When ChatGPT entered our collective consciousness in 2022, no one really expected it to shift public perception or showcase the incredible promise of artificial intelligence.

The tool was only in its infancy stage, after all. And yet, here we are, only two years later: a testament to the AI space’s rapid advancement and promising returns on investment.

It’s only natural, then, for organizations to implement AI-powered solutions in the hopes of producing equally remarkable results. But, how do companies navigate this intimidating, ever-evolving terrain? In partnership with Kickstart Ventures and Deeplearning.ai, Swarm held AI in Action: an event that brought together founders, builders, and key decision makers all looking to answer this very question. 

Covering these topics were Tim Santos, Joshua Arvin Lat, and Michelle Alarcon, three key figures in the local tech scene who brought their interdisciplinary expertise and wealth of experience to this crucial conversation. 

Below are four takeaways from their panel discussion:

#1: AI is not as easy to understand as we think.

Before the recent AI boom, people adopted a conservative standpoint when engaging with data science projects. “People used to think it was hypothetically good but lacked engineering capabilities,” explains Tim. But after the rise of ChatGPT, it’s become, as they say, for every all.

While the proliferation of knowledge about the tool should be a signal of progress, Michelle says it’s not always a good thing. “Now, everyone thinks that it’s so easy; that all you have to do is feed [a tool] your data or talk to it. We have to transition back to tempering this excitement so people don’t get turned off like before.”

Introducing AI to an organization or building an internal solution goes beyond planning the model’s mere infrastructure: “You have to plan it all out way ahead of time, including security, ethics, and governance,” explains Arvs. One example he says that a lot of business owners don’t always think about: what if a tool meant for the company is suddenly utilized by a team member for an external project? How do you deal with the costs and resources this will use up?

“We have to slowly implement proper education and we see now that [people] are doing their homework to get their data prepared, better structured, and correct,” Michelle observes.

#2: Start with the pain point, rather than the solution.

With the right tools and team, there’s no limit to the AI solutions that can be built in service of a company. But once again, Arvs thinks this comes with a pitfall of its own. “Sometimes, leaders and owners are tempted to build what people tell them to. The solutions could end up being too expensive that it might not make sense to execute it,” he observes. 

This could stem from starting with an AI solution in mind and then trying to force a business problem that it could address. “Some of them will look for metrics that are AI or machine learning-related, when in fact, they can map it back to simple business metrics.”

Michelle also encountered similar scenarios, as the head of her own consultancy firm. “[Clients] sometimes do not have the right statement of the problem. They should ask themselves: Do we just want to try out the technique, or is there something in need of an enhancement that the model can specifically address?”

#3: Invest in AI as a strategy, not a tool.

AI develops at such unprecedented speeds that by the time a solution is ready for implementation, it might already be obsolete. In true Taylor Swift fashion, Arvs managed to write two books about machine learning during the pandemic, saying: “Right now, one of the books deserves an entire second edition because of how much the generative AI space has changed so much.” 

Rather than using it as a one-off tool for selected projects, it’s wise to incorporate it into the company’s holistic strategy, using it to develop business or domain understanding. “From an implementation and planning standpoint, leaders should also know the limitations and actual performance of the AI solutions,” Arvs warns, stressing the importance of research and intentional planning. “Sometimes, what’s on the internet doesn’t encapsulate the limitations of AI solutions.” 

This could involve something as minor as checking the news for the latest developments, upskilling every six months, or even instating a chief AI officer dedicated to formulating the company’s AI strategy.

#4: Don’t leave humans out of the decisions.

Online fear-mongering may get the best of us from time to time, but one thing’s for sure: regardless of how advanced an AI solution is, it cannot call the shots and make the changes themselves.

Michelle uses a client project as an example: “Let’s say [they] already have a perfectly functioning machine learning model for predicting employee attrition. The results will be very contentious and will still require a lot of human discussions. You don’t just deploy it without thinking or allow it to decide or terminate employees.”

What remains necessary is getting the right people: knowing who to hire and how to develop them further once they make the team. This is where platforms like Swarm come in, connecting exceptional tech consulting teams to businesses looking to work on transformative projects – whether it’s building infrastructure or launching a new feature using generative AI. Today, Swarm also offers Office Hours, where founders and innovators can book a call with our in-house AI experts to build their business’ strategy.

Needless to say, the AI hype is real and here to stay, thanks to the tangible value it provides. “The performance level, quality, accuracy are all much better now,” Arvs details. “The innovation and announcements and research papers and open-source solutions… it’s like there’s an upgrade every week.” While it’s always intimidating to navigate unfamiliar (and in this case, ever-evolving terrain), companies can easily be champions with a robust strategy and reliable team.