AI can ship code or flag a risk on its own now, but when it gets one of those wrong, who actually owns that? Danny Fields and Ee Lyn Khoo run different halves of Avalara. He's CTO, she's Chief People Officer, and they both agree that the solution for AI accountability is a joint initiative.
"I don't think there was a single light bulb moment," Ee Lyn says. "We keep landing in the same place, even though we come from different directions, which is that you can't talk about AI without talking about talent."
On this episode of Keep Moving Forward, Danny and Ee Lyn trace what that convergence actually looks like in practice. For them, accountability means a human always signs off on AI’s work. Nobody should get blamed for surfacing what went wrong. And, above all, someone in leadership needs to own what happens next, not the engineer who found the problem.
For Danny, accountability starts with a hard rule. AI can do the work, but a human has to sign off before it counts. In Avalara's tax filing workflow, AI prepares the return, "but before it's submitted and filed with the jurisdiction, the customer has the opportunity to review it and click the approve button." The same rule holds for any critical step inside engineering, where guardrails limit what data an agent can touch and what tasks it's allowed to run.
When something still breaks, Danny runs what he calls an RCA/CAPA review, which is root cause analysis paired with preventative action. He opens this review by telling engineers there's no blame in the room. "The truth always comes out about what went wrong in the technology," he says, because a team that isn't afraid of the meeting will actually explain what happened instead of covering for it.
Danny has run that review for eight years, since before AI entered the picture. The same process that used to catch a bad code deploy now catches an agent does something wrong.
The obvious risk in a no-blame process is that it quietly becomes a no-accountability process. "Resilience isn't about just moving on, it's about coming back stronger," Ee Lyn says. "The only way you can prove that is if the team works differently the next time."
Danny's review gets the facts about what broke. The real accountability sits with the leadership team that has to act on what the review found, Ee Lyn says, not the engineer who reported it. She sees that responsibility moving up and out.
"Managers will own the outcome of the human-agent combination," she says, describing a shift that reaches beyond engineering into "a shared accountability across engineering, IT, and everybody who's driving outcomes for our partners and customers every day."
Two different departments, two different vocabularies, and one answer. Accountability has to be designed into how the work gets checked, not assumed because everyone means well.
Danny Fields:
Our engineers are allowed to experiment, but at the same time, everything they do, we have to follow strict processes, stay within guardrails.
Make sure the technologies we're building. The agents only have access to certain data and making sure we keep humans in the loop.
Ee Lyn Khoo:
If AI is infrastructure instead of a tool, if my job changes at the root, a tool is something people opt into. Infrastructure on the other hand, is something everybody depends on. And so in that case, I'm not planning for just a handful of AI roles anymore. I'm planning the whole workforce around it.
Gemma Versace:
Hey everyone, and welcome to Keep Moving Forward, the podcast from X-Team for tech professionals focused on growth and leadership in a world that never stops innovating.
I'm your host, Gemma Versace, Chief Client Officer at X-Team.
Danny Fields, Chief Technology and Customer Operations Officer at Avalara, and Ee Lyn Khoo, the company's Chief People Officer, run different halves of a business that builds the tax compliance technology behind transactions at massive scale. Avalara has used AI in its products for over a decade, long before generative AI became the industry's obsession. Danny and Ee Lyn keep landing in the same place from different directions: you can't talk about AI without talking about talent.
In this episode, Danny draws a hard line between task AI and workflow AI. He says task AI is the one-off prompting you'd do with a chatbot, while workflow AI is what runs Avalara's compliance processes end to end, with a human still checking the final step.
Ee Lyn explains why pressure doesn't build performance and design does, and what she had to redesign first to get her team there. We also get into Avalara's no-blame process for when an agent gets something wrong, and why Danny's idea that AI is infrastructure, not a tool, changed who Ee Lyn plans and hires for.
Let's get started.
Well, welcome Ee Lyn and Danny to Keep Moving Forward. Thank you so much for joining us today.
Ee Lyn Khoo:
Thanks for having us. Happy to be here.
Danny Fields:
Hi, Gemma. Great to be here.
Gemma Versace:
Great. Let's get straight into it. We do like to start with asking all of our wonderful podcast guests just to tell us a little bit about yourself, a little bit about your background, and what has led you to where you are today.
