August 17th, 2026 by Camille Baumann
AI QA QE AgenticAI SpiraPlan AIAssurance
AI is not making quality assurance less important. It is expanding the QA mandate.
For decades, quality teams have focused on a familiar set of questions: Did the software meet its requirements? Did the integration work? Did a new release introduce a regression? Did the system produce the expected result?
Those questions still matter. But as organizations begin deploying generative AI and autonomous agents into production, quality leaders face an additional challenge: assuring systems whose behavior cannot always be specified in advance.
AI Agents Raise The Stakes
An AI agent may retrieve information, interpret instructions, select tools, call APIs, generate content, interact with other agents, and take actions across a business workflow. Each individual component may work as designed while the overall system still produces an unacceptable outcome. This does not make traditional QA obsolete, instead it makes traditional QA incomplete on its own. For Heads of Quality, Engineering, Delivery, and Product, the opportunity is to modernize the quality function around three emerging gaps: human capacity, enterprise governance, and operational control.
Gap 1: Human Oversight Cannot Scale With Agentic Complexity
Traditional testing works especially well when teams can define expected behavior in advance. A user performs an action. The system receives an input. The application should return a specific result. QA can compare actual behavior with expected behavior and determine whether the test passed or failed. Agentic systems make that model more complicated. A single agentic workflow may involve multiple prompts, retrieved documents, model calls, APIs, tools, decisions, and interactions with other agents. Different inputs or contextual changes can cause the system to take different paths through the same process. The challenge is therefore not simply that AI is probabilistic. It is the number of possible interactions that emerge as AI systems become more autonomous.
Consider an agent that has access to customer information and several business systems. The individual API integrations may all function correctly. Authentication may work. The user interface may pass every regression test. But the agent could still interpret an instruction incorrectly, retrieve inappropriate context, select the wrong tool, and execute a technically valid action that produces an undesirable business outcome. Human reviewers cannot manually anticipate every possible permutation of those behaviors.
That is the first gap modern QA organizations need to address: quality teams need technology that augments human judgment rather than relying exclusively on human capacity.
Inflectra.ai helps extend that capacity by applying generative AI within the software lifecycle. Inflectra's platform uses Amazon Bedrock to help generate tests, tasks, code, and risks from business requirements and user stories. The objective is not to replace quality professionals. It is to help them analyze more information, explore more scenarios, and identify potential risks earlier in the lifecycle.That distinction will become increasingly important as the number and complexity of AI-enabled systems grows.
Gap 2: AI Experimentation Needs an Enterprise Governance Backbone
Many organizations begin experimenting with AI outside their established software delivery processes. A team may prototype prompts in one tool, record model evaluations somewhere else, track defects in another application, maintain compliance evidence in a spreadsheet, and manage production releases through a separate DevOps platform. That may be manageable during experimentation. It becomes much harder when AI starts participating in regulated, customer-facing, or business-critical processes.
At that point, quality leaders need to be able to answer basic governance questions.
- What requirement was this AI capability designed to satisfy?
- Which model or configuration was tested?
- Which evaluations were performed?
- What risks were identified?
- What defects remain open?
- Who approved the release?
- What evidence demonstrates that the system met the organization's quality and compliance expectations?
These are fundamentally traceability questions.
And they are already familiar to mature QA organizations. The difference is that AI expands the types of artifacts and evidence that need to be governed. SpiraPlan provides a lifecycle management backbone for connecting business requirements with development and testing assets. Inflectra's existing platform architecture is designed around complete traceability across these artifacts, helping organizations maintain a continuous view from business requirements through development and validation.
That traceability is especially valuable in regulated and high-assurance environments. The importance of this kind of lifecycle discipline is not theoretical. In an ORIX-METRO implementation, Inflectra's Spira platform supported compliance, traceability, auditability, requirements management, project management, and advanced QA. The customer reported a 45% acceleration in time-to-market, a 30% increase in project delivery efficiency, and a 35% reduction in costs. Production issues were also reported as being reduced by 10%. Those outcomes came from improving the broader software delivery lifecycle rather than treating quality as an isolated testing activity.
The same principle becomes even more important for AI. AI assurance should not become another disconnected toolchain. It should become part of the existing quality and governance lifecycle.
Gap 3: Quality Teams Need Control Over AI Behavior, Not Just Software Functionality
The third gap is perhaps the most significant.
Traditional quality assurance typically evaluates whether software performs a predefined function correctly.
AI assurance must go further. Quality teams need visibility into what an AI system was instructed to do, what information it received, which tools it used, what actions it took, and whether the resulting behavior stayed within acceptable boundaries.
That introduces a new layer of quality questions:
- Did the system follow the user's intent?
- Did the agent retrieve the appropriate source information?
