The Challenges When Adopting AI in Business

Artificial Intelligence has gotten a lot of press and attention in the last decade. Businesses are attempting to integrate AI in order to realize its full potential, but it comes with its own set of obstacles.

AI is becoming a major topic in the business world, with companies like Google, Netflix, Amazon, and others benefiting greatly from AI and machine learning algorithms. Small and medium-sized businesses, as well as huge corporations, are all affected.

Businesses have been under pressure to adopt AI technologies in order to stay competitive. There are numerous publications demonstrating the need of incorporating AI into business procedures. Because artificial intelligence has proven to be advantageous to the successful operation of enterprises. According to an Accenture analysis, artificial intelligence may raise corporate efficiency by 40% and profitability by 38%.

However, we can't ignore the difficulties that organizations have faced in implementing AI. These obstacles make the prospect of successful AI integration appear far-fetched, if not impossible. There are several challenges in adopting AI in the business sector, for example: Computing Power, Trust Deficit, Limited Knowledge, Human-level, Data Privacy and Security, The Bias Problem, and Data Scarcity.

How AI Affect Business

Artificial Intelligence provides many benefits in the business world. The company puts hope on artificial intelligence to provide more sales, determine the most optimal marketing strategy, and also optimize the company's internal performance. AI developments for business purposes include personalizing customer shopping experiences, automating customer interactions, providing real-time and progressive assistance for business operations, data processing, compiling recommendations, trend prediction, automation, and so on.

As part of a company that provides AI development services in Indonesia, GLAIR has contributed to various AI projects for various business objectives and has aided substantial AI implementation progress to its clients. Some examples of AI projects that have an effect on business development and were developed by GLAIR include the following.

• In its project in the field of Food and Beverages, GLAIR has developed a technology with a Demand Forecasting Prediction system on several vending machine products spread across client companies by presenting product stock usage data that can increase revenue. In addition, the developed AI system is also able to make efficiency in the process of providing reduced product stock at vending machines up to 80% and has been proven to increase revenue and customer satisfaction in the mechanism for regulating product stock spread in each vending machine and adjusted to the level of customer distribution.
• In the scope of projects around Media Companies, GLAIR collaborates with client companies to develop recommendation systems that are capable of engagements in terms of the number of views and user personalization up to 200%. This is certainly also successful in increasing the revenue of the company.
• GLAIR Insurance company clients are able to develop AI models in the form of proficiency modeling for insurance product offerings that are in accordance with user profiles in the distribution of customers and their needs. This is also one of GLAIR's contributions in the development of AI that is able to match the type of product offering from the company to the appropriate segmentation of insurance users.

These projects are some of the examples of projects that GLAIR has developed to help and contribute to companies trying to implement AI to develop their business sector. 

AI Development Opportunities in Various Sectors

Currently, AI technology has been implemented in several business units and sectors that support automation and compatible human work. Novan Parmonangan, Head of AI Engineer at GLAIR said that there is a huge opportunity in implementing AI in the future, which includes several technology developments such as.

• In the Medical and Health sector, AI can be developed with Medical Image Analysis technology to identify the type of disease and its treatment mechanism. This technology was developed to analyze image data from patients with diseases such as X-rays to be analyzed with AI technology and to assist medical personnel in disease detection efficiency.
• In the credit company industry. AI can help increase revenue, reduce costs, increase automation, and help develop performance tasks that are limited in resources. The technology is credit scoring which is able to assist in the approval of the credit amount as quickly as possible according to the user profile. Previously in the industry, the credit scoring process from customers was done manually, now AI can help predict which customers will accept or reject their loan applications.
• In the phenomenon of handling a pandemic. AI can be developed into a technology that is able to detect crowds of people in the form of Crowd Detection during a pandemic. Crowd Detection was developed to detect the number of visitors and the allocation of restrictions in congested areas to prevent passive transmission of the virus from the crowd.
• In the transportation sphere, AI can be developed in License Pre-Recognition in the form of an Automotive Entry System for traffic ticketing and violation systems that can assist traffic police in taking action against traffic violators who endanger other users.
• While in the sales field, AI can be developed in the form of a Chatbot technology that can help customers to answer questions about products and product recommendations. The existence of this chatbot is able to efficiently use customer service to serve customers.

Some of the opportunities from the implementation of AI can cover various aspects of human life that can be efficient and help humans in the efficiency of tasks and scope of work that cannot be done by humans.

Challenges Towards Developing Artificial Intelligence

In the development and implementation of AI, there are many obstacles for business development. There are several challenges that need to be considered in the process of implementing AI in detail. These challenges include the following. 

• Lack of support from stakeholders. There could be many causes that lead to Stakeholders. The team itself does not want to take risks, whether adopting technology is capable and can be implemented. Doubts that the team could not understand the benefits. Why AI is important and still doubt AI in helping its business processes. Lack of knowledge of AI including in terms of benefits, implementation, and clear expectations of AI in supporting business, expectations are not clear, and lack of knowledge in AI implementation. This can raise doubts from stakeholders in implementing AI in the future. After getting support, the next problem is it's not clear how it will be executed. For AI execution, where do we start, the starting point, what the use case will be, and what the AI ​​strategy will be like. Whether to make an in-house solution or use a vendor. Create our own data team and must know the current state of the company. Does the company have to invest first? Do they have to invest in dataflow, how will the infrastructure be run? These problems need to be reviewed for the implementation process developed by the next system.
  

