[ad_1]
Chris Hillman, international director of data science at data management company Teradata, recently observed that the cost of data science and artificial intelligence teams has received increasing attention as companies seek to prove the value of their investments in emerging technologies.
However, he believes that while data scientists are capable of building AI models on a technical level, business stakeholders often hinder the success of AI projects when they do not understand how AI models work or are unable to put the model recommendations into action.
“In data science, everything is a technical problem, and we solve it with technology,” Hillman explains. “But I absolutely believe that a lot of the reasons why these things don’t make it into business processes are basically cultural, political, or human issues — not technical issues.”
Teradata’s experience in building models for many international customers shows that:
- Business executives must understand AI to support and achieve project success.
- Executives can learn better from use case examples than from a “Data Science 101” course.
- Companies should conduct impact assessments before launching AI projects.
Culture, politics, and people: barriers to successful AI projects
Hillman believes that AI project failures are often caused by business stakeholders:
- Don’t trust the results of AI models because they are not part of the process.
- Failure to translate model outputs into actual processes and actions.
Hillman explained that AI problems are not technical problems, as long as data is fed to data science and AI teams. Instead, business stakeholders often struggle to understand the technology and translate AI outputs into business actions.
Executives should be involved in the AI development process
As long as the data existsHillman’s team can successfully train, test, and evaluate AI models.
“We write the output of that model somewhere, and the job is done,” he said. “Production is when the model runs once a month and something is stuck in a table somewhere.”
However, this is also where it may fail.
“It fails because the business owner has to be involved in the process,” Hillman added. “They have to decide based on the score, ‘What is the signal?’ If I say something has a 90 percent likelihood of being fraudulent, what does that actually mean?
look: Evidence of Australian innovation in the pursuit of generative AI at scale
“If the signal is to block payments and they decide to do that, then someone has to do it. In a lot of companies, that means at least three or even four teams involved; data engineers and data scientists, business owners and application developers.”
This can result in a dysfunctional process where teams fail to communicate effectively, AI fails to impact business processes, and AI fails to create the desired value.
Business owners must understand how AI models work
The rise of artificial intelligence means all business executives must understand how these models are created and operate, Hillman said.
“They should understand the outcomes because they need to guide the process,” he explains. “They need to ask: ‘What does this mean for my customers or my business process?’”
While a technical understanding of the algorithms may not be necessary, business executives should understand the basic mathematics involved in AI, such as the probabilistic nature of AI models. Business stakeholders need to understand why the accuracy of AI models differs from what is expected from traditional business intelligence reporting tools.
“If I went to the CFO with a report and they asked, ‘How accurate is it?’ and I said, ‘It’s about 78% accurate,’ I’d probably be fired,” Hillman said. “But for an AI model, 78% accuracy is pretty good. If it’s over 50%, you win.”
“We have clients coming in and saying, ‘We want this model, we want it to be 100% accurate with no false positives.’ We have to tell them, ‘Well, we can’t do that because that’s impossible.’ If you do get that kind of model, you’re doing it wrong.”
Use case: Effective tool for training business executives using AI models
Hillman believes that business owners should not take “introduction to data science” courses because they may be “useless” in practice. Instead, he says AI Use Cases It can be used to demonstrate how AI models can serve business people more effectively.
“I think a use case-driven approach is definitely better for people on the business side because they can understand it and then you can engage in the conversation,” he said.
Tips for making sure your AI project is actually up and running
Hillman offers several tips for business owners to ensure their AI projects can move from ideation and proof of concept to production:
Conduct an impact assessment
An impact assessment should be conducted in advance. The assessment should include key considerations, such as why the organization is undertaking an AI project and the specific business benefits.
“I rarely see this in the original specs,” Hillman noted.
In contrast, impact assessments are often initiated while a project is ongoing or after technical work is completed, which can result in projects being put on hold and never moving into production.
Choosing the right use case
Although Transformer models were growing in popularity before ChatGPT, the hype generated by OpenAI’s chatbot push has prompted companies to launch generative AI projects to stay relevant. This has led to some use case choices that may be misguided.
look: 9 innovative use cases for AI in Australian enterprises in 2024
Hillman often asks if businesses can “create reports instead” because there are often easier ways to achieve business goals than creating an AI model. AI models often fail to get off the ground because of a lack of impact assessment or the wrong use case, he said.
Have strong commercial sponsors
AI projects achieve better results when there is a strong business sponsor driving them forward. Business leaders can ensure that other teams in the enterprise understand the potential impact of AI projects and that they are aligned in applying AI data to their processes.
“IT may own the technology budget, and others may own the data and security and privacy aspects, but really, the driving force always comes from the business side,” Hillman said.
[ad_2]
Source link