Ee Lyn Khoo:
So I'm currently at Avalara where I'm the Chief People Officer, and I've been here for about four years. And prior to this I was fortunate enough to have a career in human resources that spends different industries from a company called General Mills to about 13 years at Amazon, and then most recently at a property tech company called Redfin.
Gemma Versace:
Fantastic. Excellent. Thank you and Danny.
Danny Fields:
Hi, Gemma. Well, I'm Irish. I've been in the US for a little bit over 30 years, and my journey to Avalara began at Oracle. I spent over 20 years working with large enterprise companies, and from there I worked, I moved to a security company where I learned all about securing hardware devices. And then what attracted me to Avalara was two things.
One, the culture of the company, and secondly, the technology challenge. And we want to be part of every transaction in the world. So from an engineering point of view, the systems we have to be to build, to be able to support that scale, that level of speed is just something like amazing to work on if you're an engineer.
So that's why I came to Avalara.
Gemma Versace:
Fantastic. Wonderful. And for the listeners listening to this podcast, you have heard right, we are lucky enough to have two leaders from the same business joining us today, and it's gonna be a really great conversation talking around the synergy in the intersection between the people and HR and process as well as technology and all things AI as well.
So. We'll start off with a question for you both. You both have said that the outcome is determined by design rather than effort, and you run both different halves of the same company. How did you realize that you were making the same argument, and when that light bulb did go off, what happened next?
Ee Lyn Khoo:
I can kick us off. I don't think there was a single light bulb moment. Danny and I have been on the same page about AI's potential to unlock business value for the whole time that we worked together, which is about four years. We have a shared North star in driving profitable growth for Avalara, and we keep landing in the same place. Even though we come from different directions, which is that you can't talk about AI without talking about talent. So on one hand I'd be, you know, redesigning how we hire and develop talent at Avalara, and Danny will be redesigning how the engineering organization works. And so every time we compare notes, we realize that we're solving the same problem.
Danny Fields:
Right. And we didn't wake up one day and just say, oh, AI is here. Over the past couple of years, the world has changed and everybody is focused on AI and how to bring AI into their company. At Avalara, we've been using AI for over 10 years. We've been using different machine learning techniques all through our products, and it's only over the past two years that the conversation has really focused on generative AI.
When we started our big discussions internally over the past two years, it wasn't just about efficiency. How do we make the company more efficient? It was really about how do we use AI for growth? And that's the conversation Ee Lyn, myself and the other members of the executive team have been having.
It's not just an engineering thing, introducing AI, it's something that goes across every department in the company and across every employee.
Gemma Versace:
Yeah. Wonderful. And we've just recently released a white paper on this and the two of you are the. Are the two perfect people to have available to have this chat around a business that is pushing heavily with AI, and as you've said, you've been doing it for many years now.
But around the creating the AI strategy. But also then making sure that you've got the right talent and people to be able to support that strategy. And what we have found and what clearly is between the two of you is that you have such great alignment and you are very very connected and very collaborative around making strategic decisions, but also understanding.
Where the direction and you mentioned it too, Ee Lyn, around, you know, being aligned on what the North star of your business looks like.
Danny, you draw a hard line between task AI and workflow AI. Tell our listeners where that line is.
Danny Fields:
For me it's very simple. So everybody knows how to use Claude or ChatGPT or Gemini, or your favorite LLM but most people are just asking it, how do I do something? You're when you ask the question, it's a simple task. But our products inside of Avalara, they're built around workflows. So everything we do.
In compliance, it's a set of steps, one after another. Do step one, then step two, define steps to file a return or remit sales tax or process a notice. So these workflows or these compliance workflows lend themselves to being agentic. So agentic means an agent is performing a set of tasks, one after the other.
So everything we're doing inside of Avalara is we're building agentic workflows to automate the activities that would be performed in a tax department. So imagine if you're filing a tax return. You have to collect all of the sales tax transactions from the business. You have to prepare the tax return, you have to check it, and then you have to submit it to the Department of Revenue.
All of those steps put together, create an AI workflow, and that's what we're doing inside of Avalara. That's the technology we're building.
Gemma Versace:
Fantastic. And being the industry that the financial services taxation that you are I can imagine that, that is incredibly regulated. How do you make sure that there is always that level of accountability and a human in the loop to make sure that. The whatever AI workflows or tooling that you are using is correct, and that you do get that opportunity to be able to to check the work that is getting created.