• Do not have regular, consistent, and appropriate data. This problem includes various components, including the following: it could be that important data is not available, the company may not store the data that is needed and must be available for AI development, the data that is owned has not been digitized but is still in paper form, and the system the format is not structured even the main problem could be that the data has not been captured at all. The data itself can be not clean, noisy data causes instrumentation errors, human errors, and makes mistakes that cause machine learning to be incompatible. If viewed from the quality of the data, there may not be a good data flow, the data is still scattered everywhere so it is not centralized, the bureaucracy in data documentation is still poor and unstructured, so the ETL data is not clear. This problem is very complex and needs to be fixed from the start if you are to build an AI architecture.
• Data Privacy and Security. The use of data is currently highly regulated by regulations, including part of privacy and security. The AI ​​application learns from the data, lest the data violate the ITE Law, including the privacy law. The data must be legal and concerned from those who have the data so that it does not violate the rules and policies related to user privacy. Then how long will the data be stored, including the period of time for storing data from the company. This regulation must be clear and well structured so that there is no wrong policy in the future that violates the misuse of data.

Apart from technical problems, there are also some non-technical problems such as. 

• Including People support was very important. Regulations must also be understood in terms of the business domain, general regulations, ITE laws, we can solve this problem from a technical point of view.
• Consultation with experts related to data regulation. This will help in providing an overview in good data documentation to the company through the transfer of knowledge from experts to stakeholders in the company. 
• Less education about AI. Many people think AI will replace humans. In fact, AI makes humans more effective. The Ai development process still requires fully human responsibilities in the loop.  For example, there is high accuracy data that humans can not control and robots in a dangerous environment which can be operated if humans are not available. Therefore, a correct understanding about AI's important benefits and how it works. Besides, rather than saying that AI eliminates human jobs, AI can also create new jobs. If there are problems that AI can help to solve, humans can focus on other problems to complete.

How to Respond The Challenges

To respond to the challenges faced in AI implementation, collaboration with several components strats from human resources, AI development architecture, to the relevance of data that will be collaborated in the system that will be built.

• Collaboration of important stakeholders in the organization.  Leaders such as CEOs, shareholders, venture capitalists, and commissioners. are important stakeholders in the company. Leaders who must lead and know the importance and benefits of AI itself. If the leader knows he must be able to educate and get advice from AI experts regarding the important components of AI and convey his understanding to the team under him. In addition, leaders can also learn from competitors regarding the implementation of AI. The important point is to study and educate. After knowing this, we know the objectives and expectations of AI that leaders want to expect from the company. There must still be a clear human in the development of AI.
• Start From Small Things. Try from the vertical section of the business that was developed to review the business's objects. So, for example, if we fail to develop the AI, it doesn't cause a big impact and the cost is not too big to cover it. Most importantly, we can identify the AI ​​environment that will be developed and ensure that the AI ​​results are good. It doesn't mean that we create a perfect end-to-end solution, but that it is progressing, running, and its performance can be measured to be improved over time. The team can be more confident and can handle more complex use cases if the collaboration between teams can be managed better from the beginning.
• Advice from an AI Expert. We need advice from AI Experts so that we know the readiness of our AI adoption, the type of AI development that will be planned: in-house or vendor, the needs of the AI ​​development itself, and the core business that the AI ​​will develop. AI experts also can help with many suggestions, in terms of data flow, BI, and whether we have a framework, clear architecture and readiness to operate AI.
• Getting Started With Structured Data ETL Documentation. Start collecting minimum data and documented flow of ETL, we can start analyzing data starting from clean data, if necessary, digitizing data can be transformed and make it easier for AI to be developed.
• Manage Data Privacy Authorization. Determine data regulations: in the form of access rights, data use, and consent of sources, differential privacy: in the form of a mechanism for publicly sharing information about a dataset by defining the patterns of groupings within the dataset while keeping information about individual dataset participants private , and a zero knowledge policy:  approved which is a mechanism by which one person (the prover) can demonstrate to another party (the verifier) ​​that a particular assertion is true without providing any extra information other than the fact that the statement is true.

What We Need to Prepare to Implement AI

There are several technical and non-technical things that we can prepare before developing AI in operational or business services within the company. But broadly it consists of.

• Investment in AI starts from knowledge that is able to answer all questions related to AI implementation from stakeholder readiness, AI use cases, implementation and development, data, and so on.
• If we have invested in AI knowledge. Leaders must make an appointment with someone as a PIC or person in charge of the AI ​​project. So there is progress from the development of the AI ​​itself. Stakeholders can review the progress. and the running of the planning, execution, development, operation to maintenance processes can run well.

Today's AI is widely adopted, AI is important like when the internet or cloud computing was booming. At this time we have to handle this phenomenon seriously.

AI education and AI implementation are needed in enterprises, because competitors and industries may die if they don't adopt AI. So that a company's business may die because it is unable to compete. For example, with offline stores vs e-commerce, as well as competition from the conventional and digital entertainment industries. AI implementation is needed for the sake of interest. We need AI for all sectors and including its implementation as simple as the internet to sustain business.