Should My Startup be Using Machine Learning and AI?

Daniel Faggella

Daniel Faggella is Head of Research at Emerj. Called upon by the United Nations, World Bank, INTERPOL, and leading enterprises, Daniel is a globally sought-after expert on the competitive strategy implications of AI for business and government leaders.

Ben Levy Emerj

Almost all young companies and companies in the start-up ecosystem today would like to talk about how they leverage AI or machine learning. In most cases, this is to make their product sound special and exciting. However, the fact is most young companies do not have the resources or the talent needed to leverage artificial intelligence in a way that would add to their value proposition.

The purpose of this article is to help companies understand whether their particular business model needs artificial intelligence from the start, or it is something that may come along later as it is not required for the basic value proposition of the company itself.

By the end of this article the reader should have a better idea of:

  • Whether artificial intelligence is critical to reaching one’s near-term objectives or not
  • When AI should be considered in the future (if not today)

This will help people understand whether their business is what I call an “AI first” or “AI later” business model.

The article will also present a framework of thinking to identify where artificial intelligence might fit into a business in the future if it is not an AI first model. The framework will provide a level-headed approach to AI adoption for companies who can’t (or shouldn’t) adopt AI immediately.

(Readers interested in additional “intro to AI” resources might enjoy our articles in How to Apply Machine Learning to Business Problems, and Everyday Examples of AI.)

1 – The Great Promise of AI: Winner Takes All

The reason AI is in the pitch decks of so many startups, whether they are using it or not, is because artificial intelligence actually has serious ramifications for some business models. On the one hand, many people are excited about AI simply because it is new and somewhat mysterious (some investors fall prey to this curiosity weakness as well – as we’ve heard from many AI startups who raised money with “we use AI” as a major fundraising hook).

However, the more important business aspect of artificial intelligence is it tends to lead to a winner-takes-all dynamic.

Some companies leveraging artificial intelligence can develop a significant advantage over other companies in their industry that would not have been possible without AI – and this makes some AI-powered companies into extremely appealing business models for investors (interested readers can see our full list of investor interviews here).

What kind of companies are these? These are companies with access to proprietary data (data to which only they have access), and the ability to acquire and use this data to continuously improve their product in real time. Their advantage is predicated on proprietary data rather than proprietary AI or machine learning, allowing them to stay ahead of the curve and develop much better products than their competitors (we break down this idea in greater detail in my in-depth Huffington Post article on AI investor perspectives).

Ben Narasin Emerj
Ben Narasin of Canvas Ventures

Canvas Ventures’ Ben Narasin observed in an interview with Techemergence regarding startups that succeed in raising venture capital:

“…very few people are pitching that they have some proprietary ML; what they’re pitching is that they have proprietary data sets which can inform that ML, which is what creates great power.”

An example of a very successful company that leverages proprietary data is Amazon. The company was an early adopter of the online commerce framework, using the Internet to sell books online initially. It has since expanded to other products, but because of this early adoption, many more people have made purchases on Amazon than any other competitive site. Amazon is the top online retailer in the world as of 2017 with $178 billion in revenue.

However, the real marketing power behind Amazon is its ability to recommend books and other products with a high degree of relevance based on a user’s purchase history as well as the purchase history of similar users. Because it has access to a huge amount of purchase history data, i.e. proprietary data, the company can and does make excellent product recommendations for any user, even if that user has a limited purchase history on Amazon. In many cases, this results in more purchases and proprietary data, hence the winner-take-all dynamic.

Another good example of this dynamic is Google. There are hundreds of search engines today, but it is undeniable that Google dominates in volume of users – and this usually ensures that they’ll continue to have the most relevant results (as they have more searches to train their algorithms on).

The main driver behind this search engine dominance is search data. It is highly probable that large tech companies such as Microsoft with all their resources could develop better technology and algorithms than what Google has today. However, that would not be enough to outperform Google in search relevance simply because they would not have access to the huge amount of search data provided by the billions of Google users. Google uses this data and machine learning in real time to understand what people are looking for better, which in turn allows it to provide highly relevant search results indexed in degrees of relevance for a particular search term and user.

Since Google results are typically better than any other search engine, more people continue to use it to make searches, feeding Google search algorithms to make it even better. Because of this cycle of more data improving search relevance resulting in more data, people have no reason to use any other search engine.

Gary Swart Emerj
Gary Swart of Polaris Partners

One of our other investor-interviewees, Polaris Partners’ own Gary Swart, used the following startup example to highlight the same kind of data ownership opportunity:

“…Here’s a good example from our portfolio, a company called Inside Sales…they take the learnings from all of their customers, and those learnings benefit all of their other customers. So they have more than 100 thousand users on their platform, and they’re passively observing the behavior of sales results…

…so it’s a dialer that enables you to more effectively and efficiently contact your customers, and every time you contact a customer, the platform learns from that behavior and those best practices, and then that in turn benefits everybody else when making a call.”

