Shensi Ding is the CEO and co-founder of Merge, a unified API platform that helps companies connect existing apps and systems with AI. Integrating AI offers many benefits: enhancing performance, automating tasks, and improving decision-making. But as Shensi points out, it’s not without looming challenges. Product leaders often wonder, Are my data and systems compatible with these new tools? Is my product team well-versed in AI technology? Can my organization’s data governance framework mitigate exposure to risk? Shensi joined co-hosts Sean Murray and Dan Sharp to provide her insights in what Sean referred to as “a firehose of information and an AI integration master class.”

Here’s what we learned:

What Does High-Quality Data Look Like?

High-quality data for AI integrations means ensuring all data is accurately synced, is normalized into consistent models across platforms, and accounts for potential edge cases. This is essential for reliable AI functionality and avoiding errors that stem from inconsistent or outdated information sources.

“There’s a lot of things that can go wrong with the AI integration if you haven’t synced all your data,” Shensi explains. “You also want to make sure that it’s normalized properly. When you’re integrating with multiple platforms in a category, you need to have some kind of canonical data model that you end up normalizing it into. And that’s very difficult to do.”

Integrating AI: Challenges and Roadblocks

Many API providers have limitations, including outdated documentation and lack of real-time update notifications. These situations often require full data synchronization (over and over again), which leads to performance issues.

Product managers need to be able to adapt to these limitations, Shensi explains. “Unfortunately,” she adds, “for each platform it becomes a little bit different, and you have to be a but hacky for how you solve it” and come up with creative, manual solutions to maintain data accuracy for AI use cases.

Skills Development for Effective AI Integrations

The simplest, most effective way to get comfortable with AI is to just dive in, Shensi says.

“I just think the best way is just like testing it out and doing it yourself. Dive deep, actually understand how an integration might be built.” She suggests starting with hands-on testing using tools like Postman to authenticate and then explore API endpoints. Teams should experiment, set up sandboxes, add sample data, and perform load testing to build fluency with integrations and account for real-world edge cases as part of their learning process.

Trends and Expectations for Widespread AI Adoption

We may already be beyond the point of consumer acceptance of AI’s role in building software – these days, they expect it. Products that lack AI integration may be seen as outdated. As AI becomes normalized, both users and businesses anticipate AI-driven insights and automation as standard product features rather than optional add-ons.

“I think everyone expects it now. And if it doesn’t, it’s kind of weird.”

Catch the entire episode for Shensi’s thoughts on these important topics:

• The difference between a typical product owner and the technical product owner?
• Why product leaders should become comfortable with AI tools and how to use them.
• The new product recently launched by Merge that enables agents to safely make calls to third-party enterprise tools.

Our conversation with Shensi Ding is the third of five episodes the team recorded at INDUSTRY: The Product Conference. We’ll publish the final two in-person episodes over the next few weeks, including chats with Axel Sooriah (Atlassian) and Michelle Parsons (Lex). Great insights from outstanding product leaders!

The post 175 / Seamless AI Integration: Challenges and Opportunities, with Shensi Ding appeared first on ITX Corp..