Danny Fields:
You know, that's a great question. From the beginning when we started building our new AI products, we made sure from right from the beginning to keep the human in the loop. So the example I just gave about filing a tax return. We will use software and AI to do all of the activities required to prepare the tax return, but before it's submitted and filed with the jurisdiction.
The human, the customer has the opportunity to review it and click the approve button. So we always keep a human in the loop. And it's the same with other tasks that we perform inside of engineering. Using AI technology if there's a critical step in a process, we will always have a human in the loop to double check with tax and compliance.
You can never make a mistake. You always have to be correct.
Gemma Versace:
Yes, absolutely. And some good confidence for not only your current clients, but also potential clients listening as well around how you can give that level of assurity, which is fantastic. Ee Lyn, you say pressure doesn't build performance. That design does. What did you have to redesign first?
Ee Lyn Khoo:
I think Gemma, the first thing we had to redesign was what good look actually looks like. We learned that if you tell people to just use AI more, everybody guesses differently. Some people fake it. Some people ignore it and some overdo it. And so before anything else, we defined what success look with AI looks like for each role type.
And that's the real difference between pressure and design. Pressure tells you to move faster, but doesn't tell you where design tells you what you're building towards. So that speed actually has somewhere to go. And so once we clarified that bar. Hiring, onboarding and training, for instance all started to point towards the same target so that people don't have to invent their own version of what it means to use AI well.
Gemma Versace:
I love that saying that you've just said that pressure tells you to go faster, but it doesn't tell you what specifically to do. I think that's a really interesting point that that you've called out there and some really good advice to HR leaders that are listening to.
Today as well. Danny, back to yourself. So, an agent files something and it turns out that it was incorrect. Who owns that inside your organization? Where do you go to have a chat around making sure that doesn't happen again?
Danny Fields:
It's engineering. So in engineering we're the ones building the agents, and we're the ones that make sure that strict guard rails are in place and we have to make sure agents only have access to the data they're supposed to have access to. We have to make sure they can only execute tasks that they are allowed execute and so on.
So we're a little bit like Dr. Frankenstein. If the thing we built misbehaves, we've gotta go fix it. Now if something goes wrong, we have a process in place to make things better. Whether it's AI or non-AI. In the engineering team, we're the ones building the software. Software breaks, and when it breaks, we have a process where we will dig in and we will ask questions about what happened, what went wrong?
Why did it go wrong in the first place? And then what we do is we put process changes in place or technology changes in place to make sure that problem never occurs again. And when we're talking with the engineers to find out what happened, we have a culture of not blaming because. If the engineer can come into an environment that's safe and they can just explain what went wrong, they will open up explain all of the detail, and then we'll put technical fixes in place or process fixes in place to make sure the same problem doesn't occur again.
And that's a process we've had in place over the past eight years since I've been at Avalara, and it has helped make all of our products better.
Gemma Versace:
Yeah. Fantastic. I love that. I think it was a little bit similar to the Jim Collins Good to Great that when something does go wrong, to do an autopsy without blame, because obviously it does allow for a safe environment where people feel that they can genuinely say where or if they were involved in it going wrong in a safe way, but also to make sure that there's some really good learnings and as you said, to ensure that it doesn't happen again in, in the future as well.
Danny Fields:
We call it a root cause analysis in engineering. If I use the word autopsy, I think I had to freak out my engineers.
Gemma Versace:
Yes. Very good point. Very good point. Danny. Ee Lyn, you have quoted Danny on AI being infrastructure rather than a tool. What does that change about your job specifically?
Ee Lyn Khoo:
Oh my gosh. If AI is infrastructure instead of a tool, if my job changes at the root, a tool is something people opt into. Infrastructure on the other hand, is something everybody depends on. And so I, in that case, I'm not planning for just a handful of AI roles anymore. I'm planning the whole workforce around it. That means that AI literacy isn't just a training module, it's. Something as basic as knowing how to send email. And so it changes who we hire for. We need people who are strong systems thinkers have sound judgment and high agility because with infrastructure, the ground will shift much faster and we can't afford to recreate the playbook every time it shifts.
Gemma Versace:
I think for all of the HR leaders listening, again, some really fantastic words of wisdom for them to be able to also potentially look at their. Talent strategy for bringing on AI talent as being having more of that breadth and depth of experience. And I think that's a really good way of looking at it, that not just a tool, but also something that, you know, becomes every single.
