> ## Content Index
> Fetch the complete content index at: https://www.edgewisely.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# Why Companies Are Building, Not Buying
- URL: https://www.edgewisely.com/why-companies-are-building-not-buying/
- Published: 2026-09-06T04:36:55.000Z
- Updated: 2026-09-06T04:36:55.000Z
- Description: McKinsey says 32% of enterprises have skipped a software purchase because agentic coding tools let them build it themselves. CodeRabbit's $143M raise is a bet on what that shift creates.
- Author: John Karpentar
- Tags: Opinion, Enterprise

**How agentic coding tools are quietly turning a third of enterprises from software buyers back into software builders.**

**A McKinsey survey says 32% of organizations have already skipped a software purchase because they could build the thing themselves. The funding market just placed a very specific bet on what happens next.**

For two decades, the safe default in enterprise technology was to buy. Building software in-house was slow, expensive, and usually worse than what a specialized vendor already sold. That calculation is now visibly cracking, and the evidence isn't a thought experiment — it's a named percentage in one of the industry's most-cited surveys, sitting next to a fresh nine-figure funding round for the company selling the tools that make in-house building plausible again.

## The number: 32%

McKinsey's [State of AI 2026 global survey](https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai-in-2026?ref=edgewisely.com) — 1,719 responses across 97 countries, collected between May 4 and June 8 — found that 32% of organizations have decided against buying a software product or feature because they can now build it in-house using agentic coding tools. That's not a hypothetical preference. It's a third of surveyed companies reporting a purchase decision they actually made differently because the build option got cheaper.

The number isn't evenly distributed. Among the 6% of respondents McKinsey classifies as "high performers" — companies that attribute at least 5% of their EBIT to AI — nearly half are skipping software purchases, against 31% of everyone else, according to [reporting on the survey](https://finance.yahoo.com/technology/ai/articles/build-vs-buy-shift-32-113806700.html?ref=edgewisely.com). Large enterprises with more than $1 billion in revenue are moving fastest of all: 40% are now scaling agents in at least one function, up from 27% a year earlier. By industry, technology companies lead at 41%, with healthcare payers and providers at 39% and professional services and energy both at 38%.

Lieven Van der Veken, a senior partner at McKinsey, frames it as a deliberate pivot rather than an accident of AI hype: "Leaders are asking what their organisations need to build AI tools themselves." The more useful part of his framing is the discipline attached to it — the companies getting real value aren't building everything, they're getting more selective about which category each decision falls into, and treating the ongoing cost of running what they build as a design constraint from day one rather than a bill that shows up later.

## The company betting the same thesis is durable

Three days before McKinsey's number started circulating, CodeRabbit — a startup that reviews code for quality and security before it ships — closed a $143 million Series C at a $1.5 billion valuation, according to the [company's official announcement](https://www.businesswire.com/news/home/20260812311754/en/CodeRabbit-Raises-%24143-Million-at-%241.5-Billion-Valuation-and-Introduces-Agentic-Change-Management?ref=edgewisely.com). The round, co-led by Atomico and Smash Capital with participation from BMW i Ventures and Datadog, came less than a year after the company's $60 million Series B, on the back of revenue that grew more than 5x year-over-year. CodeRabbit now runs more than 2 million code reviews a week for over 17,000 customers, including Adyen, BMW, Indeed and Nvidia.

CodeRabbit isn't a coding-agent company itself — it doesn't write the code. It's the layer that checks whether code written by people, or increasingly by agents, is safe to ship. CEO Harjot Gill's framing of the problem is the same one McKinsey's survey is measuring from the other side: "Code changes now originate across the software organization, including from developers, non-technical personnel, as well as from coding agents to issue trackers, support systems, and production alerts. Every change creates a decision for the team." When more of an organization's software gets built internally instead of purchased, the volume of code that needs reviewing, governing and trusted doesn't shrink — it multiplies, often faster than the humans available to check it.

That's the honest read on why investors wrote a check at a 1.5 billion valuation for a code-review company right as the build-vs-buy number was making headlines: more building means more code nobody outside the vendor relationship has already vetted, which is a durable problem regardless of which specific coding agent wins any given deal.

