Dr. Pinnaree Tea-makorn, Lecturer at Sasin School of Management, discusses how startups can use AI to improve their operations at the Sasin Bangkok Venture Club General Meeting. Here are some points to consider:  

Consider what AI can do for businesses and the domain expertise of your startup

Dr. Pinnaree said that to determine the sweet spot between your expertise and AI, you have to consult AI experts and your domain experts, like the executives of your business. When defining the domain of your business, consider consulting the executives before implementing AI in your operations. Another point she raised is that businesses should not try to transform or implement AI on a grand scale. She advised businesses to think about optimizing tasks rather than automating jobs or changing the entire system. “Instead of using AI to completely replace call center persons, you can use AI to help the routing before sending your customers to the right department to reach a human representative or use AI to transcribe the conversation and use that to improve the caller’s experience. Instead of replacing the radiologist unit, think about how to use AI to diagnose X-rays much faster, or use it as a second opinion for radiologists to consider,” said Dr. Pinnaree.  

Think about your business value and consider the pain points in your business

“You can have the latest model of GPT-4, but if it doesn’t solve your business pain points, it doesn’t have value at all if you were to add that into your business,” said Dr. Pinnaree. She added that even though AI is driven by data, you can make progress without big data. Instead of using millions of data, in some tasks, you can use a hundred or a thousand data points. “For example, you can train an AI model to recognize dogs with only several hundred photos of dogs instead of millions,” she added.  

Do Your Due Diligence

“Before you start a project, you think about what the AI can do, what is good for your business, do your due diligence on both sides, and find the intersection. That is the sweet spot for us to use AI,” Dr. Pinnaree said. Moreover, ethical concerns are a very important issue as people are concerned about how and where AI is going so it is crucial to think about the impact that AI would have to make your project sustainable. In order to enhance your startup with AI, Dr. Pinnaree divides AI usage into two considerations:

• AI that is applicable to everyone: Analytics that could be performed on any type of data, in marketing, sales, inventories, and customer service. Examples include using AI to analyze customer’s responses, facial recognition to do access control, and quality control for manufacturing sectors.
• Using AI to create business opportunities: New technology can create new business models and types, providing businesses advantages to tap into the newest technologies.

Focus on AI to strengthen your vertical industry

“You got to know how your data behaves in order to be able to create a successful AI model. You can’t just throw your data to your IT team and say create me some great algorithms without communication with the business side,” she emphasized. In addition, Dr. Pinnaree advises startups to foster academia-industry partnerships to accelerate development. Collaborating with academia and universities to gather insights and research into the newest technology is an opportunity for ventures to create a lot of new and successful business opportunities. “On the flip side even if you don’t collaborate with universities, you have to have your own Research and Development team to keep up with technology,” she added. Dr. Pinnaree concluded with a few points on the Risks and limitations of AI:

1. Machine learning systems have low interpretability
2. Machines may have hidden biases from the data: For example, if your company hires more men than women, the CV pool used in AI algorithms will select one gender over the other based on the past data you’ve given.
3. Neural network systems deal with statistical truths rather than literal truths: AI gives the most likely answer but not definite answers.
4. Diagnosing and correcting exactly what’s going wrong can be difficult: Scientists are still trying to find how to debug the AI system.