This winner-take-all dynamic drives fastest growing and most profitable companies like Google, Amazon, Facebook, and Netflix. They are pulling farther and farther ahead of their competitors because proprietary data and artificial intelligence gives them an advantage over everybody else. The great promise of AI is providing the backbone for market dominance, as illustrated by some of the most profitable and fastest growing companies in the history of the world.

It is not surprising that every startup in Silicon Valley wants to dominate their industry in the same way as these companies by leveraging artificial intelligence in one way or another. However, AI is not always going to foster that winner-take-all dynamic for every business model – and certainly not in a company’s early days before achieving scale. In fact, AI is not going to help for many business models at first, or even at all. Some startups are simply “AI-first,” and some are not (neither is inherrently better than the other).

2 – Is Your Startup “AI First” or “Not AI First”?

Any startup interested in leveraging artificial intelligence should first determine if it has an AI-first business model or not. There are two criteria for this.

  1. The product or service requires AI to deliver the value proposition
  2. The founding team includes someone with robust AI experience and/or skills, either through academic training or marquee company employment

The first criterion relates to the nature of the product or service, which directly affects the value proposition. If the product or service is not made possible without AI or machine learning – then by definition it is an “AI first” company.

Medical diagnostics is one area in which machine learning has some traction because it requires processing a large amount of data to get a relevant result.

For example, if the value proposition of startup is using machines to diagnose cancer based on medical images, then that company must begin with a significant investment in machine learning talent and applications. There is no other way to start the company.

Another area where artificial intelligence may make a difference is predictive analytics. For example, let’s say another startup provides predictive maintenance services for heavy machinery manufacturers (like Predii or a similar vendor). The clients use sensors to monitor and collect heat, friction, and vibration data in manufacturing plants. The predictive maintenance service company can use this data to train artificial intelligence software to detect potential problems with the machines. Again, it would be near impossible to achieve this aim without machine learning – and a company with this value proposition would have to be “AI first.”

In other words, if the startup’s value proposition needs AI to deliver the best possible product or service, then it is an AI first company.

However, there is a second part to this determination: An AI-first company also needs AI-related talent on the founding team.

A startup company that requires artificial intelligence to deliver its value proposition needs to have at least one founder that has robust artificial intelligence experience in academia or from working marquee companies like Amazon or Google. In a previous article on how to cut through “AI hype”, I mentioned that a lack of at least one AI expert founder (through academic training or cutting-edge hands-on experience in a previous position) is a tell-tale sign of a company that’s faking their supposed use of AI.

Somebody on the founding team needs to have a deep understanding of the core technology required to deliver on the value proposition. If nobody on the founding team understands artificial intelligence technology on this level, business failure is a high likelihood, not the least of which is because it would be almost impossible to get funding from a venture capitalist.

A startup pitching an AI-related product or service looking to raise money to hire AI people is not likely to convince a venture capitalist about the soundness of the business plan. A venture capitalist will want to see concrete evidence that the startup can deliver on its promise, and that typically means having basic technology in place and some evidence of traction in the industry.

Therefore, the definition of an AI-first startup company is one where artificial intelligence is a necessary element of the value proposition and has a founding member with the requisite AI experience. Such a company has a good chance of attaining a winner-take-all dynamic.

That said, it should be clear that this article does not promote the idea that founding teams need to build a business model as an AI-first company to attain success. Solving an important customer problem and growing revenue are far more relevant than using machine learning for it’s own sake. This aricle is merely a guideline for determining if a startup should be called “AI first” or not.

Qualifying as an AI-first company does not guarantee success. Many extremely successful companies today were not AI-first companies, although some may eventually “grow into” artificial intelligence as the need arises.

3 – How to “Grow Into” Artificial Intelligence

Any startup that does not qualify as an AI-first company may also eventually apply artificial intelligence in order to stay competitive in their industry. While it may not have had the talent, money, or data access to allow the use of machine learning in the beginning, it can certainly plan to keep AI in the horizon as part of long-term business strategies.

Artificial intelligence and related technologies today is at a stage in which Internet companies were in 1995. The tools and skills necessary for building a website capable of online commerce were too complicated and expensive for most people. Two decades later, these tools and skills are much more accessible, and companies use the Internet routinely for internal and external purposes.

The same thing is likely to happen with artificial intelligence. Today, AI is new, inaccessible, and expensive, but it will be everywhere in a decade or two. As the landscape shifts and AI becomes a natural part of software products and online tools, it is important to think about how to leverage it when the time comes. A non-AI-first startup company today should plan to take advantage of AI in the future to stay competitive in the market.

