Irina Kolesnikova
April 29th, 2022
For those business leaders who don’t have a computer science background, it can be challenging to keep the focus on artificial intelligence implementation. Statistically, more than 85% of the AI projects fail, and it costs companies millions of dollars of wasted investment and a lot of time. Luckily, we have colleagues who can share their AI solution experiences.
We’ve conducted a series of interviews with startup founders from different fields, asking them what they wish they had known before implementing AI solutions. Check their experiences of what should have been done and learned before AI project implementations to smooth the path for future leaders thinking about using AI as part of their process.
Artificial intelligence (AI) is often pitched as a magic wand, an amazing solution to a lot of problems. Yes, artificial intelligence looks pretty much like the closest thing we have to magic, overcoming or at least improving challenges. When dealing with your first AI project you can easily discover that there are plenty of ready, out-of-the-box AI solutions on the market. But, for startup founders who think that it’s a shortcut to receiving the quickest result, we have disappointing news: there are so many AI solutions, that it’s easy to get lost.
So first answer the questions about your business priorities, then about your AI project.
Hyperhuman is the omnichannel video content platform for the health & fitness industry to offer the most streamlined path for creating and distributing quality video content on any channel and format. Create once, publish anywhere.
Its mission is to make the world’s health & fitness video content universally accessible, reusable, and licensable.
The startup is founded by ex-Fitbit product leaders around a top team of tech experts and former professional athletes.
“I think AI is such a hyped topic. And everyone can think of so many things that artificial intelligence can do for them and their business. But it’s important to have a clear understanding of the AI technology landscape and what problems can be solved by AI in order to evaluate your options.
Then you need to invest, prioritize and iterate. When we started, it was hard for us to navigate around all the noise and all the options to find the right setup for success and properly use AI to satisfy our business needs.
Over time, you understand more and more how AI can help and make your business and product unique.
Here are some questions that you should ask yourself:
If you have all this information, then you can make the right trade-offs and celebrate a true AI success story.”
“Learn your customers and their pains and needs”. Everyone starting their own business knows this hackneyed common truth. But it appears that knowing the needs and pains of all stakeholders brings bigger business value to an AI project.
Talk to all parties interested, this input helps to identify sources of high-quality data.
MAX Sports Health is a digital health platform and mobile app aimed at helping people adopt the “whole-person” approach to health that elite athletes have been using for years.
It is both an educational tool, and a full suite of health resources ranging from custom curated content and health improvement assignments, to 24/7 Artificial Intelligence chat that can assist with most health-related issues, and a toolbox for all of your wellness needs.
“We started looking closely at youth sports and that was our target customer from the beginning and we evolved so much, especially over the last year.
We are really talking to educators and businesspeople, teachers, superintendents, and principals of schools, all the people who will be on the other end of the app and listening and finding our direction based on our feedback.
The process is no different for the product, and it is no different for artificial intelligence.
Ask those questions, honestly, I would not change this process. Even though it has been difficult, it has been incredibly important because we are building a business to serve people, so we need to listen to them, take their feedback and use that to build our business.”
In coming up with your big idea, think of the architecture ahead. It could be tempting (and sometimes even convenient) to get rid of all complex or exciting unnecessary features but starting simple and planning for the future is the best idea. Your AI solution strategy should better include (or keep a place for) all AI models you were thinking about from the very beginning, as fitting a new feature in later could be a backbreaker.
Start small, plan for the features, collect data, test, and try implementing them when the data is ready, and the resources are available. You will always have time to go back to simplicity if needed.
La Koketa is a digital wardrobe, a personal stylist and a shopping app for iOS that solves the “What to wear?” problem.
La Koketa helps users find the perfect outfit for any occasion out of their own clothes by suggesting personalized outfit ideas thanks to the exclusive back-end AI algorithm.
La Koketa app reduces your investment in clothing by showing how new items fit in your closet before making a purchase, thus making shopping smart and easy.
