SmartBear provides high-impact, easy-to-use tools for software teams to plan, build, test, and monitor great software, faster.

An application breaks. An AI engine, model, service or application (or all of the above) wrote the code. The AI ran the tests... so the AI is responsible, right? You’re not even close, says Dan Faulkner, chief product & technology officer at SmartBear Software, a company known for its developer tools like Swagger, SoapUI, and TestComplete, which help teams test and optimise their software. Faulkner explains that the software development pipeline has grown longer and more complex, with AI agents and non-technical contributors joining what was once an exclusively human chain. But length doesn’t dissolve accountability. It concentrates it – with a human at the end. “The scale of this shift is hard to overstate. Our own research suggests that more than 9 in 10 development teams have adopted AI coding tools, with most generating at least 40% of their code through AI,” said Faulkner. Meanwhile, non-technical colleagues are entering the build pipeline faster than ever, using natural language interfaces that let them describe what they want without necessarily understanding what they’ve changed. What are AI agents doing? The SmartBear team say that AI agents are “taking on a growing share of testing work” today and absorbing the repetitive, lower-risk tasks that once consumed engineering hours. Together, these developments redraw the map of who does what in software development – and who answers for it. “Before AI, the chain of custody was clear. A human developer wrote code, understood its dependencies, and could trace how a change might ripple through the system. Now, code generation is distributed across agents and contributors who may not have full visibility into what they’ve touched or affected. No more centralised responsibility. It’s a shared one – stretched across large hybrid teams with widely varying levels of technical depth,” explained Faulkner. As agents absorb the lower-level work, we know that (in theory at least) humans move up the stack. Verifier VIPs at the vanguard The developer or quality engineer of today is less often the one writing the code or running the test, and more often the one standing at the end of the line, making the call. “That role – the verifier – is not a consolation prize for being replaced. It’s the job that matters most,” Faulkner says. His feeling is that verification demands judgment: knowing what to trust, what to question, where the model’s confidence outpaces its accuracy, and what a passing test actually guarantees. Those are not things AI agents can do for themselves. The chain of custody for software quality still exists. It has grown longer. Extra helping of honey The Computer Weekly Developer Network (CWDN) quizzed Justin Collier, senior director of product management at SmartBear for more. CWDN: If AI now writes 40% of code, how do you pinpoint accountability when a specific production failure originates from an AI-generated commit? Collier: You’re trying to determine where the issues are so that you can ensure your end users have the best digital experience possible because application integrity is intact. Accountability matters for preventing future issues, but the fact is that bugs get shipped, period. AI-generated code hasn’t changed the steps in the SDLC; the step after committing the code is testing. If a bug reaches production, that means it was missed in testing. And now the goal is to discover and resolve it as quickly as possible. SmartBear’s Collier: There’s a concept from Jocko Willink & Leif Babin’s book Extreme Ownership: If we all operate in a model of extreme ownership, then I own it, you own it, we own it. AI is a tool in our tool chest, and it’s only as effective as the person wielding it. Observability solutions, like BugSnag, help users pinpoint production failures by capturing runtime data, including stack traces, breadcrumbs, performance, and trace data, that helps teams find the needle in the haystack so that developers can fix issues swiftly, deploy a fix, and ensure a significant portion of their user base never experiences the issue. Also, AI testing solutions like BearQ can map your application, run user journeys end to end, and then update the tests as your application changes. The beauty of BearQ is that it can test the areas of your application that you couldn’t traditionally cover or didn’t even think to cover. All of us want to deliver the best digital experience we can for our customers. Once you’ve pinpointed the issue, repository tools and commit history will help you uncover where the commit came from, human or AI. But the engineer who approved and shipped it owns the outcome regardless of it being AI-generated. CWDN: Technically, how does SmartBear’s tooling help “verifiers” distinguish genuinely trustworthy AI test results from confidently wrong ones? Collier: You have to understand who wrote the tests. Did humans write the tests and AI execute them or did AI write the tests and also execute them? In either scenario, being crystal clear about the intent of a particular feature helps both humans and AI build tests that accurately reflect what we want the end user experience to be. SmartBear’s tooling delivers that trustworthiness by anchoring tests to something verifiable, like an API definition or a contract, so the test isn’t just asserting an outcome; it’s checking that outcome against a spec. For UI testing on web and mobile, our AI testing solutions can turn natural language test steps into automated actions and update themselves automatically as the UI changes. Additionally, our fully agentic QA Agents validate how the application actually behaves and produces reports when issues arise so that a human can validate and modify if needed. Regardless of your AI testing strategy, having a clear understanding of your test steps, product intent, and AI tests that surface edge cases is critical for both humans and AI to better understand where problems arise. CWDN: Beyond the “9 in 10 teams” stat, what other numbers show how verification workload has shifted since AI agents joined pipelines? Collier: A few stats that stand out to me about how AI has improved the pipeline: A SmartBear customer was using runtime data and AI agents to suggest code fixes based on errors that were being seen in production. They’re approving more than 70% of the pull requests, which means their developers are writing features, not fixing bugs. Customers using AI-powered no-code automation reduced manual testing time by 50% and accelerated release cycles by 20%. A customer used runtime data and an AI agent to pinpoint an issue in 15 minutes versus the one hour it previously took them for similar issues. Customers can come in the morning, see the AI tests that were written overnight, validate and approve them, and save an estimated 75% of their time. What’s fascinating is the amount of data that AI can process, and how much of it points a human straight to what actually needs attention. We take the data, fix issues fast, and create new tests, so we don’t see the same regression twice. Coding is getting quicker with AI, but so is the rest of the SDLC. CWDN: Who exactly holds accountability now – the developer, the non-technical prompter, the AI vendor, or the verifying engineer signing off? Collier: At the end of the day, the engineer who approved the code is responsible. There’s a concept from Jocko Willink and Leif Babin’s book Extreme Ownership that sums this up well: if we all operate in a model of extreme ownership, then I own it, you own it, we own it. AI is a tool in our tool chest, and it’s only as effective as the person and team wielding it. Just like the physical tools in my garage, I have to take care of them and service them to ensure that they are going to work properly, and then I’m also responsible for the outcome once I’m done with the job. AI is no different. I must make sure I provide the correct context, limit the scope, verify that the outcome is what I’m looking for, and either accept or decline it.

