AI-powered tools such as chatbots and image processing apps have become the new way forward, and it’s not hard to see why as more and more companies invest in AI-based solutions.
Arthur Eichele was out hunting turkeys, he wore a turkey disguise. But the disguise was so good that a mountain lion bought it and attacked. Arthur fought back.
When the lion realized its mistake it thought, ‘my bad,’ and went to look for real turkeys.
Source: byjessicala.com
However, just like Arthur Eichele and the lion, it can be difficult to tell whether AI products work as advertised at a glance. When you find such apps on sale, you need to test some features and talk with sellers to understand the product.
You also need to know how to evaluate such apps so that you can make an informed decision.
That’s why in this article, we’ll go through the key questions you need to ask yourself when considering an AI product purchase.
But first, let’s define Artificial Intelligence (AI) products.
What are AI-powered websites and applications?
AI-powered websites and applications are applications that deliver value by deploying AI to perform specific tasks that would normally require human intelligence. The tasks might be predicting outcomes, recommendations, creating content, or any other intelligent tasks.
Such apps may deploy various AI technologies such as natural language processing, machine learning, deep learning, image processing, recommendation engines, speech recognition, and content generation.
Machine learning is the process by which machines learn how better to respond based upon structured big data sets and ongoing feedback from humans and algorithms.
AI solutions mainly fall into 3 broad categories:
Source: AI Marketing Solutions
- Core AI solutions
Core AI solutions focus on improving aspects of the field of AI by building technologies that perform or improve processes such as:
- Data cleaning
- Data modeling
- Deployment
As you can see, core AI businesses will usually produce applications that will be used by other AI specialists to create more particular AI-powered solutions and tools.
- Application AI solutions
Application AI solutions do specific intelligent tasks. They can be deployed across a wide spectrum of industries.
Typically, these applications don’t have a direct benefit and need to be applied to specific scenarios for them to be of value. Therefore, AI professionals also tend to be the highest number of people who use application AI solutions.
Examples of application AI include:
- Image and video processing
- Content creation
- Text and voice processing
- Industry AI solutions
These AI solutions are trained and optimized to do intelligent tasks in very specific applications. Because of this, they provide an immediate utility and do not require much setup and customization. Non-techies can easily use these applications.
Examples include AI-powered solutions for:
- Customer service and improved customer experience
- Search engine optimization
- Self-driving cars
- Cyber threat prevention
You will find that most AI-powered tools and websites for sale on Flippa fall in the industry AI category.
Now that you understand the main categories of AI products, let’s get into how you can evaluate them.
How to evaluate AI products
The goal here is simple: there are many general questions you should ask yourself when buying a website, but for AI products, these following 9 questions can help you get right down to understanding whether the application works as a product and as a business idea.
First, you want to find out whether an AI-based app or website is useful. Then you’ll want to determine how much you want to bid for the product once you’ve established its usefulness.
The steps below will help you vet the usefulness of an app.
1. What problem does the app solve?
For any tool or application to be useful, it must provide a solution to a real and relevant problem.
For instance, Google’s search algorithms make it easy to find information. Without Google or other search engines, you’d have to scroll through lots of content and find what you want. Thankfully, the algorithm uses artificial intelligence to recommend search results based on your search intent. Google’s AI predicts intent from lots of factors, including your browsing history, location, current events, and legal jurisdiction.
To understand what problems the AI application solves, try to answer these questions;
- Who faces the problem?
- Are there resources being used to solve the problem without AI?
- Is the solution monetizable?
- What value comes out of solving the problem e.g monetary benefits, comfort?
- Can the problem be solved efficiently even without AI e.g through automation?
You can ask the seller these questions. After all, they must have done some research before investing their time and money to build the solution.
2. Can you estimate the addressable market?
AI-based websites and apps are quite technical by nature. So most of them will work for very specific use cases.
The addressable market is an estimate of the number of potential customers who might need the application.
Many sellers have these estimates; ask them if they can share as you do your due diligence.
Then you can look at industry research and see if any has referenced the total addressable market. Be sure to also understand competitor products and see what sets apart the solution that you’re checking out on Flippa. You may also ask for any customer insights the seller has on their sales conversions.
3. Is essential data accessible?
AI-based applications will always need data. This raises the question of where the data will come from, and whether the data is ready for use or needs to go through a cleaning process with new clients.
Consider an application that audits local marketing and customer retention data and then offers recommendations. For this app to work, it requires access to sales and retention data. Do prospective clients already keep this data? If the answer is yes, the application will be easy to deploy. If the data is unavailable, it presents an obstacle to the adoption of the solution.
4. What are the legal and ethical considerations?
As established, AI web and mobile applications rely heavily on data and this comes with some risks that have legal and ethical implications. Consider privacy laws, ethical issues, and any potential risks.
Privacy laws
Privacy laws are not standardized globally, and this creates boundaries and hurdles for businesses that run online. For instance, the EU has its GDPR, the US has its own data privacy laws which are quite strict. But some countries in Africa, Asia, and Latin America may have more lax laws.
