The latter lets them own the user experience, but at a much larger up-front cost and a risk of unnecessarily boxing their technology into a niche.
Pini Yakuel, the CEO of Optimove:
The ability to perform better math is no longer enough to build a successful AI-driven startup because in the rush to be an AI-focused company, startups have stopped taking into account the business problem they are solving (and not every business problem is best solved by AI). What can increase chances of success, is knowing that a certain industry challenge has a strong likelihood to be solved ‘well enough’ with an AI approach. For that, you need to excel at defining and framing the problem you are out to solve. Your creativity and expertise will be your trump card, not a better algorithm.
Dimitri Sirota, the CEO and co-founder of BigID:
One mistake companies in tech sometimes make is getting their message lost in the jargon they use to describe it. AI and ML are a means to an end. Saying you are an AI company alone is not a durable difference when AI becomes commonplace. The essence of a successful start-up is solving a problem that a large enough universe of customers understand themselves to have. Telling that story effectively is how a company wins. AI can be the adjective but it’s not the noun.
Mahesh Ram, the CEO of Solvvy:
AI companies should pick a domain where consumers or businesses are currently willing to adopt AI/automation solutions and there’s social and cultural acceptance. This way you are not fighting against the current early on. For example, people don’t want robots as their doctors (yet); however, they are ok with automation when it comes to customer service.
Andrew Filev, the CEO and founder of Wrike:
Don’t include “AI” in the name of the company. Although AI is far less sci-fi and more commonplace now, it can still scare a few folks off and you don’t want to frighten your potential key buyers by making them think you are offering an AI solution that will one day replace them.
This article originally appeared in Forbes.