Enroll in TestComplete Training at BITA Academy. Learn automated testing tools online from experts and boost your QA testing career today.

SmartBear this week extended its ability to apply artificial intelligence to application testing with the addition of a computer vision capability that can be used to detect objects in applications. Prashant Mohan, vice president of product at SmartBear, said the company’s TestComplete platform now combines computer vision and optical character recognition to help automate the […]

Curate Feeds | Make Collections | Customize Email Briefs

QA Automation Engineer OneMain Financial holds a strong commitment to our customers, communities and team members. This approach to lending has helped us

QA Automation Engineer OneMain Financial holds a strong commitment to our customers, communities and team members. This approach to lending has helped us

SmartBear today announced the first platform to unify property-based, OCR, and vision AI — perceiving applications the way humans do.

SmartBear this week extended its ability to apply artificial intelligence to application testing with the addition of a computer vision capability that can be used to detect objects in applications. Prashant Mohan, vice president of product at SmartBear, said the company’s TestComplete platform now combines computer vision and optical character recognition to help automate the […]

SmartBear provides high-impact, easy-to-use tools for software teams to plan, build, test, and monitor great software, faster.

SmartBear provides high-impact, easy-to-use tools for software teams to plan, build, test, and monitor great software, faster.

SmartBear provides high-impact, easy-to-use tools for software teams to plan, build, test, and monitor great software, faster.

SmartBear provides high-impact, easy-to-use tools for software teams to plan, build, test, and monitor great software, faster.

SmartBear provides high-impact, easy-to-use tools for software teams to plan, build, test, and monitor great software, faster.

エクセルソフトが開催するハンズオンセミナーで、TestCompleteを使用したテスト自動化の操作を実際に体験できます。参加無料！

エクセルソフトが2025年7月25日に開催するTestCompleteオンラインハンズオンセミナーの詳細をお届けします。

Lead QA strategy; manage QE teams; automate & standardize testing (Jira, TestRail, Selenium, Jenkins, TestComplete); partner w/ Product, Engineering & DevOps across SDLC/CI-CD. Req. master's degree in Comp. Sci. or SE or for. eq. & 2 yrs of exp. as a Sr. QA Anal. or Qual. Eng Mgr. or Mgr. Testing. 2 yrs prof. exp. w/Jira, TestRail, UAT Testing, Jenkins, Confluence, Selenium, & TestComplete Reqd. Sal: $108,200 Med, Dental, LI, EAP, Disab., 401k. Occ. U.S. travel up to 20%, avail for empl. at various sites in U.S. No fixed itinerary. Contact: Qual. Applicants send resume to talent@bounteous.com and Ref. job title & Job #B0013 in subj. line. ----------------------------------------------------------------------------------- JOBS.NOW Note: To tap into these hidden job opportunities, it's crucial to adhere strictly to the application process outlined in each job ad. At JOBS.NOW, we ensure that every listing includes detailed employer instructions. Follow them precisely to be considered for these unique positions! The "Log Application" button simply allows you to log the application for your records - JOBS.NOW does not submit any applications to employers directly. Remember to still apply through the method indicated in the job ad (mail, email, or via link). Please note that JOBS.NOW is an independent website and does not post this listings on behalf of any employers nor do we receive any compensation for these listings. All listings are sourced via media or internet channels required by the PERM process.