AI Accelerator: Hanh Nguyen, Senior Product Analyst

In this installment, we speak to Hanh Nguyen, Senior Product Analyst who moved from cautiously testing AI to designing how it works and into a product that turns hour-long document reviews into minutes. Here’s why hallucination risk is a QA problem, not a mystery, and why the harder question isn’t what AI can do but where it belongs.

The AI Accelerator series is an ongoing collection of conversations with people across Confluence who are at the forefront of how we build and use AI.

From the intelligent features embedded in our solutions to the tools colleagues increasingly leverage to get their work done, each installment hears directly from people across teams and disciplines. Find out what’s working, what isn’t, and what they’re still figuring out. Ground-level perspectives from the people living it every day.

Q: Tell us about your relationship with AI, and how it has developed?

A: I started out slow in my use of AI, just poking at it to see if I could shave time off manual processes and analysis. But gradually I found myself building agents and prompts to handle the tedious, repetitive work. It also started filling in technical gaps that used to pull me out of my flow; things I knew conceptually but had to stop and look up. That gave me the freedom to focus on the parts of the job that required judgment. That shift from using AI to designing how AI worked for me was when it really clicked.

The first project that got me hooked was building a beta version of a tool to help streamline our proposal process. A lot of what goes into a proposal is repeated across engagements; the same core information, just slightly adjusted for context. I built something that could pull from that institutional knowledge and surface the right content at the right time. Instead of starting from scratch every time, you’re refining and customizing. Seeing that work, watching it collapse hours of work into minutes, was the moment I knew this was something worth seriously investing my time in.

Q: How are you using AI within the products you build to deliver a better solution to Confluence clients?

A: Our clients use our solutions to produce financial reports that need to be cross-checked and validated before they can be considered “finished”. So, where our solutions produce the report via well-established automation, there was still a manual process that came after. A single document could take 45 minutes or more to review, so this can really add up! 

I am Product Analyst for POINT Validation, Confluence’s agentic document validation solution. Simply put, it uses AI to automate that workflow. It ingests a wide range of document types, including PDFs, spreadsheets, and custom checklists, runs independent validation across sources, flags inconsistencies, and produces a fully cited audit trail. What used to take the better part of an hour takes minutes, which I think is a great use case for AI really delivering for our clients. Teams only need to intervene where POINT has identified something that needs a human decision.

Beyond validation, POINT can also convert unstructured data into structured output for downstream systems. A PDF capital account statement, for example, can be transformed into a CSV file ready for use elsewhere. And because the rules that govern validation can be written in plain English and adjusted on the fly, clients can tailor the workflows as needed.

What used to take the better part of an hour takes minutes, which I think is a great use case for AI really delivering for our clients. Teams only need to intervene where POINT has identified something that needs a human decision.

Hanh Nguyen, Senior Product Analyst

Q: Were you originally skeptical of AI? Do you still have any skepticism?

A: Honestly, yes; early on I was skeptical. The models were inconsistent, and hallucinations were a real problem. It was hard to trust something that could sound completely confident while being completely wrong. That skepticism hasn’t fully gone away; every time hallucinations happen, it’s a reminder that AI isn’t infallible.

At Confluence, hallucination risk is just something we treat as a QA problem, not a mystery. Models have gotten a lot better on their own, but the bigger lever is how we index the underlying data. We build that indexing around the actual structure of whatever content a product is working with, so the model’s pulling from real source data instead of just generating plausible-sounding text. We also run accuracy tests against benchmarks built for each product’s specific domain instead of generic ones, because “sounds right” and “is right” are different bars.

AI should augment what people do and handle the repetitive work but not replace their judgment. A frustration I have is the assumption that AI is the answer to everything. It’s not. There are real problems it solves brilliantly, and there are places where it just doesn’t fit — and confusing the two can make things worse. The risk is leaning on it as a default crutch instead of being intentional about where it truly adds value. For me, the work is figuring out the right problems at the right time. Where does AI genuinely move the needle, and where are we just using it because we can? That’s a harder question than it sounds, and I don’t think enough people are asking it.

Q: How are you using AI day to day, and how has this changed the way you work?

A: I am using AI across pretty much every domain.

At work this ranges from research, data analysis, scripting and design, to considering how next generation tooling can help our clients get more out of the solutions they already use.  I lean on it heavily to ideate; thinking through problems, pressure-testing ideas, and exploring directions I might not have considered on my own.

It’s hard to point to a single moment in the day where AI isn’t at least in the background. Notes from a meeting get captured and organized without the usual scramble. Emails and replies get a first pass that I refine rather than write from scratch. AI has become less of a specific tool I reach for and more of a layer that runs through how I work and think.

Q: Has AI truly surprised you at any point?

A: It has, in a few ways. Some of the hallucinations are genuinely hilarious. When a model confidently explains something that is completely, spectacularly wrong, there’s a dark comedy to it that never gets old. On the bad side, anything new at this scale is a little unsettling. When a technology has the potential to reshape entire industries, the future can feel murky — and that’s a big thing to sit with.

What gives me confidence is that Confluence has always been at the forefront of adopting new technologies, and our products have always been built around what’s right for our clients. The tools change, but that commitment doesn’t.

Hanh Nguyen, Senior Product Analyst

Q: What is the AI development you are watching right now – and why?

A: Like everyone else, I’m watching the Artificial General Intelligence (AGI) conversation closely because the pace of capability growth is real and the workforce implications are significant. The part that doesn’t get talked about enough is adaptability. Humans can adapt, but it takes time; retraining, reskilling, finding where human value fits in a shifting landscape. That gap between how fast AI moves and how fast people can adjust really has my attention. The technology will keep accelerating. The question is whether we’re setting people up to move with it or just watching them get lapped.

I think a lot about people entering my profession now. The market is shifting fast and it’s easy to look at AI and wonder where you fit. What I’d say is every major disruption has come with opportunity, and adaptability has always been the skill that matters most. I graduated during the great recession and ended up taking roles that weren’t exactly what I had envisioned for myself. But looking back, they let me build some of the most valuable skills I have today. The path isn’t always obvious when you’re on it. Stay curious, stay adaptable, and don’t underestimate the value of the work in front of you even when it doesn’t look like it leads where you hope you’re headed.

Q: How has using AI impacted your enjoyment of your role as a product analyst?

A: Oh, it has increased it, without question – but there’s a nuance I didn’t expect. I do miss some of the “figuring-it-out” moments. There was something satisfying about wrestling with a problem on your own, testing a new approach, getting it wrong a few times, and then finally cracking it. When you Googled your way to a solution at 11pm, you really owned that solution. AI shortens that loop so dramatically that sometimes you don’t get the struggle — and the struggle, it turns out, was part of the fun. That said, I’ll take the trade. The problems I get to spend my time on now are definitely more interesting than the ones I gave up.