## Why the enthusiasm outruns the results

The uncomfortable part of this story is that building more doesn't mean building successfully. Research from [MIT's NANDA initiative](https://www.aigl.blog/content/files/2025/09/The-GenAI-Divide-STATE-OF-AI-IN-BUSINESS-2025.pdf?ref=edgewisely.com) found that internally built AI systems succeed roughly 33% of the time, compared with about 67% for vendor-purchased tools — the inverse of what the build-vs-buy enthusiasm would predict. [Gartner has separately forecast](https://www.gartner.com/en/newsroom/press-releases/2025-06-25-gartner-predicts-over-40-percent-of-agentic-ai-projects-will-be-canceled-by-end-of-2027?ref=edgewisely.com) that more than 40% of agentic AI projects will be canceled by the end of 2027, citing escalating costs, unclear business value and inadequate risk controls — the same gap this publication has [tracked in enterprise AI pilots more broadly](https://www.edgewisely.com/why-enterprise-ai-pilots-fail-genai-divide/), where the distance between a working demo and a production system that survives contact with real usage keeps turning out to be wider than teams expect going in.

That gap doesn't make the McKinsey number wrong — it makes it a leading indicator rather than a finished result. Organizations are exercising a build option that's newly available to them; whether that option pays off is a separate, harder question that most of them haven't answered yet. The survey measures a decision that got made, not an outcome that's been proven.

## Stakeholder read

**For enterprise software vendors,** the threat isn't losing to a competitor's product — it's losing to a customer's own engineering team armed with a coding agent and a plausible reason to try building the feature themselves. The vendors best positioned to survive this are the ones selling something genuinely hard to replicate internally: deep domain expertise, compliance certifications, or infrastructure a single engineering team can't stand up in a sprint. A narrow point solution that a competent internal team could plausibly rebuild with a weekend and an [AI coding assistant](https://www.edgewisely.com/the-7-best-ai-coding-assistants-in-2026-github-copilot-cursor-devin-desktop-amazon-q-cody-replit-agent-and-tabnine-compared/) is exactly the category McKinsey's number says is now at risk.

**For engineering leaders inside the 32%,** the McKinsey data is less a victory lap than a warning about what comes next. Building the thing is the easy part with agentic tools; maintaining it, securing it, and keeping institutional knowledge about why it was built that way when the original engineer leaves the company is the part that doesn't get any cheaper. That's precisely the governance gap products like CodeRabbit, and the broader push toward [treating autonomous agents as something that needs a leash rather than free rein](https://www.edgewisely.com/obsidians-bet-on-the-agent-leash/), are built to fill.

**For investors,** CodeRabbit's valuation is a bet that the build-vs-buy shift is structural rather than a temporary side effect of a hype cycle, and that whichever layer of infrastructure sits underneath all that newly built internal software — reviewing it, securing it, explaining it to the next engineer who has to touch it — becomes as valuable as the tools that generated the code in the first place. It's a bet on volume and risk, not on any single coding agent winning the market.

## The zoom-out

The build-vs-buy pendulum has swung before — cloud computing pushed companies toward buying, open source periodically pushed them back toward building, and each swing looked permanent from the inside while it was happening. What's different this time is the mechanism: it's not that building got cheaper because engineers got cheaper to hire. It's that a category of software that used to require hiring engineers at all can now be produced by a much smaller team pointed at an agent, which changes the actual unit economics of the decision rather than just the sentiment around it.

The organizations that come out ahead in this cycle won't be the ones that build the most. They'll be the ones honest enough to notice, the way McKinsey's own data quietly shows, that a third of companies have already made a build decision the industry's own failure-rate research says is more likely to disappoint them than the purchase they skipped — and who build the governance and maintenance discipline to close that gap before the bill comes due.

## Frequently Asked Questions

### What is the "build vs. buy" shift in enterprise software?

It refers to a growing tendency among companies to build software internally using AI coding agents, rather than purchasing it from an outside vendor, because agentic tools have lowered the cost and time required to produce working software in-house.

### What did McKinsey's State of AI 2026 survey find?

McKinsey found that 32% of surveyed organizations have decided against buying at least one software product or feature because they could build it themselves using agentic coding tools. The trend is strongest among high-performing companies and large enterprises, and most pronounced in the technology, healthcare and professional services sectors.

### Why did CodeRabbit raise $143 million, and what does that have to do with build vs. buy?

CodeRabbit, which reviews and governs code before it ships, raised the round to expand its platform for validating software regardless of whether it was written by a human or an AI agent. As more companies build software internally rather than buying it, the volume of unreviewed, internally produced code grows, increasing demand for independent tools that check whether that code is safe to ship.

### Do internally built AI tools actually work as well as purchased software?

Not consistently, according to available research. MIT's NANDA initiative found internally built AI systems succeed about 33% of the time, versus roughly 67% for vendor-purchased tools, and Gartner has forecast that more than 40% of agentic AI projects will be canceled by the end of 2027 due to cost overruns, unclear value or inadequate risk controls.

---

*Editor's note — sources: McKinsey (State of AI 2026 report); Forkast News via Yahoo Finance; CodeRabbit (official announcement via BusinessWire); MIT NANDA initiative; Gartner.*