The first app your team builds with an AI assistant deploys from a laptop, and that’s fine. Someone runs one command, the app goes live, and the people who asked for it start using it.
The fiftieth app is where it gets messy. By then a dozen people deploy, each from their own machine, signed in as themselves. Ask which version of an app is live, or whether anything was tested before it shipped, and you get a shrug. When something breaks on a Friday, the one person who knows how it was deployed is on holiday.
AI assistants solved the coding. What breaks at scale is everything after the code is written.
Software engineering already has an answer for that part, and it’s decades old: continuous integration and continuous delivery. I’d argue vibe coding needs a pipeline more than hand-written software ever did.
More code, less stability
Google’s DORA research program has measured software delivery for over a decade. It watched AI arrive in real time.
In 2024, DORA found that AI came with a cost. As adoption rose, delivery throughput dropped an estimated 1.5% and delivery stability dropped 7.2%.1 Teams wrote more code and shipped it less reliably.
A year later, 90% of software professionals were using AI. Throughput had recovered. Stability hadn’t. DORA reports that AI adoption “does continue to have a negative relationship with software delivery stability,” and it names the reason:2
Without robust control systems, like strong automated testing, mature version control practices, and fast feedback loops, an increase in change volume leads to instability.
That list is a fair description of a CI/CD pipeline.
DORA’s headline for the report is that AI amplifies whatever a team already has. So a team with a pipeline gets faster. A team without one gets more change, more often, and nothing to catch it.
of software professionals now use AI, up 14% in a year
DORA, 2025
estimated drop in delivery stability as AI adoption rose
DORA, 2024
of application security exposures will come from vibe coding practices by 2027
Gartner, April 2026
Verification is the new bottleneck
Gartner comes at the same problem from the other side. In its Market Guide for Enterprise Vibe Coding Platforms, it calls the prototype-to-production gap the market’s central constraint, and describes what it calls verification debt: “The volume of AI-generated code is growing faster than teams’ capacity to verify it, shifting the cost from writing code to understanding and securing code that no human authored.”3
A hundred apps from a hundred builders, each changing weekly, will outrun any review board. The only check that keeps pace with machine-written change is one that runs on its own, on every change. That’s a pipeline.
Gartner is specific about it. Its guidance on governing citizen developers who vibe code tells organizations to “integrate automated code scanning into CI pipelines” and to “provide standardized deployment pipelines with built-in security checks, environment separation, and policy enforcement.”4 The Market Guide tells leaders to require security checks on generated code before deployment, “integrating automated testing and remediation into CI pipelines.”
It also expects the pipeline to decide who wins the market. Code generation is a commodity now, and differentiation “centers on governance controls, deployment pipelines, back-end integrations and team collaboration features.”3
Why the pipeline is the right control for AI-built code
You could put the controls somewhere else: training, a policy document, an approval committee. For AI-built software, I think the pipeline wins, for four reasons.
It doesn’t care who wrote the code. Gartner’s recommendation is to subject all platform-generated code “to the same enterprise application security controls applied to human-written code before deployment.”3 A pipeline does that by construction. The tests and the reviewer can’t tell whether a person or an assistant wrote the change, and they don’t need to.
The merge becomes the record. Every change that ships went through version control first. So when an auditor asks what changed, when, and who approved it, the answer is already in the history. The builder never had to keep notes.
Deploys stop belonging to a person. When a pipeline deploys, it signs in as itself. Production credentials stay out of laptops, and when a builder leaves the company, the release process stays behind.
A red check means it didn’t ship. When a build or a test fails, the pipeline stops before anyone touches the broken version. That’s DORA’s fast feedback loop. You find the problem in a check instead of a support ticket.
Merge to live
- BuildWith an AI assistant
- MergeA reviewed, recorded change
- CheckTests and scans in CI
- Live and governedDeployed by CI, not a person
The pipeline has to be the easy path
There’s a catch, and it sinks most governance programs. More and more of the people building apps with AI assistants aren’t engineers.
