Many Artificial Intelligence (AI) software companies are started because they have specific, promising mathematical algorithms as core intellectual property. Their ultimate success lies in identifying the killer application for these algorithms. Product Management best practices dictate that you should start with identifying a market need – then design a solution to that need that customers will pay for.
Starting from specific algorithms and working to find the business problem they solve is getting to product-market fit the hard way, a.k.a. backward.
Many AI companies fit this model and all market participants must be aware of this dynamic – whether it is to position your product or make a buying decision.
So why does this matter?
The law of the instrument says that if you have a hammer as your instrument, everything looks like a nail. Think about the AI algorithm as a hammer.
As the AI vendor…
…founded by a recognized mathematical genius, this does not appear to be a business problem. You know, academically, that there are many applications for your algorithms that have already been thoroughly researched, advanced, and theorized in numerous journal articles. Surely, if you swing this hammer enough eventually you will find a bunch of use cases companies will spend handsomely to solve.
Your key challenge appears to be getting enough potential customers to try it to see just how universally great your mathematics is. Orders will start pouring in when you find the right set of use cases.
As the potential client…
…you have limited resources and specific business problems to address. You have lots of questions to figure out if this is the right solution to any one of your needs.
- Do you need a hammer?
- Has the hammer been used to solve the same business problem you have?
- Maybe the hammer can solve multiple problems for you?
- Is there a less expensive, lower risk, alternative way to address your need?
The AI vendor just raised a huge amount of money from venture capitalists. Investors must know the value of the offering to make such an investment, right? Sitting on the sidelines while leaders in your industry seem to be buying hammers, saws, and screwdrivers does not seem like a good option.
In the age of rapidly evolving AI technologies, these fundamental questions are difficult to assess for a business unit. Many enterprises have created the concept of an innovation lab designed to test out promising technology while reducing investment risk to your line of business.
Implications for AI Startups
From a product management perspective, I strongly prefer to identify the market needs and users first. Then come up with solutions to those needs. This creates the most likely chance to succeed in building a valuable product or services business. You can focus wholly on the problem and incrementally deliver a valuable means to solve it. This is a proven easier path to deliver value and aligns with most lean agile methodologies organizations like to employ today.
Back in the real world, we do have mathematical algorithms discovered and companies believing they can develop commercially valuable solutions out of them. Venture Capitalists agree to this approach by pouring significant capital into the market each year.
Startups have two main product strategies to employ at this point:
Sell the Algorithm
Selling mathematical algorithms as-is is not viable since they cannot be patented directly and have minimal intrinsic value. Therefore, it is necessary to build enabling systems and methods that can unlock the power of the algorithms. Leverage data scientist crowdsourcing to tackle the challenge of finding a multitude of business needs your solution can effectively address.
Following this route requires either enabling your technology to be incorporated into other third-party data science platforms or building out a data science platform of your own. A data science platform is necessary to achieve scale in your crowdsourcing efforts and working with tools already popular will do this. The risk is competing against thousands of other open-source and proprietary algorithms – on someone else’s platform. This tends to commoditize your perceived value reducing commercial potential.
On the flip side, building out your own data science platform is a much larger investment with its own challenge of gaining significant adoption. Commonly this is tackled through a combination of opening the platform up to complementary algorithms and data science tools coupled with a freemium packaging and pricing model.
Vendors also will attempt to develop an active and engaged user community to socialize the benefits of the algorithms and the platform. Over time, the data science platform which was initially secondary to your business may become the core offering. This effectively reduces the risk of building operating a backward AI company because of the data science platform, which would be a solution designed and built based on customer need first.
Sell a Packaged Business Solution with the Algorithm
While you have not proven your AI to deliver significant value at solving specific business needs, you likely have some strong hypotheses where it fits. Through research and analysis, you can work to validate the market need, identify a market category to target, and derive features that must be available to compete.
This is the product strategy that will directly solve your customer needs best, which can often translate into premium pricing over platform approaches. However, this strategy will typically require the greatest investment to achieve a minimally viable product. The core reason for higher investment is the need to embed significant domain expertise into the solution.
Critically, in this strategy, you must be willing to invest in development beyond your original AI algorithms to reach product-market fit potential.
Some startups began based on specific advanced AI then search for market problems to solve with it. While this is backward according to traditional product management best practices, it is not uncommon in the AI startup world.
The risk for vendors in this situation is to not be aware of the impact of this on their positioning and product investments. Good product strategy does not just emerge as there are too many competing forces in any organization. The key is to start by recognizing current dynamics and then planning with intention.
In a dynamic market, there is no single strategy that is right but options to consider. Consider wisely.
In my next post, I will dig further into the segmentation of AI startups using Four Quadrant Positioning Analysis.