Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Repair Windows errors before they cause bigger problemsFix Now →iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
A legal software organization says it roughly tripled R&D output per engineer over 18 months by redesigning how work moved from requirements to release—not by relying on AI to write code faster alone. In Greg Ingino’s 2026 InfoWorld account, the biggest gains came from reducing handoffs, giving teams more end-to-end ownership, and putting quality and security checks into delivery pipelines. The figures are the organization’s reported results, not independently audited benchmarks.
What changed: fewer handoffs across the delivery lifecycle
The organization’s central change was to reduce the gaps between product, development, QA, security, and deployment operations. Instead of passing work through a chain of separate functions, teams moved toward end-to-end feature ownership, with AI agents assisting at stages of the lifecycle.
Ingino summarized the shift this way: “We assumed most of the productivity gain would come from AI writing code faster. It didn’t. The biggest gains came from getting rid of the handoffs between stages.”
The practical implication is that faster code production is only useful when the rest of the workflow can absorb it. A feature can wait as long for review, testing, security approval, or release preparation as it took to write. Reducing those queues can improve end-to-end throughput even when no individual coding task becomes dramatically faster.
#1 Best Overall
Where AI fit into the workflow
Requirements and shared context
The team created a shared repository for product and engineering context and used AI assistance in requirements work. Ingino reports that requirements work that had taken weeks could be completed in an afternoon. That result depends on having usable context available to the people and systems doing the work; AI assistance cannot compensate for missing or contradictory product decisions.
Implementation and pull requests
AI assistance was reported in about 3% of pull requests at the start of the period and 68% by the time of Ingino’s article. That is an adoption measure, not a measure of how much code AI wrote, how much review time it saved, or how many changes were accepted without modification.
Rank #2
Test creation
The organization reported that AI generated 99% of new tests and that its suite included more than 39,000 AI-developed tests. The account describes tests as tied to code changes and product requirements. Test volume alone does not establish test effectiveness: coverage, relevance, maintenance burden, and the ability to catch real regressions remain important.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
How quality and security controls kept pace
The approach paired AI assistance with quality and security gates embedded in delivery pipelines. Confidence scoring was used to route some routine approvals automatically, while lower-confidence work went to people. This is a risk-based division of labor: automate predictable checks where the process supports it, and retain human judgment for cases that need more scrutiny.
Rank #3
For another organization, the lesson is not to copy a particular confidence threshold—the account does not publish one—but to make the control path explicit. Decide which checks must pass before a change advances, which cases can be handled automatically, and which signals trigger human review. Then measure both throughput and outcomes, so automation does not simply move defects or risk further downstream.
What results the organization reported
Ingino’s InfoWorld article, published in 2026, reports the following changes over the 18-month effort. These are claims from the author and organization; the article does not provide an independent audit or the underlying data needed to reproduce them.
| Measure | Reported result |
|---|---|
| R&D output per engineer | Roughly three times the prior level |
| Releases | Nearly twice as many per quarter |
| Deployments | Increased from 82 to more than 155 |
| Customer-reported defects | Fell 65% per million lines of code |
| Pull requests with AI assistance | Rose from about 3% to 68% |
| Vulnerability density | Fell 76% |
| New tests | AI generated 99%; the suite included more than 39,000 AI-developed tests |
| Feature delivery example | A feature reportedly went from specification to working pull request in about four hours, versus 15 days previously |
| Requirements work | Reportedly shifted from weeks to an afternoon |
The article says the organization tracked DORA metrics, cycle time, pull requests merged per developer, and lines changed per developer against a fixed baseline. It does not publish the baseline values, metric definitions, team-size changes, or the underlying dataset. Those omissions make the reported gains useful as a case study, but not proof that another company should expect the same multiplier.
Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Repair Windows errors before they cause bigger problems3Scan for outdated or missing drivers - takes under a minuteWhy faster engineering can create a new bottleneck
As releases accelerated, the organization found that go-to-market readiness also had to accelerate. Documentation, enablement, customer-success briefings, and customer readiness could no longer be treated as work that happened after engineering was finished. If those activities lag, teams may produce deployable features faster than customers and internal teams can understand, support, or adopt them.
That means a throughput initiative should map the full path from product decision to customer-ready release, not only the coding and deployment stages. Look for queues and rework at each transition, including documentation and support preparation, then decide whether the constraint has moved after a change.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How to apply the case study without assuming the same results
- Choose one lifecycle area to improve. Start with a concrete bottleneck—such as requirements turnaround, review queues, test creation, or release approvals—rather than introducing automation everywhere at once.
- Establish a baseline and definitions. Track end-to-end cycle time and relevant quality or security outcomes before changing the workflow. Define how each measure is counted so that apparent improvement is not caused by a changed denominator or team size.
- Connect assistance to shared context. Make requirements and engineering knowledge accessible where work is performed, and keep product decisions authoritative and current.
- Build controls into the delivery path. Specify automated checks, confidence-based routing, and the conditions that require human review before expanding automation.
- Evaluate the whole system. Watch for changes in defects, vulnerabilities, release readiness, customer support load, and downstream queues alongside output and deployment frequency.
- Expand from evidence. Learn from the first area’s results and failure modes before applying the approach to more teams or lifecycle stages.
For any AI coding or delivery tooling under consideration, useful evaluation criteria follow from the workflow: which lifecycle stages it supports, how it uses existing work and knowledge systems, what quality and security controls it enables, how it escalates uncertain work, whether it measurably improves end-to-end cycle time, and whether downstream release preparation can keep up.
What the case study does—and does not—show
The account supports a focused takeaway: organizational design and fewer handoffs may matter as much as, or more than, faster code generation when the goal is more completed engineering work. It also describes AI use across requirements, implementation, and testing, paired with pipeline controls and human escalation.
It does not establish that AI alone caused the reported improvements, that the metrics were independently verified, or that the same results are typical. The published account lacks sufficient detail about baselines, team changes, measurement definitions, and data to isolate the effect of each intervention.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
