Closing the AI Trust Gap with Andy MacMillan

Hey everyone, we're back here at Inspire 2026 Alteryx User Conference. I'm really happy to have this gentleman sitting next to me right here. It's Andy MacMillan.

Andy is the CEO of Alteryx, and so we're going to hear it right from his mouth what's going on here, what his impressions are, and I think it's going to be a great interview. Andy, welcome to Techstrong TV. It's great to have you on here.

Thanks for having me. I appreciate it. It's my pleasure.

So Andy, you took the reins here about 18 months ago. Well, let's start with a simple question. What were you doing before that?

So before I was at Alteryx, I was at a company called UserTesting, which was focused on getting customer feedback so people could build more engaging products. Had a great run. We took it public, took it private, did a big combination with the number two player in the space.

I had a great time doing it. But it had sort of run its course, and I was looking for something a little bit more AI-centric for my next role. Really?

18 months ago you realized it was AI? We were even earlier. 5 wall to wall, so almost two and a half years ago.

Wow. And what we found was it was really good at creating creative things from other creative things. Yeah.

So feed it 10 pieces of marketing content, it'll give me an 11th piece that sounds kind of like my marketing content. What it was really not equipped for or good at was dealing with my actual enterprise business processes and my data. And I actually left thinking I was going to go start a company that would solve that problem.

So I'd actually started to build some prototypes. I'm not a founder by nature, I'm more of a scale guy. But I felt so strongly there's a great need.

And while I was working on that project, I got the call about the Alteryx job, and I was like, "Oh, this is exactly it. " So that's how it happened. Sometimes things just work that way.

It did feel that way. It felt that way. It's interesting, I tell people this all the time.

So I'm a founder. I've founded, co-founded four or five venture-backed-- Well, this company wasn't. Techstrong was not venture-backed.

I bootstrapped it. It just worked like that. Amazing.

Sometimes it just strikes right. But the other ones I've done are both in venture-backed, so and I know what you mean. You're building it.

The thing about now, though, is you don't have to build an engineering team of 20 people anymore and go raise 5 million bucks to get a prototype out the door. With AI, it seems like back to two guys in a garage, right? To build a company, and that's one I think one of the magical things about all of this.

So you were looking to do something in AI, and Alteryx was already the perfect vehicle to do it. What made it perfect, you think? I think in a lot of ways, and I joke that I've been doing this long enough that stuff starts to look like a movie I've seen before.

Even movies feel that way now. My wife and I watch a movie and I'm like, "I've seen this movie before. " We wrote this one.

Yeah. And so it sort of rhymes with the past. And I think one of the things that I observed about the opportunity with Alteryx was getting into something that had a meaningful amount of scale that was solving an original problem, which is there are people in your business that know your business really well, and they know your data really well.

These are the kind of ops people that work in sales ops and rev ops and finance ops and accounting and things like that. And they're the kind of people when you ask them a question about the business, they ask you three questions back. Do you mean like this or like that?

And what about partners? And how does this fit in? And I think those are the folks that need to be empowered to bring the business logic, the data, and sort of the experience of how the business works into an automation world to solve a problem.

Historically, that was building something like a Tableau dashboard. Yeah. Right?

And so you go to Alteryx and you put all that logic in place to have the dashboard be truthful to your business process and your metrics and how everybody thinks about the data when you present that data. I think that new universe is all the way the same right up until you get to, now I want an agent to interact with that data or do something about my business, but I still need all that trust and that understanding about the business backing up that agent or that LLM. And again, I think that's something, while it's new that we're doing it for AI, it's not new to doing that job for Alteryx.

com. Doing that one since 2014. That concept of tribal knowledge, right?

That's tied up. So you have technical debt, but you also have tribal knowledge. If you ever read the book "The Phoenix Project" which is all about that.

One of the conflicts there is the person who has the tribal knowledge may not be the best person for the new organization you're building, but how do you not live with that tribal knowledge? " Now we have the ability to almost capture that tribal knowledge from folks. You're presenting actually, we're recording this on a Tuesday.

You're presenting on Wednesday here. Tomorrow morning. Yep, that's right.

The main keynote. This isn't live. People, by the time they see this, you would have already spoken through the magic of TV.

Let's hope it goes well. Well, it's going to. There's no doubt in my mind.