Ask an operations analyst to write a CI workflow file, set up environments and manage deploy tokens, and they’ll go straight back to their laptop. The pipeline becomes one more control that gets routed around.
Gartner’s research on scaling vibe coding calls for “approved vibe paths,” built by software engineering and handed to builders ready to use:5
Make the safest path the easiest path by providing approved templates, reusable components, secure integration patterns, standard APIs and deployment guardrails.
DORA’s data lands in the same place. 90% of organizations have adopted at least one internal platform, and DORA finds “a direct correlation between a high quality internal platform and an organization’s ability to unlock the value of AI.”2 The platform team builds the pipeline once. The builder never has to.
So for vibe coding at scale, the question I’d ask is how little a builder has to do to end up in a pipeline. The answer should be close to nothing.
What that looks like in Helix
That was the design goal for Deploy from CI in Tray Helix. One command, helix ci init, writes a GitHub Actions workflow into the app’s repository. The team picks how deploys start: by hand, on every merge to the main branch, or to staging on merge with production waiting for a reviewer in GitHub.
$ helix ci init --provider github --strategy staging-on-merge
# Writes deploy-staging.yml and deploy-production.yml into the repo
From then on, a merge is the deploy. CI signs in as a Tray API user, so every deploy is attributed to the pipeline. The job waits until the new version is live and fails if the deploy does, so the check on the commit tells you whether it shipped.
The workflow files are the team’s own. The tests and security scans Gartner recommends go in above the deploy step, and a failure stops the release.
Teams that already run their own CI keep it. Any CI service can install the Helix CLI and run the same deploy command.
Once the app is live, Helix runs it on managed infrastructure and governs it, with SSO, managed credentials and an audit trail. The pipeline decides what gets that far.
The worked example, with the three setups, the API user and what to watch for, is in Deploy from your CI pipeline.
Where this leaves IT
Gartner expects enterprise vibe coding platforms to help create 40% of internal tools at large enterprises by 2029, up from under 10% in 2025.3 Most of those apps will come from people who have never opened a CI configuration file. They shouldn’t have to.
The companies that get value from that wave will ship every AI-built app the same way, starting with the first one: committed, checked, deployed by the pipeline, and running somewhere IT can see it.
The ones without a pipeline will still get the code. DORA’s numbers show what comes with it.
Footnotes
-
Google Cloud, “Announcing the 2024 DORA report,” October 2024. As AI adoption increased, it was accompanied by an estimated decrease in delivery throughput of 1.5% and an estimated reduction in delivery stability of 7.2%. Back
-
Google Cloud, “Announcing the 2025 DORA report: State of AI-assisted Software Development,” September 2025. AI adoption reached 90% of software development professionals, a 14% increase from the prior year; AI adoption is linked to higher delivery throughput and continues to have a negative relationship with delivery stability; 90% of organizations have adopted at least one internal platform. Back Back
-
Gartner, Market Guide for Enterprise Vibe Coding Platforms, C.A. Swan, Manjunath Bhat and Bill Blosen, 28 April 2026, G00844703. GARTNER is a registered trademark and service mark of Gartner, Inc. and/or its affiliates in the U.S. and internationally and is used herein with permission. All rights reserved. Gartner does not endorse any vendor, product or service depicted in its research publications and does not advise technology users to select only those vendors with the highest ratings or other designation. Gartner research publications consist of the opinions of Gartner’s research organization and should not be construed as statements of fact. Gartner disclaims all warranties, expressed or implied, with respect to this research, including any warranties of merchantability or fitness for a particular purpose. Back Back Back Back
-
Gartner, “Govern Vibe Coding for Citizen Developers With Self-Service Platforms,” G00858202, 29 June 2026. Back
-
Gartner, “How to Scale Vibe Coding Using Low-Code Engineering Principles,” Mukul Saha, G00857758, 11 August 2026. Back