Day occurrence for people working within not only your business, but also for many businesses out there at the moment with a lot of leaders listening today. This question is for both of you. So Danny, you have warned against over reliance on AI and Ee Lyn your running weekly experimentation across every function. Very, I guess contentious question here but who's right?
Ee Lyn Khoo:
I think it's such a timely question, and I suspect every company is going through this right now. I think we're both right because we're solving different problems. Listen, I think experimentation is a good thing, period, but we are very disciplined about it. So we reimagine the process first, or we reimagining the operating model first, and then we push on.
We push on what's actually possible with AI. It's not about try everything and see what sticks. What do you think Danny?
Danny Fields:
I completely agree with you. And in engineering, I mean, new technologies are becoming available pretty much every week or every month, and we have to experiment all the time to make sure we're at the cutting edge. So our engineers are allowed to experiment, but at the same time, everything they do, we have to follow strict processes, stay within guardrails.
Make sure the technologies we're building. The agents only have access to certain data, like I mentioned earlier on, and making sure we keep humans in the loop and so on. So we want to go as fast as we can. We want to be cutting edge. We want our engineers to learn as quickly as possible about the new technologies.
We wanna find problems and fix them faster. But we just need to make sure that we're not repeating mistakes. So we do encourage experimentation, but also within strict guard rails.
Ee Lyn Khoo:
Yeah, and if I tag onto that, it's really experimentation versus caution. It's experimentation within the guardrails that Danny builds, and that's actually a key reason why it's safe to move as fast as we do at Avalara.
Gemma Versace:
Fantastic. Thanks for that insight guys. If you are comfortable sharing. Danny. When it comes to AI tooling, was this something that at Avalara, the leadership team comes together to identify. Approve particular AI tools that will be made available to a, across the engineering team initially.
And then more broadly, the business or is it a case? So is it more kind of mandated from the top down or have you used some of this experimentation where. People across the business can use their own tooling and then provide feedback that then potentially gets taken up across, across the broader business.
What's your, what has your kind of take been on that and across all functions really, not just specifically engineering. I.
Danny Fields:
You know, that's a fantastic question, Gemma. In engineering most of the tools we're using probably started from a grounds up approach the engineers will find the new tool, the new technology they will start using it. Obviously, they have to get approval to use it internally. And then when we find a tool that is something that we have to use in our products or to build our products or to make us more efficient we have a team of people that will review the tool, check the pricing, figure out how much this thing is going to cost because AI tools can be very expensive.
And then we'll roll it out for all of the teams that need it. So today, I'm very proud to say in Avalara, we're using probably the most cutting edge tools that are available, and we're willing to invest in purchasing new tools in order to build the best products as fast as we can, and the best way that they can be built.
So we're very careful and deliberate about checking that the tool is a fit for purpose in the organization, checking the cost of the tool. And we do have a centralized function for that again, but we do encourage engineers to experiment and go out and find new tools that they can bring into the team.
And then across the entire company, if other departments that are experts in their function, if they identify a third party tool that makes sense to bring into the company, that's AI powered, we have a team that will review. The tool the team sits in it and they will examine it, make sure it's secure, look at the expense, and we have a formal process to adopt that and bring it into the company.
But we do it very quickly because with AI, everything moves fast and there are processes to review and buy new tools has to be fast as well.
Gemma Versace:
Yes, absolutely. The speed at which AI is moving. I we had a podcast, a guest a couple of weeks ago when we asked him, how quickly do you need to be adapting and moving with AI, and his response was, you need to feel like you're on a bike riding down a hill.
That if, if you don't feel the breeze in your air, in your face because you're moving so fast and you're probably not going fast enough. So that's brilliant that you and your team are able to have that efficiency with responding to business needs as well. Last question and this one is for both of you as well.
Again, Danny, you run a no blame incident review, which we've touched on. Whilst Ee Lyn, you talk about resilience when things don't land, how do you keep either one from turning into no accountability?
Danny Fields:
Again, a great question and it goes back to what I said earlier, Gemma when we start the meeting and we call it an RCA/CAPA process, so that's RCA, root cause analysis. And the PA is preventative action. So with the root cause analysis part of the meeting, we say it upfront, we tell the engineers we're not going to blame anybody.
We tell them they're in a safe space. They're not going to get into trouble. There won't be any finger pointing. And again, we've been doing this for a number of years. And by saying it upfront and the setting the tone of the meeting the woman who runs the meeting is amazing at just setting everybody at ease at the beginning.