But give our readers and listeners, watchers, a preview of the keynote. Yeah, the keynote is really focused on how do we empower those folks that know the business so well to take that logic and implement it in a way where AI can make it accessible. I have this belief that when technology transformation happens, that often the early adopters are the technologists.

They're the people that love to tinker with the technology. The real value comes when that-... ability gets put in the hands of people that are working part of the business problem.

I think it's like the dot-com boom was the same way. When e-commerce and websites took off was when we empowered marketing and product managers to sell stuff online, not when we empowered technical people to put stuff on the internet. So first we're going to be focusing on our user base to help them get ready for empowering their businesses with AI, how to empower their agents with AI, and their knowledge around the business.

We're also going to unveil a lot of product. I came here a year ago as a new CEO at Alteryx and laid out a vision for our Alteryx 1 product, working with the big data platforms, the BigQueries, the Snowflakes, the Databricks of the world, and doing all of that to empower business users to build logic that powers AI. Mm-hmm.

We're back a year later, and again, because of your point on building with AI, you can build so much faster. We've got just a ton of innovation we're going to show our customers, and as a product person, my background, I'm really excited about where we've taken the product. Absolutely.

And while we're here, let's just make sure we hit it. Where can people go? com.

I actually have written an article about these product announcements, and you can read about it there. But where can they get their hands dirty with it? Will it be available on a website?

com. We have a great community website as well. We just relaunched with this amazing community of 750,000 members of people that just love Alteryx and build stuff with it.

And so there's a whole bunch on that community. Anybody can join the community. We do a lot on LinkedIn as well.

So tons of places to learn about it. Cool. Andy, I want to return to the theme you were hitting on, which is these holders of the tribal knowledge.

They're really called gurus. Call them whatever you want, but they're the holders of the tribal knowledge. One of the things I see is, quite frankly, they're a little, I'm trying to think of a good way of saying this without being negative.

They're wary of AI. How does that sound? Yeah.

Okay. They're wary of AI, right? They don't trust it, if I have to be blunt, right?

They just don't trust it. If I put it out there, who's going to have access to it? Is it going to delete things?

You hear horror stories. For sure. What do you say to those people?

How do you gain that trust? It's understandable. Well, I'm not saying it's irrational.

Well, technology disruption, any disruption, people go through this. Is it good for me? What does it mean for me?

What does it mean for my business? So I think all that's fair. All of the data that we've seen, we surveyed our users, and they came back and talked about this exact issue.

Their organizations are trying to figure out how to trust what comes out of AI. It is this new technology that does this amazing stuff that looks like magic. Magic's really fun to demo.

Magic's hard to trust. Yeah. Right?

And this is why so many companies have these amazing pilots. They showed somebody some cool magic, and then they went, "Great, are we going to close the books on that? " Yeah.

" And go, "I don't know. '" The AI said it was, yeah. So there's a trust gap.

Right. A real meaningful and legitimate trust gap. What we're talking about during the keynote tomorrow is what's a framework to think differently about solving that gap.

And the framework that we're laying out for our customer base is can the answers you get from AI be visible- Mm-hmm ... understandable- Yep ... repeatable- Yes ...

and auditable? If I ask it the question a bunch of times, do I get the same answer? Can I go visibly see how it came up with that?

And this is a big part of the power of our product is it's used by people who are data technical, right? The kind of people that can build spreadsheets and pivot tables, typically not coders. Yeah.

And so if I go to my CFO and I say, "Hey, I vibe coded 8,000 lines of Python, it's going to do our Sarbanes-Oxley compliance," probably not going to happen. But if I go to my CFO and I say, "Hey, we use Alteryx. We've been using Alteryx for years.

I had maybe Claude help me build three new workflows. Here they are. They're going to run just like this.

They're Alteryx workflows. And I can have my auditor look at it. I can have my legal team look at it.

And we can put this in production. If we have to make a small change, we can go into that Alteryx workflow, and maybe AI helps us make the change. But after we make that change, it's visible, understandable, repeatable- Trustable ...

auditable. Trustable. So that's what it all comes down to.

And I think when that happens, and we get to a place where people can trust what's going to happen with AI, the magic becomes a superpower because it's no longer a magic trick, it's a capability. And I think that's the shift we have to go through. I agree with you.

Now, getting from here to there is the issue, right? We're at this, call it awkward adolescent stage, where we have, even at TechShop, we're a small company. We have all these pilots, we have all these different projects that we're running agentics in and everything.

But moving from pilot to production, and I'm talking now from my personal experience, the AI gets me 85, 90% sometimes there. You still got to tweak and push and shove and- Yeah ... pull that 10% out of there.

And you got to know what that 10% is. First, you have to identify this is the- Back to, again, if I've got a Claude-coded Python script closing my books, and I don't know which 10% it got wrong- And I don't know Python- ... it's not very helpful ...

from rattlesnake. That's right. It's just rattlesnakes to me.

That's the difference. And so I agree. I think it's incredibly powerful.

How do we give you that 90% and not take the 10% away from you, let you do that? Yeah. And here's the good news.

I think that's just a time thing, right? That's a get to know you thing, and we're going to rapidly move through that as well. I think one of the other things, and I've spoken to a few people here already about it, is generative AI, 2025 was the year of generative AI.

Well, 2024 was, too. 2023 probably was. Frankly, sort of said, yeah.

I agree. But certainly this year, agentics are real. It's all about the agentsAnd that, again, plays right into Alteryx, right into our strategy.

But how do we get our companies eager to say, "Hey," it was a nice parlor trick, as you mentioned, right? Generative AI did things that were auto-magical, right? Yeah.

Not quite sure how it did it, but I loved what it did. Yeah. It was so cool.

Right. But now stuff's getting real, and agentics are where it gets real. Are people eager to embrace those agentics, having had enough of the parlor tricks?

Or are they saying, "Well, I learned my lesson there. " I think they're anxious. And what I think they're anxious about is, AI's not going away.

No. " Don't like business happening. This is happening.

Yet, we haven't seen a lot of stuff go into production as an industry. So how do we fix that? And another trend I've seen along with the agentics piece is the line of business budget is now becoming where AI's happening.

When I meet with CFOs or CLOs, what they're telling me is that last year the CIO had a budget for AI. This year, I have a budget for AI. What I don't have a budget for necessarily is to hire a bunch of people.

So I have to take this budget that I got for AI, and I have to solve real business problems. Without headcount. Right.

In a way- Operational ... in a way that I trust. So how can I take the infrastructure that the IT team has provided me?

Maybe they've given me access to BigQuery, maybe they've given me Gemini, but I don't just need Gemini and BigQuery. I have to bring in this business logic, and I got to, whatever, close the books, cut territories, figure out how to do forecasting, whatever those things are. And, again, my whole thesis is AI takes off when those folks that have to solve those problems can say, "Okay, I got it.

" That's where agents take off. So yeah, I call that operationalizing AI, right? And that's one of the interesting things.

I was talking to your CMO about this this morning. Look, coding, pure software coding, marketing, they're kinds of the tips of the spear. Those functions, AI, quick.

Right. I mean, we have no doubt about it. It's writing code.

It's making great mark. As you mentioned, I give it 10 marketing docs and the 11th is a killer. Yeah.

Right? That's right. But that's different than some of the other jobs we're talking about, right?

And I don't know if we've quite cracked the formula on doing what the CFO's team is working on. That's right. Right?

Forecasting, budgeting, these kinds of things. The nuts and bolts of ops. Yeah.

The things where it's cool that we have this new probabilistic capability, but probably doesn't rhyme very well with my finance. No, it's not determining risk. That's right.

It's probabilistic. But there's some great use cases that could support using those agents. And I think it's just a matter of time.

And maybe I'm a glass half full kind of guy. It's just a matter of time because every week it's relentless. Yep.

It just gets better every week. I wanted to get to the next question, though, and this is a problem I've heard from a lot of CEOs, which is once we get to that point where we're really using AI, Katy bar the door because the scale, now we don't move at human scale. We move at machine scale.

Yep. At AI scale. And that creates yet another series of governance and trust issues, which is, it's okay when I'm turning out four things a day.

When I'm turning out 400, I just don't have enough eyeballs to make sure. It's like if I don't trust it, what the hell am I going to do? But I agree.

But you're back to trust. Yeah. And then you go, "Okay, I'm going to turn out 400 things based on what?

" Right? If I have employees building those 400 things, how do I trust it? Who are those people?

Yeah. What are they trained to do? And so I think getting back to how do I implement a set of capabilities that the business owns, that I trust if I'm that line of business owner.

I'm the head of rev ops, I need to cut territories. I have an agent that cuts territories. How do I trust it?

I think I trust it if my territory planning team owns the logic of how this thing gets done, and can maintain it, and can update it. Because that's the only thing we never seem to want to talk about with all these AI pilots is stuff changes. It's not a one and done.

Yeah. I use the example of imagine having a sales commission agent. Mm-hmm.

Right? And I joke that if you asked ChatGPT a year ago what will my sales commission be, it'll give you a three-page write-up on how sales commissions work, but it won't tell you what your sales commission will be. Right.

And what I want to know is what will my sales commission be? Sales commission. And that's a specific answer.

There's a right and a wrong answer. And again, I'm not asking for, hey, make a reasonable calculation. I'm asking make the exact calculation that my commissions team uses to generate commissions.

Yeah. The person that should own that calculation should be the commissions team, not somebody who vibe coded a commissions app because- What do they know? " Now they got to change.

You got to change the commission agent. I agree. And so that infrastructure is what lets me build 400 things and trust it is I know the commissions are right, and I know that the revenue calculation's right, and I know that the territory planning is right, and I know that the sales controls are right because those teams are empowered to be trusted.

Sometimes I feel, though, that we hold AI up to an artificial level of perfection, right? I think with people, we operate in the don't let perfect get in the way of good. We got to keep the- Right ...

the lights on. We got to keep the balls moving. With AI, it's almost like, oh, it made a mistake once.

I can't trust it. That's true. Yeah.

So again, we're going to have to do that. We don't have a lot of time. I want to pivot, Andy, real quick and really focus in on Alteryx now.

Talk about the evolution to date and then what you seeBetween now and Inspire 2027. Yeah. The big evolution has been this Alteryx One product.

So Alteryx, two years ago, had essentially two products. We had a desktop product that had a super high net promoter score, massive community around it, everybody knew it. And then we built a whole set of cloud capabilities to work with the Databricks and Snowflakes and BigQueries of the world.

We built it sort of under the assumption that that would be a new user. Mm-hmm. And what we discovered in the market was, no, that's the same user.

Same user. " Part of why I think our audience is going to be so excited because we're bringing all this new tech that we've built over the last couple of years that they weren't at their fingertips. We're bringing it right into the environment they love.

And then the second bit is showing them how to use that tech to be that business logic owner for AI. I've been doing this myself. I love tinkering with the product.

Even just this last Friday, I built a workflow that analyzes my sales pipeline performance. I built it in Alteryx using our Ask Alteryx AI copilot. Love it.

It took me about an hour to build the workflow. It all runs off our Snowflake infrastructure. It's verified the data's correct by my finance team.

It runs every morning now, and it creates a report in Gemini that I can work with, and did it all myself. And again, I think that's where when that kind of capability is put in the hands of the users trying to solve problems, that's where this thing takes off. So I got to tell you, as a CEO, I'm in the same boat.

I get so damn excited that I'm able to do this without asking the whole team and making everyone run crazy to get the CEO a report, right? It's a good feeling. Yeah.

It's a good feeling, man. Hey, one last question and we're going to run. What do you say to other executives out here who are-- We're all dealing with the same things right now.

We worry about headcount. We worry about trust issues. We worry about, are we betting the company on maybe something unproven, but if we don't make that bet, we're going to lose anyway, perhaps.

What would you say to your fellow CEOs out there? It's transformational. I don't think this is a moment to lean out.

It's a moment to lean in. I think understanding who's going to own the business logic and the automations that give you the predictability and the visibility, the understandability, the repeatability, the auditability that lets you trust AI, that's the unlock. You can't enable with AI what you can't automate, right?

If you have a way to recognize revenue, you need that orchestrated in a way that you can trust. But if you can do that and you can empower your teams to build that logic, I think you can trust AI, and I think a company that trusts AI is going to scale a lot faster than those that don't. Agreed.

Andy, thank you so much for doing this. Really enjoyed it. We appreciate it.

Check out the keynote tomorrow. Well, it'll be available, I think June 5th- That's right ... they said on virtual on demand.

So you can check it out there. You can read our coverage all around this. But for now, we'll be back with more on Techstrong TV.

Just stay tuned.