McKinsey 2026 Report: AI Agents Disrupt the Enterprise 'Build-vs-Buy' Model

Host: Hey everyone, welcome back to Allur. I’m your host, Alex Chan. Host: To help me break this down, I am thrilled to be joined by Marcus Thorne. Marcus is a veteran Enterprise Architect who’s spent the last fifteen years navigating the build-vs-buy dilemma at some of the biggest firms in the valley, and he’s currently spearheading AI integration strategies for several mid-market firms. Marcus, thanks for joining us on Allur! Guest: Hey Alex, thanks for having me. It’s a wild time to be talking about this. That McKinsey report… it’s basically putting into numbers what a lot of us have been feeling on the ground for the last six months. Host: Right? I mean, 32% is a massive number. That’s not just a few "bleeding-edge" tech companies; that’s the enterprise mainstream. Marcus, in your experience, why is this happening *now*? We’ve had generative AI for a couple of years. What changed? Guest: It’s the shift from "AI as a helper" to "AI as an agent." For a while, we were just using LLMs to help us write a function or debug a regex. But what the report highlights is the maturation of *autonomous* agents. We’re talking about systems where you don’t just ask for a snippet of code; you give it a high-level requirement—like, "Build me an internal dashboard that pulls from our CRM and our ERP, tracks churn, and gives the sales team an alert system"—and the agent actually goes and does it. It designs the architecture, writes the Go or PHP backend, builds the frontend, and handles the deployment pipeline. Host: It’s that velocity that’s the killer, isn't it? Guest: Exactly. It’s "high-velocity generative AI." I actually saw a demo recently where a dev lead used a command—it looked almost like a single line of code—something like `AI_Agent.create_dashboard` with a list of data sources and roles. And the agent just… spun it up. What used to take a team of three developers three months is now taking an afternoon of prompting and oversight. When the "build" time drops that drastically, the "buy" option starts to look really unattractive, especially when you consider the licensing costs and the fact that off-the-shelf software never *quite* fits your workflow perfectly. Host: I think that’s an "aha" moment for a lot of people. Because usually, when you buy SaaS, you’re paying for the convenience of not having to maintain code. But now, if the AI is the one maintaining the code, the "burden" of building custom software effectively vanishes. Guest: Precisely! And that’s the "seismic shift." Historically, the cost of custom software wasn't just the initial build; it was the "technical debt" and the maintenance. But if you have an autonomous agent that understands the codebase perfectly because it *wrote* it, and it can refactor it or add features in seconds? Suddenly, "Build" is the path of least resistance. Host: So, okay, let’s talk about the struggle here. If I’m a developer listening to this, I’m thinking, "Is my job just… gone?" Or "How do I even manage a fleet of AI-generated internal tools?" Marcus, what are you seeing in terms of the real-world struggles of implementing this? Guest: Oh, it’s not all sunshine and rainbows. The struggle is moving from being a "coder" to being an "architect" or an "orchestrator." I’ve seen teams try to let these agents loose, and if you don’t have a very clear set of architectural guardrails, you end up with a "black box" of code that nobody understands. Host: That's interesting! It’s like we’re moving up the stack. Instead of worrying about a semicolon in a Go routine, we’re worrying about: "Does this dashboard actually comply with GDPR?" Guest: Exactly. And honestly, it’s a bit of a learning curve. There’s a lot of "um, actually, that’s not what I meant" moments with the AI. You have to learn how to speak "Requirement-ese." Host: So, if 32% of companies are building their own tools, what happens to the SaaS industry? Are we looking at a "SaaS Apocalypse"? Guest: (Laughs) Maybe not an apocalypse, but definitely a reckoning. The "commodity" SaaS companies—the ones that provide simple CRUD apps, basic task managers, or standard reporting tools—they are in trouble. Why pay $20,000 a year for a tool that your internal AI agent can build in a week for the cost of some compute tokens? Host: It’s funny, I was talking to a friend who’s a Laravel dev, and he was saying he’s never been busier. He’s using AI agents to build these custom internal "micro-SaaS" tools for his company. He says he feels like he has a "superpower" now. It’s like he’s a one-man agency. Guest: That is exactly the sentiment. It’s empowering for the individual dev who knows how to leverage the tool. It’s no longer about how many hours you can sit and code; it’s about how many solutions you can deploy. And McKinsey is basically telling the C-suite, "Hey, your devs are now 10x more productive—stop buying expensive software and start letting them build." Host: Wow. It’s a complete reversal of the last 20 years of tech strategy. Marcus, what’s one piece of advice you’d give to a developer or a tech lead who’s looking at this report and wondering how to stay ahead? Guest: I’d say: don’t fear the agent, become its pilot. Start experimenting with these autonomous workflows now. Learn how to define system architectures at a high level. If you’re used to just taking a JIRA ticket and writing a function, start thinking about the *whole system*. Because the person who can describe the "why" and the "how" to the AI is going to be the most valuable person in the room. Host: "Become the pilot." I love that. Host: Marcus, this has been such an eye-opening conversation. It’s clear that the "Build vs. Buy" debate is no longer a simple math problem—it’s been completely disrupted by the speed of AI. Guest: You can find me on LinkedIn or at my blog, ThorneTechStrategy.com. It’s been a blast, Alex! Host: Thanks so much, Marcus! And to all of you listening, thanks for tuning into Allur. If you enjoyed this episode, please subscribe and leave us a review—it really helps the show. We’ll be back next week to talk about the latest in Go and mobile dev. Until then, keep building! Or, I guess, keep *piloting*.