And telling them that this is a safe place and we do provide psychological safety in this meeting. And the truth always comes out about like what went wrong in the technology. And that's part of our job as engineers to build something and if something goes wrong identify the problem, the root cause, and then make the software better so it'll never happen again.
Ee Lyn Khoo:
And I think that's where resilience picks up. Gemma resilience isn't about just moving on, it's about coming back stronger. And the only way you can prove that is that it is if the team works differently the next time. I think Danny's process gives us the facts about what broke, but it's on all of us, especially the executive team. To make sure that those facts actually improve how we work and where we're headed is a world where managers will own the outcome of the human agent combination, IE people and the AI or whatever AI that they're working with. And so this is going to fundamentally change what management means, and it's creating a shared accountability across engineering it, and everybody who's driving outcomes for our partners and customers every day.
Gemma Versace:
That's thank you both for your answers there. And I think it also just reinforces what has been, I guess, a theme through this conversation as well is that, you know, I think when people think about AI and its adoption that it purely sits within, you know, from a CTO and the engineering side of things, but it is, it's a full function adoption that is required and, as I said, the synergy and collaboration between yourself as the Chief People Officer, Ee Lyn, and yourself, Danny, as the CTO. It's definitely providing a lot more value and meaning. It's a lot more meaningful and genuine going across all of the functions and across your entire business. So. Thank you so much for joining us today.
We do have a one final question for all of our Keep Moving Forward podcast guests is what keeps you moving forward? What gets you going in the morning and what makes you really, you know, excited to to get up in the morning guys.
Danny Fields:
Well this morning it was a strong cup of black coffee. But most
Gemma Versace:
Not a Guinness.
Danny Fields:
I wake up and I face our engineering team. Not, that's for the evenings. But in the mornings it's… I'm waking up every day trying to figure out how can we continuously disrupt ourselves, right?
And this new world of AI, AI is changing everything. And everybody in engineering, everybody across our company needs to be using AI to make the company more efficient and to drive growth in the business. And it's sort of like the book The Innovator's Dilemma.
We're looking constantly to find new disruptive technologies that will change things.
And I'm working with Ee Lyn and other leaders across the company to figure out how can we disrupt ourselves internally, not just in engineering, but in every department. How can we reimagine and transform every business process across the company so we can go far faster? So. It's exciting. I mean, we have all these new AI tools.
It's like a big box of toys and we get to play with them every day, and we get to figure out how we can make the company better and go faster. So that's what I wake up thinking about every day, Gemma. If we don't disrupt ourselves, somebody else will.
Ee Lyn Khoo:
Yeah, and I think in a similar vein, like what keeps me bounding into work is we're in such a moment right now and I'm closely watching whether our people and our business. Come out of this better than when they went in. And that's not gonna happen by accident. It's a choice that we make every day with intention.
And so this moment we're in won't wait for us. And so either we shape it or it's gonna happen to us, and I'd rather shape it. And that's exciting.
Gemma Versace:
Yeah, absolutely. What a brilliant answer. Thank you so much. And you can definitely see the passion and the energy but also the belief that you both have in not only the jobs that you do but clearly the, clearly the business that you're working for and the people you're working with.
So thank you so much for joining us today, guys.
Danny Fields:
Thanks, Gemma.
Ee Lyn Khoo:
Thank you.
Gemma Versace:
Pressure tells people to move faster. Design tells them where to go. Pressure is the easy default, because it's fast and it feels like leadership. Design takes longer, since it means deciding what good actually looks like before you ask anyone to chase it.
Ee Lyn defined success for every role first, so speed had somewhere to go, instead of pushing people to use AI more. Danny runs the same instinct through engineering. His team can experiment as fast as they want, but only inside guardrails that keep a human checking the final step. Neither one is choosing between speed and control. They're building systems where speed only happens because control is already in place.
Danny and Ee Lyn got here independently, solving the same problem from different directions, and agreeing on what AI infrastructure actually requires before either team built anything.
AI infrastructure has to be designed on purpose, well before the pressure hits.
Join us next time for more conversations with technology leaders who help us grow and lead as we keep moving forward. You can find us on Apple Podcasts, Spotify, or YouTube. If you enjoyed this episode, please share it with your network.
We'll see you next time.