So if the seller of an app comes from a different jurisdiction than you, it is good to understand whether the application will be compliant when you deploy it to other markets.
Ethical issues
While some countries may not have any privacy requirements, it is still important to respect user privacy when building or deploying AI.
For instance, consider Google’s reverse image search: a program allowing you to search with an image instead of text. Have you considered what happens to the image you uploaded for the search?
Google keeps those images. While no one else will access the images, Google will use them to improve their products.
If you uploaded the picture of someone, then there’s a point to be raised about the ethical considerations of using someone’s picture to train your algorithm without their consent. Third-party data uploads to an AI-powered tool can have a slew of accompanying ethical considerations.
So to recap: AI is trained on data. What you do with that data needs to be considered from both a legal and ethical standpoint.
5. Is the application monetized?
Look at the monetization data shared by the seller. Also, consider the monetization strategy. The app might be offered as a one-time purchase or a subscription. Ask the seller if there is a longer-term monetization plan and marketing plan already in place.
If an application is not monetized, that doesn’t disqualify it from being a good investment. You should still ask to see a monetization plan for such apps.
As you explore monetization, remember there are costs to running an online business. Find out how much expertise and resources you will need to run, maintain and improve the AI-based solution.
6. How does the product deploy AI?
Note that many SaaS companies have added AI as a buzzword to their products when in reality, it could simply deploy automated analytics processes and workflows.
Generally, the role of the AI should be in the seller product description. But when this is not clear, follow up with the seller and ask them to provide this information. You can also easily filter out legitimate AI products from simple automation by asking the seller to explain how AI works in the application.
7. What features does the product have?
The features of an AI-based app or website are key in delivering value to customers. You need to test the features of the product and see if they work – the maxim ‘try before you buy’ rings true here.
You can do this if the app or website is online. Sometimes, the seller might need to set you up with a demo account so you can access the features.
Integrations and APIs are key aspects to look out for. This is because no tool works alone. If the AI-based software is a chatbot, then it should have the infrastructure in place to integrate with common business tools such as CRMs, webpages, and even social media.
Finally, understand the installation and deployment process, and how it plays out in real-world applications. Then you’ll have a good idea of how well built the product is.
8. What are the results achieved and what is the accuracy of the product?
Find out about the successful and failed deployments of the app by asking the seller. If the app has been in the market for a while, you’ll also expect to see client testimonials, ratings, and case studies.
As you will notice from this app currently listed on Flippa, the seller provides an overview that shows the average ratings from over 10 installs. In this case, the seller has also shared that customers haven’t posted reviews.
Now, the absence of testimonials, ratings, and case studies doesn’t necessarily mean the app doesn’t work. However, cover your bases by asking the seller if they have any case studies. If not, you’ll need to rely even more heavily on practically testing the product.
No AI is 100% accurate. That’s why it’s important that you have an acute understanding of how accurate or robust the AI tool is in predicting outcomes or providing recommendations. Ask the seller for this info since they tested the AI against historical data and competing solutions when developing the application.
You can then compare the accuracy with industry standards.
9. How does the application work?
Now that you know what problem the website or mobile app solves, you can proceed to test the application.
Data is at the heart of any AI solution, so that’s the place to begin:
Understanding data
All applications that deploy some form of machine learning or artificial intelligence use data. In some cases, data is the intellectual property that can give the application a competitive edge.
Find out what data was used to train the AI models. If the AI was trained using insufficient data, the output might look okay internally but when the AI solution is deployed in the wild it won’t achieve desired outcomes.
Also, ask the seller about the sources and volume of data used to train AI models. Other information you need to find out about data include:
- Management: What resources are needed to collect and manage the data.
- Scalability: What it will take to scale the application from a data perspective.
If you can, go a step further and test the application with your own data.
Data can be categorized into external and internal data:
External data
We have established that data is central to any AI solution. In cases where an application uses data from previous clients and other third parties to deliver solutions to a new customer, the data can be considered external.
AI products that use external data work almost immediately after deployment for a new client. Most marketing AI-based tools work like that.
Internal data
Sometimes, the AI tool might be deployed in highly unique customer environments. In such cases, data from external sources can’t train the AI to create solutions for the customer.
For example, many car manufacturers may deploy proprietary manufacturing processes and machines. An AI solution that helps manage production by controlling the machines, predicting breakdowns, or recommending maintenance will need to be trained with the client’s own data. So after a client buys it, the application will require a training period before it delivers desirable business outcomes.
Vetting AI products can be challenging but highly rewarding
If you take time to find out as much as possible about an AI product recommendation, you can make a wise investment decision.
In all honesty, the most robust test you can perform to verify the viability of AI solutions is to test the product multiple times. Test using multiple data sets and different scenarios to determine how well an application deploys machine learning. How you feel about the product upon testing (and asking yourself the above 9 questions) will give you insight into whether the product is right for you or not.
Sellers on Flippa are always willing to provide information and demonstrations where necessary – you just need to know what to ask. If you are a first-time buyer on Flippa, here are some pointers to help you vet cool businesses for sale on the platform.
