Q&A with Chee Chew, Chief Product and Technology Officer, Zapier, explores how AI orchestration is reshaping automation, empowering knowledge workers and redefining governance in modern enterprises.
You’ve been at Zapier for about a year and a half. Why did you join the company?
Overall, for me, there are three main reasons. First of all, I’m a bit of an automation nerd. I’ve been nerding out on home automation for many, many years and business automation is, of course, in the same realm. I think it’s very powerful, very creative and has the potential to be approachable by everyone on the planet. It doesn’t require a developer to do amazing things. So the notion, the concept of automation, is something that I’ve been very, very passionate about for a long, long time.
Number two, the onset of AI, of what it can do or what it unlocks in terms of approachability, that it can make it more accessible to everyone, as well as expanding the realm of what automation can do, I think AI really unlocks this on a whole new level, and I’m super excited about that.
And number three is that I got excited about the company itself, the way that the company operates, how forward-leaning it is, how fast it moves. Its willingness to make bets on the future and reinvent itself, I find that’s the right type of company and so I got excited about the company as well. You put those three together, and I became super excited about joining.
How does Zapier help companies to use AI to automate workflows?
Right, so expanding on what I said before: there are two ways Zapier really helps them. First of all, if you think about what AI is doing, it is transforming the prior data movement role and data transformation role from traditional automation and traditional IPaaS into one where, instead of just moving data from one SaaS system, transforming and putting in another, AI is allowing you to take the information, analyse it, make decisions on it and then orchestrate downstream systems as well as people, so this changes from data movement to decisioning. That’s transforming what can be done.
It’s also transforming who needs to do that work, so that pushes out the orchestration work to the knowledge workers who are most responsible for those decisions, the marketing decisions, the AI orchestration for marketing. You want the marketer to own that because they’re actually making marketing decisions, not just moving data around. That’s unlocking what can be done, the decision-making that can be done with Zapier now. And because of the way we’re approaching unlocking builders, we’re making the building itself agentic, making it conversational. That’s allowing knowledge workers, less technical people, to be able to actually do that orchestration.
What are some use cases?
We see use cases in really every department at this point. Whether it’s lead routing coming in, so handling leads that you want to route to sales, making assessments and judgment over who it should be assigned to, and how it should be handled, that’s one thing that can be automated.
From the legal teams, we’re seeing folks take incoming contracts and automatically apply first-pass redlines. That’s another example of something that can be done in the legal department, pulling in real decision-making and judgment into it.
From HR: onboarding is an example of helping people onboard into the company. Or, of course, an IT organisation can automate a lot of the tickets it gets. And more and more, we’re actually seeing tickets handled end to end without a human involved, thanks to AI orchestration.
Everyone is talking about AI governance. What are some best practices?
The honest truth? AI governance isn’t something you bolt on after the fact. It needs to be baked into how you build and deploy AI from day one.
I think about it in three main areas.
First, security has to be foundational. This means implementing proper access controls so your AI systems respect the same permissions your users already have. If someone can’t normally see certain data, your AI shouldn’t be able to pull it for them either. And if you’re using customer data to train models, that data needs to stay completely isolated. No mixing it with other customers’ information.
Second, you need visibility into what’s actually happening with your data. People want to know where their data is going and how it is being used by AI. That means comprehensive audit trails, centralized monitoring and clear explanations when AI generates content. For high-stakes decision-making, keep humans in the loop to approve everything before publication.
Third, and this is where many organisations get it wrong, don’t create a separate governance structure just for AI. Extend your current governance structure. Most companies are moving toward centralized AI governance, but that doesn’t mean rigid command-and-control. The smarter approach is adaptive governance: tighter controls for high-risk use cases, more flexibility for lower-stakes applications.
The goal isn’t to slow innovation down. The objective is to create guardrails so people can move fast, knowing the safety nets are there.
How should CIOs work with line-of-business leaders on AI transformation?
Well, it’s all about this balance of empowerment and governance. The most successful ones we’ve seen are those in which the IT organisation facilitates and empowers business owners to incorporate domain-specific decision-making with AI into workflows, while ensuring safety, compliance and protection of the company. That balance of empowerment, knowledge, education, governance and oversight, because catalysing it is where we have seen CIOs and AI transformation officers achieve the greatest success.