- Did it call an authorized tool?
- Did it disclose information that should have remained private?
- Did it remain within organizational policies?
- Did a model change alter its behavior?
- Did an agent consume an unreasonable amount of resources while attempting to complete a task?
- Could an adversarial user manipulate the system into behaving differently?
These are not replacements for functional testing. They are additional dimensions of assurance.
The modern QA stack therefore needs to evaluate both deterministic software behavior and probabilistic AI behavior.
Traditional Testing Still Matters
One of the biggest mistakes organizations can make is assuming that the arrival of AI means conventional testing is no longer relevant. An AI application is still an application. It still has interfaces, authentication, APIs, databases, workflows, infrastructure, integrations, and release dependencies. All of those components require functional, integration, regression, and performance testing.
Rapise continues to address this layer of the problem by helping teams automate testing across web, desktop, mobile, and enterprise applications. The value of conventional automation can be substantial. For example, Telstra International used Inflectra's testing solutions to support Salesforce automation. The organization reported an 80% improvement in test coverage, while reusable regression automation allowed teams to execute full regression testing weekly and reduce their reliance on manual testing.
Those capabilities remain just as important in an AI-enabled application. The difference is that they now represent only one part of the assurance architecture.
AI Systems Require Behavioral Assurance
Once generative AI and agents enter the system, quality teams also need to evaluate behaviors that conventional regression tests were not designed to measure. SureWire extends assurance into this layer by focusing on generative and agentic AI systems.
Instead of asking only whether a predetermined output matches an expected value, AI assurance can evaluate whether a system continues to behave acceptably across different prompts, models, contexts, and adversarial conditions. That includes risks such as hallucinations, prompt injection, information leakage, policy violations, behavioral drift, and unsafe use of external tools. The objective is not to prove that an AI system can never fail. No responsible quality program should make that promise.
The objective is to systematically discover failure modes, establish acceptable behavioral boundaries, evaluate systems against those boundaries, and maintain evidence of how identified risks are being managed. That is a much more useful definition of AI assurance.
A Modern Quality Stack Needs Multiple Layers of Assurance
The QA organization of the future will not choose between traditional testing and AI assurance.
It will need both.
At the foundation, organizations still require clear requirements, traceability, release management, defect tracking, testing, and governance.
Above that, conventional automation validates the predictable components of the system.
And AI-specific evaluation examines the probabilistic behaviors introduced by models and autonomous agents.
For Inflectra, those layers can be connected through a common lifecycle:
- SpiraPlan provides the system of record for requirements, risks, tests, releases, defects, and governance.
- Inflectra.ai augments quality and delivery teams by helping generate and analyze lifecycle artifacts such as tests, tasks, code, and risks.
- Rapise automates functional and regression testing across the conventional software components surrounding AI systems.
- SureWire extends testing into the behavioral assurance of generative AI and autonomous agents.
Together, these capabilities reflect an important shift in how organizations should think about quality.
Modern QA is no longer concerned only with whether the software works.
It must increasingly demonstrate that the entire intelligent system behaves within acceptable boundaries.
QA Is Becoming More Strategic, Not Less
The rise of AI has naturally created uncertainty about the future of many technology roles.
For quality professionals, however, the more important question may be how quickly the role can expand.
Autonomous systems create new risks precisely because their behavior is more difficult to anticipate. Organizations therefore need people who understand requirements, risk, testing, governance, evidence, failure analysis, and system behavior more than ever.
The QA leader's role can evolve from managing a testing function to establishing the assurance architecture for intelligent systems.
That means closing three gaps.
The first is the capacity gap: using AI to augment human analysis as system complexity grows.
The second is the governance gap: connecting AI development and evaluation to enterprise requirements, risks, releases, and compliance evidence.
The third is the control gap: continuously understanding whether autonomous systems are operating according to their intended purpose and within acceptable behavioral boundaries.
Organizations that close those gaps will be better positioned to scale AI without abandoning the quality disciplines that made reliable enterprise software possible in the first place.
The New Quality Question
Traditional QA asks:
Did the system do what we specified?
Modern AI assurance adds another question:
Can we demonstrate that the system continues to behave acceptably when the answer cannot always be specified in advance?
That is the opportunity in front of quality leaders. Not to discard the practices and technologies that already work, but to extend them. Because as software becomes more autonomous, organizations will need stronger assurance—not less. And the quality teams that modernize now have an opportunity to lead that transition.
Modernize Your Quality and AI Assurance Lifecycle
Inflectra helps organizations connect requirements, software delivery, test automation, AI assurance, risk, and governance within a traceable lifecycle.
Explore how SpiraPlan, Inflectra.ai, Rapise, and SureWire can help your organization extend traditional QA practices into the era of intelligent and autonomous systems.
