AI for Test & Measurement

Embracing AI: A Necessity in the New Era of Test

The use of AI within test and measurement has been transformed with the invention of LLMs that can extract value from large unstructured data sets. This capability creates new, highly efficient test methodologies and development processes that are necessary because AI tools are also accelerating product design cycles upstream, meaning that test teams are facing more projects, more complexity, and shorter timelines. However, this opportunity also brings risk, as technology change is outpacing the industry’s ability to formally define best practices and standards. 

Our role is to help engineers navigate this change using NI software, other paid industry tools, and open-source technology, to identify the right pace and placement of AI adoption to maximize engineering productivity, without endangering product quality. 

We build on top of leading AI technology, adding test expertise from more than 50 years as a global leader in software-defined test, to help test groups get the most out of AI. NI Nigel™ AI benefits from the rapidly evolving power of commercially available large language models, focusing and augmenting their output towards the unique challenges of test. A test-optimized AI like Nigel can help you plan your test strategy, develop IP, assign work, get actionable insights (including DUT failure points, measurement integrity, and test station operations), and generate product reports. Teams that know how to use AI effectively across more of their workflow of tasks will be more effective than those who don’t use AI or use it in a limited capacity.

Three AI Operation Modes for Test Engineers

Test engineers can leverage these three distinct operational modes to improve productivity and efficiency:

In-context advice—When compared to humans, AI excels in finding specific information within large data sets. Test engineers receive immediate value and minimal risk when using AI to help with tasks like finding hardware specifications or wiring diagrams, building tables that show current system utilization or performance, or locating sources of functionality or error within a code base. Engineers using a test-optimized AI will get more concise answers with less chance of hallucination, because a test-optimized AI like Nigel has access to structured web data, data sheets, and other product documentation. Nigel also has context of connected hardware and software modules, meaning less theoretical answers and more accurate and targeted responses towards a practical, implementable solution. 

Code Generation and AI authorship—Using AI to generate code allows engineers to work faster—and, when used properly, even quality-conscious test groups can benefit from the abstracted workflows it offers. Engineers remain accountable for the correct operation of the application; therefore, generating code that is easily read, understood, debugged, and validated becomes paramount. In this new era of AI-assisted test software development, accurate high-level representation of complex systems, standardized trustworthy code modules, coding best practice, and design review become more important—not less.

General-purpose software tools allow engineers to work much more quickly, but their lack of guardrails, lack of test specific context, and lack of integration into existing processes introduce risk and disruption. Test-optimized tools, such as using Nigel within the NI LabVIEW+ Suite, can minimize this risk by drawing from trusted IP and methodologies and keeping development visible to the engineer at every step. 

With token cost rising, test teams should also look to generate only the unique code they need by building on existing applications where possible. For example, when prompting your AI to generate a test program, getting it to build upon the powerful sequence engine that already exists within NI TestStand is much quicker, cheaper, more maintainable, and more reliable than generating a new application from scratch. 

Outcome-driven automation with agentic AI—Test engineering is a much more extensive discipline than just IP development. By looping the prompts being delivered to the LLM and providing access to skills, information, and tools, engineers can work at a highly elevated level and receive significant productivity gains. Using agentic AI prompts no longer form a prescriptive question, but rather describe an outcome or need such as: When interacting with an AI in this way, user prompts shift from prescribing specific tasks to defining the desired outcome instead. For example:

“Nigel, help me test this PMIC using available test hardware. Here’s the datasheet R&D has given me explaining the required behavior."

For agentic AI to be effective, the AI must have deeper, domain-specific knowledge than when acting as an advisor or author. It also requires the engineer to have a strong ability to define the outcome they need and recognize success or failure in delivering it. Nigel AI has proprietary skills and contextual connectivity across the NI software toolchain, enabling it to provide better quality results and integrate more easily into existing processes. Its open connectivity enables it to work with other agents, bringing test methodology skills to other development workflows.

Test Engineering Requires an Open Approach to Software—Including AI

Most successful test groups use multiple software tools within their team. Rarely does a team have the luxury of rebuilding their test platform from a “blank slate.” Inherited code, tool-specific team members, and specialist IP can all dictate tool use, but test leaders should not feel constricted to standardize too much—using a software tool optimized towards the task at hand will yield the best results. 

Test-optimized AI workflows deliver the highest productivity gains when applications require classic test requirements such as hardware connectivity, signal processing, parallel execution, real-time deployment, and system planning/design. These workflows are less differentiated as the applications move into web or database development, sequential algorithm development, or report generation. Many test groups are establishing a hybrid approach as best practice, where the test-optimized tool is used to orchestrate the main application while other tools or code are integrated into this to provide specific functionality based upon their individual merit or preference. 

Table 1 provides considerations about where a test-optimized tool like Nigel AI has the greatest impact.

The Quality of Your Data Defines the Impact of Your AI 

Historically, much of the data acquired by test systems is discarded or underutilized as an operational byproduct rather than a strategic asset. Most test organizations are now racing to update data management capabilities to support AI-driven transformation that promises a wealth of new product, and operational insights. Consider these two best practices when formulating your AI strategy:

Metadata context is essential—To fully reap the benefits of AI, good measurement data must have more than just what the instrument measures. To perform anomaly detection or other root cause analysis, the context of how and where that measurement was taken is as important as the signal data itself. By aggregating environmental conditions, asset health information, code versions, and operator data into a holistic digital thread connected to a particular DUT, we can characterize a full understanding of any given measurement. Using NI software like TestStand paired with NI SystemLink can provide the context needed to quickly take the proper actions as the manual documentation of this context is abstracted from the user. 

Hardware choice matters—Many traditional test system architectures are optimized toward edge processing to minimize the bandwidth on communication networks. This architecture made sense when GPIB bit rate was the limiting factor. But in the new era of test, we must give our cloud-hosted models access to the full waveform of data, along with the full context about the environmental situation at the time of measurement. This capability lends itself to centralized hardware architectures where multiple instruments can stream time-synchronized data over a high-bandwidth PCI Express bus to a single processor that can supply information to the user and store the complete data set for future AI analysis. The NI PXI platform is ideal here, as it simplifies streaming and synchronization across all the instruments within a system allowing deeper insights into test and operation.

The Future of Test Is Inextricably Linked to AI

Access to AI for test and measurement will fundamentally change the way engineers develop test systems. From the first requirements document to the final published report, AI can simplify the complexity of development, providing the information you need to make sound decisions and take time-consuming tasks off your to-do list, so you can focus on your area of expertise.

 Eventually, AI will help you design complete test systems. As experts in the test and measurement field, we are charging toward that future. General-purpose AI tools will drive investment into the underlying models, and by building on top of these technologies, Nigel AI will inherit their increases in intelligence and power; general-purpose models alone will not satisfy the complex needs of test and measurement users.

Because of our decades of test experience and a foundation of well-organized, contextualized measurement data, Nigel will continue to evolve as a powerful, test-specific intelligence.