To do this, there are three questions to answer:

  1. What is the company’s value proposition and what can it do for customers better than any other company does?
  2. What data can it access to help it do that job better?
  3. How can it use AI to access more proprietary data than anyone else – and improve the product in the process?

To illustrate how this framework can help prepare a non-AI-first startup to leverage AI in the future, consider the following examples.

Growing Into AI – Marketing Automation Example

A company provides an email marketing solution for eCommerce companies, many of which are in apparel and clothing. Its value proposition is allowing customers to target their campaigns to the right person at the right time to drive more sales. The product helps online retailers sell more products by automating campaigns and targeting the right readers.

In the course of working with their clients, the company gains access considerable data. This includes the various subject lines of their emails, email copy, open rates, and eCommerce sales generated from each of these email campaigns. The company also has access to data related to market segments and campaign pacing.  

The company can collect the data from their clients and use that information to feed artificial intelligence software in the future. AI will help improve the software to launch better email campaigns for existing and new apparel customers, increasing their rates of success each time. A new customer selling clothing can use this software to build a new campaign for a particular product using the right email template, subject lines to introduce the brand, and a follow-up campaign to get people to buy more products for a particular market segment.

Growing Into AI – Event Promotion App Example

Another example is a company that develops an event-promoting app for smartphones. The value proposition of the app is to help different kinds of users find fun events and activities to which they might want to go. It will also find the best available discounts for these events and activities, applicable dates and times.

Such an app would provide access to data about events all over the country, and record the ones that appeal to different kinds of users based on the ratings from the users, and the number of users that check out different kinds of events. The company can access data every time that a user makes a search for a particular type of event, clicks on an event, or rates an event.  

The company can use this data to get a good understanding of the type of events that are most popular for different kinds of users, and eventually build a robust recommendation engine much like the ones used by Netflix and Spotify. The basis of the recommendations is the preference data of all their users over time, and with artificial intelligence can determine the right recommendations to make their users happy, encouraging them to use the app as often and for the longest possible time. An event-promoting app that people find useful builds a loyal following, providing more data that can make it better and more valuable to both users and clients, which in case is event companies.

These are just two examples of how a company can grow into artificial intelligence. The key concepts here are data access and gathering. AI can only learn if it has data and the more data it has, the better it learns. A company can only grow into artificial intelligence if it treats any data it can access and gather as a valuable resource, which will come into its own when the time is right.

Concluding Thoughts – How to Think About AI Adoption for a Startup

Many startup companies will put artificial intelligence in their pitch deck to increase its appeal. Most of the investors we speak with tell us that if “artificial intelligence” isn’t on page 1 of the pitch deck, it’s on page 2 (whether the founders know what artificial intelligence is or not).

Most startups today will not need AI at the beginning of their journey to product-market fit or initial profitability. Artificial intelligence will lead to some very powerful companies, and beginning with AI expertise is going to be very advantageous in some cases, but it will not be relevant for many business models, and may even cause a startup to fail.

For a good chance of success, the most important thing for a startup company to do is provide solutions to important problems and find somebody willing to pay for it. If a startup can determine a value proposition and deliver on it without AI, then it should not leverage AI. If it can acquire users or customers, build a strong brand, earn revenue, and grow in terms of revenue and users without artificial intelligence, then it is a successful company.

Ben Levy Emerj
Benjamin Levy of BootstrapLabs

Growing in traction, users, and revenue is more important than saying or trying to use AI. In an interview with BootstrapLabs co-founder Benjamin Levy, he explains the purpose of AI from a business point of view:

“…I think it actually starts with basic lack of understanding of how the technology is going to impact their business…if you’re a business person, at the end of the day, you’re going to be caring about your ROI—is my product better, faster, cheaper; am I selling it to my customers and are my customers happier, and am I competitive.”

Companies that are not AI-first should focus on core activities: finding and analyzing the problem, finding solutions, finding clients, and growing revenue. If this entails doing something other than AI (which applies to most companies), then that is what they should do.

For the most part, companies that can do these core activities will find a use for artificial intelligence in the future. They should keep it on their strategic horizon using the framework discussed in this article. There will come a time when AI will be more accessible, and when that time comes, these companies should be ready to take advantage.  


Header image credit: PanamericanWorld

Stay Ahead of the AI Curve

Discover the critical AI trends and applications that separate winners from losers in the future of business.

Sign up for the 'AI Advantage' newsletter:

Stay Ahead of the Machine Learning Curve

Join over 20,000 AI-focused business leaders and receive our latest AI research and trends delivered weekly.

Thanks for subscribing to the Emerj "AI Advantage" newsletter, check your email inbox for confirmation.