Non-tech organisations often have a completely different set of conditions that call for unconventional strategies for AI deployment
As we’ve got used to everyone being able to crunch numbers using a computer, we are now rapidly entering an era when computers can see, hear, and make decisions on humans’ behalf with the use of Artificial Intelligence (AI).
Democratisation of AI-based tech is now leading to even the least tech-savvy companies using this technology to their advantage. Companies operating in healthcare, travel, insurance, retail, education, and many other industries now embrace AI software development to streamline their decision-making and make workflows more efficient.
For example, Johnson & Johnson uses AI to discover new drugs and make vaccines. Bloomberg uses AI to automatically generate financial news articles based on companies’ financial reports. Costco has managed to attract millions of new customers by utilising AI to detect the most effective locations for their new store locations.
Other uses of AI firmly resemble decades-old sci-fi movie scenarios. For example, Ping An, a Chinese insurance company, uses facial recognition to detect dishonest clients. Potential borrowers can now apply for loans through an app by answering questions about their finances using a mobile camera.
An embedded AI algorithm monitors facial expressions to spot lies and figure out whether a prospective borrower needs to be further interviewed by a human professional.
Common AI adoption pitfalls
While tools that incorporate AI have become as accessible as never before, the lack of AI understanding hinders the realisation of the full potential of this technology.
Furthermore, non-tech organisations often have a completely different set of conditions that call for unconventional strategies for AI deployment. This is why non-digital companies often struggle with AI implementation.
Here are the main challenges that these companies face and ways to overcome them:
Small data
Unlike tech giants who have access to inordinate amounts of data to train their AI algorithms, smaller non-tech companies often have not enough data for powering their models. Unfortunately, there is no one-size-fits-all solution to this problem. Given that data is the bread and butter of any successful AI model, it’s critical for non-digital companies to turn to consultants on the matter.
For example, let’s say there is a bicycle manufacturing company, which is looking to implement AI to detect bike frame scratches and defects. It’s highly unlikely that such a company would have millions of dented bike frame photos lying around.
However, new sophisticated AI algorithms can generate artificial images based on a very small number of similar images, which would then be used again for algorithm training. Alternatively, companies can feed algorithms with relevant data gathered from external datasets, but it would take significant input from data scientists to make it work.
Lack of change management
AI deployment often has much more impact on an organisation than it’s expected. When employees, stakeholders, or customers are not ready for AI implementation, workflows often get disrupted in a negative way. To overcome this, companies need to think about their change management strategies in advance and ensure that everyone is on the same page regarding AI implementation.
People need to be informed how exactly AI will influence day-to-day operations and educated about the basics of the technology. Workflows need to be adjusted, retraining initiated and stakeholders notified. Non-tech companies have to deal with much more uncertainty and reluctance to change than other companies that have technological innovation at the top of their agenda.
Unrealistic expectations of AI and what it can do
Far too often, non-tech businesses struggle to achieve the same model accuracy as they expect. This is especially relevant when AI feasibility is justified based on research, where experiments were likely conducted in perfect environments that are hard to replicate in the real world.
For example, our imaginary bicycle manufacturer can be convinced about AI viability based on comprehensive research about automated AI-based scratch detection software. However, it rarely becomes apparent that such type of research is often conducted in closely controlled environments with high-quality images. However, when it comes to deployment, it becomes apparent that a manufacturer’s image quality is not sufficient, and the production environment requires dramatic adjustments to become appropriate for AI.
In this particular example, it could be possible to rely on human employees to double-check the AI system output. In essence, though, it’s paramount to conduct rigorous pre-deployment tests in an environment that would resemble real-life conditions as much as possible.
As AI goes beyond the tech industries, it becomes increasingly important for companies to start considering the technology. It will inevitably be the X-factor in reshaping how business functions like HR, finance, and customer service will work in the future.
With media often portraying AI as a Swiss Army Knife that can solve any possible business issue, some business owners are still uncertain about the technology and struggle to separate hype from reality.
In a nutshell, the technology’s most disrupting feature is its ability to make predictions way cheaper and faster than it was possible ever before. Similar to the democratisation of electricity fuelling economic growth in the 19th century, we can expect AI to have a dramatic impact on business by lowering the cost of making predictions.
With forecasting becoming a readily available business instrument, those who won’t learn how to make use of it will most certainly fall behind.
Source: e27