“At this point, I kind of regret not adding a class-specific feature that could make the core experience more interesting and better, I would say.
We decided to get rid of this feature in the beginning; it was like a canvas where you can put different items and make collages.
This is very relevant for fashion-oriented users, but we decided just to get rid of it because it did not fit within the concept of the algorithm.
We would have to provide a separate space for users to experiment and to be creative and make collages, which is something that would have made a bit of a mess. Now, as I’m thinking back, I would love to have it because it’s fun, but it’s not fitting with what we did as a final result. I do not see how it is going to fit anyway, but I would love to have it.”
We live in a world of endless learning processes, a feeling that becomes even stronger when we talk about AI solutions. Some end-users could be hard-nosed toward the innovation.
Research the end-user’s reaction and think about the onboarding process and support you can provide.
upmedic is a web-based platform for creating, storing, and analyzing textual medical documentation.
It solves the problem of time-consuming unstructured medical documentation by using templates that act like TODO lists. Thanks to this upmedic enables clinicians to create high-quality medical reports that contain standard phrases.
Their template library is created by experts that use the software every
day in their work.
One thing is to develop an innovation in the field of healthcare, another is to have it adopted among doctors, as it is often introducing changes to the way they have been working for years.
The initiators of the discussion about deploying an innovation are very often the owners of the facility because they look for ways to serve more patients in a unit of time.
After they are convinced, the doctor has the final word because he or she is the end-user of the innovation, so they need to feel the clinical added value from the very first moment they use the innovation in order to see the potential returns from investing their focus into it.
That is why we are making the onboarding process as seamless and efficient as possible, taking into account the human factor that everyone has slightly different expectations.
AI models need the data to learn from. Finding clean and usable data answering the requirements of both machine learning and business goals can be a challenge.
Ask the tech team or partner questions about data governance before you implement AI.
Tryp.com Traveltech is an AI platform that enables multi-destination traveling, creating trips at the best price, to anywhere in the world, using AI.
They have more than 28 million places to stay and 6000+ possible destinations.
By using real-time pricing and an advanced algorithm, Tryp.com will find the perfect trip with flights and hotels included.
“AI is a huge buzzword, startups feel like they need to say they are AI, that they have AI, and they also think AI solves everything. It’s like this magical thing that will solve all your problems. And if you are a business person, you just say: “I’m going to use AI to solve this”. So, it’s the problem with buzzwords, people want to use them.
They don’t understand, where I am going to get my data from.
Many times I get questions like what are your data sources? How are you going to write the scripts? These are very good questions to ask.
On the contrary, sometimes you can spend time discussing how we are going to technically do this and that, and then you come to reality.
I think the best solution is a collaboration of businesspeople with a very strong technical team.”
The startup doesn’t need to be an artificial intelligence developer itself. To boost business value, it is much easier sometimes to outsource machine learning to a partner with years of experience and successful projects in their background than to devote resources to educating an in-house team to solve your specific issues.
If AI systems are the core of your product, then it is probably handy to have an in-house team, even though growing it could be time- and resource-consuming. In this case, data scientist’s consultancy could bring data insights or a new vision of machine learning technologies needed for AI development.
Finding an AI partner or consultant will benefit your business
Hepta Airborne provides power grid operators with an easily deployable end-to-end inspection suite and analysis tools.
Their software produces a birds-eye view of the grid, its most critical parts, and detailed reporting for work crews.
Hepta’s solution helps power grid operators to have faster inspection cycles, find more relevant defects and save on inspection costs.
“Collaborating with MindTitan helped us to kick-start the AI development faster,
we had experts in-house but with our partner MindTitan, the processes of creating the data for the AI models accelerated more than if we would get a new person in the team and train them for our use case.
Thus, the process developed faster using MindTitan’s help.”
Here is the list of the recommendations for a better business outcome after machine learning implementation:
