Boomi World 2026 | Matt McLarty, Boomi

Clips
More from Boomi World 2026

Matt McLarty

CTO

Boomi

play_circle_outline Originated from Chris Cappetta's work to create a "digital expert builder" using LLMs.

play_circle_outline Explosive Adoption: Early Partners Praise High Satisfaction as Product Integrates Claude Code, Copilot and Codex

play_circle_outline Companion relies on Boomi platform APIs and composability for connectivity and automation.

play_circle_outline Marketplace automation: Companion will build marketplace recipes and document how solutions were created.

Info
Transcript

Matt McLarty, Boomi

Matt McLarty

CTO Boomi

In this interview from Boomi World 2026 in Chicago, Matt McLarty, chief technology officer of Boomi, joins theCUBE's John Furrier and theCUBE + NYSE Wired's Gemma Allen to discuss the launch of Boomi Companion and what it signals for the future of enterprise agentic engineering. McLarty introduces Companion as a collection of agent skills built on Anthropic's open Claude skills standard that goes far beyond conventional code generation — fully building, deploying, testing and iterating integration solutions from natural language instructions. He explains that... Read more

explore Keep Exploring

Who worked on creating a digital "me," and what is Companion? add

Who are the early adopters of this offering, and who is it currently serving? add

How do you categorize the different types of agentic AI (e.g., agentic analytics, agentic automation, agentic engineering), and where does your product Companion fit into that landscape—what does it use under the hood and how does it enable business automation through APIs? add

How is the team's strategy evolving for building marketplace assets and using Companion, and how is the team's role changing as a result? add

bolt Powered by CUBE AI

Matt McLarty, Boomi

search

Gemma Allen

>> Welcome back to Boomi World 2026 coming to you live from Chicago. I'm Gemma Allen with theCUBE and I'm joined by my cohost, John Furrier, and we are talking all things the future of the Boomi ecosystem, some of the great technology developments of the last year and the year to come. And joining us now is one of the main brainiacs behind some of these developments, Matt McLarty, CTO of Boomi. Welcome, Matt.

Matt McLarty

>> Great to be here. Great to be back.

Gemma Allen

>> So I think we're just going to get straight into it, and talk about one of the exciting things to hit the stage tomorrow, but you mentioned off camera, it's already kind of out in the ether, and that is Boomi Companion. Maybe break that down for us. What exactly is that?

Matt McLarty

>> Yes. What is Boomi Companion? I like origin stories, if it's okay.

Gemma Allen

>> Of course.

John Furrier

>> Absolutely.

Matt McLarty

>> So we've all been tracking this ridiculously explosive innovation in the AI space. Since its inception, we've been kind of looking at how do you use LLM technology to do more with our platform, as well as where do we fit in the ecosystem? So one of the areas is just like, we have so many customers that are building stuff, building lots of stuff. As AI comes in, demands get higher. How do we leverage the new technology to just make it easier and faster to build solutions on our platform? So a gentleman named Chris Cappetta on my team who's here, he's been working for the last several months on doing exactly that. He's an expert builder himself. He's like, "How do I turn this technology into a clone of me, digital me?" And he started this work early, before there was a Claude Code, before there was Claude skills and all this, and he's just been iterating it ever since. So what Companion is is now a collection of agent skills, which is something that Anthropic made an open standard of, which is ways that you can tell agents, instruct them how to perform repeatable common tasks and repeatable roles, that is just really good at building Boomi solutions. And because it's created by an expert Boomi Builder, he's got all sorts of best practices in there, and we've expanded it so other members of the team have been contributing to it. So as opposed to like a typical old-fashioned code generation thing that might come up with like a little skeleton thing, "Hey, here's what you might do," and then you have to go fill in the details, Companion actually builds a fully robust solution. It deploys it, it tests it, it checks its results, it iterates through, it learns, it talks to the user, and so it's one of those things that it's almost a little too magical in terms of what it can do. So now as we brought this out to the Boomi community, which we've been sort of slow rolling it. You know how a restaurant might open up for a couple months, and then they have a grand opening?

John Furrier

>> That's kind of -

Matt McLarty

>> Yeah. Kind of what we're doing with Boomi Companion. We've been actually sharing it with partners, sharing it with customers, and it's had this explosive adoption. So it's like you just give some natural language instructions, you give it maybe some links to documentation, some structured files, whatever you want to give it. And it's incredible at just building these robust solutions.

Gemma Allen

>> Well, help us profile some of these early adopters. Who exactly is this serving right now?

Matt McLarty

>> Well, we were starting with some of our closest partners, because I think a lot of people were sort of experimenting with how to do this, and we have actually some incredible work that's been done in our partner community and customer community. But we're like, connect with them and say, "Hey, here's what we're building." And in some cases we're able to sort of ... Because it's skills, it's built on this open standard, people can sort of bring their own skills, and that is really our goal, is to make this very collaborative, and take in all the intelligence of the community, add new skills, augment the skills we have, but the reaction has just been like, "Where have you been all my life?" So it's especially appealing to long-term customers, users of Boomi, because they can really see the value. They can see not only is it building this stuff, but it's like building it in a way that's going to really be useful and scalable and follow best practices. But they're sort of our test subjects right now. But I think even those who are ... The way it works is you take these skills, and then you can implement them into your coding platform of choice. So it could be Claude Code, could be Codex, could be Copilot, because it's this open standard. The idea there is we want to go to where developers live. We have incredible capabilities in the platform that you can use, but what happens when there's developers out there that are building stuff, maybe they don't even know about Boomi, right? But if you have these skills available in that enterprise Claude Code deployment, whatever, as people are coming up with, "I need to build something like this," it can come back and say, "Well, yeah, you build that with Boomi." So we haven't really tapped that potential yet, but I think that's where we go with that.

John Furrier

>> There's so much headroom. I mean, look at this digital twin concept, with the builder. Amazon built an agent for DevOps, because they know what the grind is for, the undifferentiated heavy lifting, their term they use.

Matt McLarty

>> Exactly.

John Furrier

>> And so they basically know what a DevOps engineer does, so they just built it, because they do it.

Matt McLarty

>> Yep, exactly.

John Furrier

>> They are the subject matter experts. How does this scale? Or two questions. At what level has it gotten from a data activation standpoint, what's available, and how much room does it have to grow? In other words, is it remedial tasks? Is it full data access? What level of data and privileges?

Matt McLarty

>> Yeah, so-

John Furrier

>> I can only imagine there's a lot of compliance and governance involved.

Matt McLarty

>> I see kind of two buckets right now that don't necessarily have to be distinct, but I think right now they are distinct. When it comes to agentic AI, you've got agentic automation. I mean, there's actually three. So there's what I would say is agentic analytics, but that's kind of its own thing that's less where we play, where you're getting into sort of agents doing analysis on pure data. I mean, we can do some stuff there, but the big payoff space I think is agentic automation, which is when you can get agents actually running your core business operations for the company. But then there's agentic engineering. So this is how you run the business, this is how you build the business. So Companion right now, we're more in the how you build the business space. So in terms of what we're actually using under the hood of Companion, the beauty of it is it's built all on the APIs of the Boomi platform. So we've been helping companies in the automation space for a while. I won't say agentic automation, but helping companies in the automation space for a while run their core business in a more efficient way and a more connected way. And so that's that particular domain, and we've been helping them create APIs and make things more connected, make things more composable. But ourselves are like, "We need to be composable too. We need to be more automatable as well." So we've been, over the last few years, opening up the platform more through API. So this is really a realization of the potential we have with the platform to go and build. So the connectivity layer for Companion is more around our own APIs. Now what we build with Companion are those processes and solutions that are going to then hit the data connections and automate the business. We're building skills now to bring in more agentic processing, so we can build the solutions through Companion that then run on the platform, and do the automation and ultimately the agentic automation. I think those worlds can come together though, because if you think about like long term, if you have agents that are running processes in production, and we talked about what we're introducing with Orchestrate that's more monitoring. And you can only have so much time in Steve's keynote, you got to get the Fruity Pebbles in, so you can't give all the details away. But you look at what's in Orchestrate, a lot of it is understanding the ... Really, being the bridge between the business and the technology running the business. So if we can actually provide a layer of abstraction that is like, "This is the process that does your order to cash," or, "This is the process that does your employee onboarding," if we actually have the way of mapping that business process to the underlying technology, that's very powerful. And if you think about how that could be used in an agentic context, if we can detect that there's slowdowns or suboptimal processing going on, then you can update or rebuild the processes running it, which is where I think those streams cross. You can get Companion to come in and build that on the fly. We're not there yet, so there's still distinct worlds, but I think they do come together. But for now, Companion is very much in the build realm, and it's really instantiating the stuff that would make those other connections.

John Furrier

>> As you said, the cat's out of the bag, so you're going to talk about it on the stage tomorrow. What's the key learnings, besides the fact that people love it? What's the takeaway? Is there going to be an update tomorrow?

Matt McLarty

>> I think that since ChatGPT exploded in December 2022 or whatever it was, everyone's been seeing that it's ... the inevitability of like the AI revolution. But in practice, a lot of enterprises are like, "It's hype, but I'm not seeing the value yet." It's pretty clear that the industry, not surprisingly, you've got a bunch of software engineers trying to build solutions and stuff. The fact that they've built incredible coding capability in the LLM space, like the agentic engineering space I think is really more refined than anywhere else yet. We'll get there with agentic automation. I think we're there-

John Furrier

>> I like that term, by the way, agentic engineering. It's not DevOps per se, but it builds on DevOps.

Matt McLarty

>> Right. Well, and we specifically differentiate between agentic engineering and vibe coding.

John Furrier

>> Of course.

Matt McLarty

>> Because vibe coding is kind of like hobbyist-

John Furrier

>> It's a -...

Matt McLarty

>> shot in the dark. Yeah, exactly....

John Furrier

>> it's like going out for dinner.

Matt McLarty

>> Agentic engineering is like, how do we get the best practices of ... And that's a whole other topic, because the whole software development life cycle has been built up on these deterministic solutions, and that's out the window now. Anyway, but I think because the technology's come so far, like huge leaps around agentic engineering, it's like, here's a way that you can really ... That's what we're seeing with our customers and partners is they're like, "This is the payoff. We were excited, we've been excited for three years, but we've also been wondering when's the payoff?" This is the payoff.

John Furrier

>> Gemma and I have been talking about this on theCUBE at the NYSE. The enterprise adoption has lagged, on the hype's been high, but it's been lag ... Yeah, marketing gets better, better marketing copy, good RAG. That's not going to move the needle.

Matt McLarty

>> I would argue on the better marketing copy, by the way. Because my LinkedIn inbox is just a ... Anyway-

John Furrier

>> The marketing slop, I would say, some call it, but that didn't really move the needle. But this year, the coding moves the needle, because enterprise has engineers and developers. They see the product too. They see the top line results from new apps that can go faster. You guys are-

Matt McLarty

>> Well, I think that where ... I'm biased, let's be honest, I'm biased. I think where this takes it to a new level, though, is it's not just barfing out a bunch of bespoke code and saying like, "Here's a big blob, go implement this." It's actually built on best practices of like a tried and true enterprise platform. It's structured in a way so that ... I think in the low code space, like one of the ... You could say Boomi's been low code. I would argue Boomi's had a principle around usability and great user experience. So for a generation, low code has been the, "Hey, you can drag and drop, it's easy to configure, you don't have to write code." But now with conversational interfaces, it's like a leapfrog thing. So we're not beholden to that, but I think that the graphical interface takes on a new role, which is, if I do build all this stuff, what did I just build? Well, we have a way of showing how it's built-

John Furrier

>> Get the kill switch, too-...

Matt McLarty

>> it's built, structuring it. Yeah.

John Furrier

>> You can kill switch.

Matt McLarty

>> And you can still tweak it that way if you want. It's like a bidirectional, best of both worlds.

John Furrier

>> Yeah, it's awesome. I mean, it's going to usher in the results that's going to bring in the agents. My question for you is, big discussion around the future of software engineering. I use the word engineering deliberately. It's not software development. It's more engineering now than dev. You went from, when I graduated in the '80s, software engineering, then you became a developer, then you were a coder. So we're back to engineering, but what is the future for an enterprise engineering organization? You mentioned you guys have firsthand experience. What advice would you give folks that are sitting out there trying to figure out their makeup of their team? How do you measure productivity? How do I marshal the agents? How do I watch the budget? I've been, heard a rumor that Amazon's been, they've been writing OpenClaw to juice their token budget, because they're managing employees on token usage.

Matt McLarty

>> Well, we know anytime you're focused on output metrics instead of outcome metrics, then that's a potential trouble area. And then a lot of companies are doing that -

John Furrier

>> What's your vision on engineering?

Matt McLarty

>> I think that the biggest challenge for enterprises in the software engineering space is, I think the biggest fallacy is you're going to throw everything out and replace it all with AI. That's just not efficient. Deterministic processing is solid for a reason, and deterministic processing is the foundation of the whole software development lifecycle. Like regression testing, everything is built on predictability, right? When you introduce LLM and probabilistic technologies, you're challenging the founding assumptions of software development. So I think it's really getting the balance right between deterministic processing and probabilistic processing is the biggest architectural challenge that's going to hit the software architecture, software engineering communities. I feel good, because I feel like we are a balancing point for doing that. I actually, I wrote a piece for IT Revolution blog a few months ago. If you're familiar with the book, Thinking Fast and Slow, Daniel Kahneman-

Gemma Allen

>> My husband loves that book.

Matt McLarty

>> Yeah. So system one, system two. System one in the brain, this is how the brain works. System one is the autopilot functions where you do everything on intuition, you don't ... stimulus response. System two is the thinking part of the brain. And the way the brain works is it tries to get system one to do everything. It's like, "Don't bother me. I'm lazy. System two's lazy, don't bother me. Get system one to do it." And that works, because it's optimized, right? I think we have to think about, that should be like a guiding principle as we design these agentic systems is to say, "If you can do it the deterministic way, do it the deterministic way." It's going to be more efficient operationally. It's going to fit more with your practices. You're going to need to evolve those practices to introduce probabilistic and see ... And there's already work being done. We actually are showcasing Marcus Mueller from my team, Steven Fishman, who's co-author of the book I wrote, he's in the COE team. They're showcasing an agent scorecard tool, which is using some new practices around testing the reliability and trustability of agents. So there's new thoughts out there, but the established practices are on the deterministic -

Gemma Allen

>> Yeah, go deterministic if you can, but convert the non-deterministic to deterministic.

Matt McLarty

>> Right. Augment with -

John Furrier

>> How should organizations manage this? Because this is the number one question we see people asking us is like, "Okay, how should I think about framing that, organizing it? What's the best practice?"

Matt McLarty

>> Well, so this is another thing I'll be talking about very briefly tomorrow, because we've actually created a website called agentic-academy.ai, so agentic-academy.ai, to put out some of those best practices around, specifically around, not to say, "In the purely agent space, here's the things that you do," there's lots of communities doing that. But really, how do you balance the world of agentic AI with the existing world of deterministic software processing? And patterns, primitives, I think we have to kind of go back to kindergarten on software architecture. This is a new paradigm of software architecture. We went through object-oriented, we went through service-oriented, we went through API first, microservices, like we have all these different movements. This is a bit of a bigger leap, but it's the, what is the prevailing software architecture practices that introduce probabilistic processing into real-time workloads? I don't know what we're calling it yet. I don't have a name for it yet, but that's-

John Furrier

>> Agentic engineering.

Matt McLarty

>> Yeah. Agentic engineering, could be, or agentic architecture. I'm not sure, but it's not going to be agent-oriented architecture. That's all I know. We're not going with AOA.

Gemma Allen

>> So talk a little bit about the longer tail opportunities and the new TAM that Companion brings to Boomi though, right? You mentioned that a lot of your long-term customers see a lot of value. They understand the construct and the concept very well, but there's also a whole other world out there of mid-cap companies, SMB that also need this opportunity, right?

Matt McLarty

>> Oh yeah.

Gemma Allen

>> They don't have the tech budgets that others do. So how are you guys thinking about that, and how are you messaging and landing that with them?

Matt McLarty

>> Well, this is, again, I mentioned the book that we wrote. In the subtitle of the book we wrote, it's called Unbundling the Enterprise. In the subtitle is optionality. It's called APIs, Optionality ... So we're offering options. Companion is very much aimed at the developers who might be using these tools. I mean, it does work with Cowork and other productivity tools. But we have Orchestrate in the platform that we just announced, which is very much aimed at any persona, and can go at the ideation level, does a lot of the same capabilities for building out that way. What we're finding is, if you provide the right set of options, then you can be a fit for all these different company sizes. So a big enterprise might be more utilizing companion with the full platform, and MCP registry, and so on. Smaller company might be like, "Hey, all the AI stuff we need is in Boomi and we can use Orchestrate to create these workloads." So we're just creating the options there. On Companion specifically, though, I wouldn't rule it out for ... Because it wouldn't surprise me if those companies as well are going to be experimenting with Cloud Cowork and other productivity tools, and ChatGPT, they're putting that type of capability in there. But what we're finding is even as we put it out there with this purpose of, "Hey, it's easier to build with Boomi," it's also, because it understands how to build with Boomi, it understands how to manage Boomi. It becomes sort of like this whole ability to talk to Boomi, anything you would do with the platform. So as we have more capabilities in there with MetaHub, which just came out, which has more of the semantic layer for understanding the business that helps do that mapping between business processes and the technology foundation, that can be something that helps provide more understanding as we get more Orchestrate capabilities in there. So it really becomes a way of really, a window into the entire fabric of your organization.

John Furrier

>> Just as a point of clarification, it's agentic-academy.ai

Matt McLarty

>> Yeah.

John Furrier

>> Just checked out the site, looks really good.

Matt McLarty

>> Cool. Well, yeah, power of AI, my friend. What might have been a multi-month project can be a weekend hobbyist-

John Furrier

>> Hire a design agency, six month waterfall, projects, meetings. I mean, the world's changed.

Matt McLarty

>> There are experts in those fields that I don't want to diminish, but for doing a prototype like that, running experiments? Unbelievable.

Gemma Allen

>> So Matt, aside from trying to ensure that you keep the messaging somewhat new tomorrow, and this technology has already gotten so much excitement, talk to us a little bit, final question, what's ahead for you? For the next couple of days, but also for the next year for you and the team?

Matt McLarty

>> Yeah. Well, I mean, I love this event, because it's just such an explosion of opportunity to meet with partners, customers, and just really take the pulse of our community. So it's looking forward to this next couple of days. But ahead, I mean, our team's in an interesting space. We've historically been building all the assets on our marketplace. If you go to marketplace.boomi.com, solutions and recipes and templates. This was our way of accelerating development, and we still do that, but Companion is, just unlocks everything, because now we're building those with Companion. And we're going to be providing like, for every recipe we put out there, we'll say, "Here's how we built it using Companion." And now we have a whole backlog of new skills that we want to incorporate into Companion. And we've got a lineup of partners and customers saying, "Hey, I built a skill for this. I built a skill for that." So our team has now become, into the whole product management realm-

John Furrier

>> You have a fleet of software agents on the Boomi team. Head count's going to go up.

Matt McLarty

>> Yeah, yeah.

John Furrier

>> Infinite head counts.

Matt McLarty

>> Infinite head count.

Gemma Allen

>> Matt, thanks so much for joining us here in theCUBE.

Matt McLarty

>> My pleasure.

John Furrier

>> Great to see you.

Matt McLarty

>> Thanks for having me.

Gemma Allen

>> I'm Gemma Allen. We are at Boomi World 2026, live in Chicago. I'm joined by John Furrier, where we are talking all things Boomi, the ecosystem, and the future of agentic tech. Stay tuned.