Kevin Thompson is Tricentis‘ Chief Executive Officer and Executive Chairman of the Board.
Since ChatGPT launched in November 2022, generative AI has promised to transform the workforce. Countless businesses are implementing integrations that streamline and optimize processes and systems. In fact, the results of a recent Gartner survey show that 70% of the business leaders who responded are now exploring generative AI for use in their companies.
Four Best Generative AI Practices
With more companies beginning to leverage generative AI to bolster various processes—particularly around software development and quality engineering—it’s becoming clear the tool is here to stay. As such, there are four best practices businesses should consider as generative AI becomes a more permanent part of operations.
1. Put together a cross-functional team.
Although many companies have an AI team or department, relying on them alone can cause bottlenecks, hinder productivity and lead to disconnects between teams, or silos. Instead of a lone AI team, companies should look to establish a cross-functional support team with members from a variety of different backgrounds.
With a cross-functional team, companies can investigate a more diverse set of questions around AI’s impact: What are the opportunities and how do we apply them and become more knowledgeable? How do we build a roadmap and know what to do with it? How do we create standard approaches rather than a series of custom approaches for the use of generative AI? How do we adequately train our employees on the appropriate use of AI? What are the risks in terms of laws, regulatory compliance and ethical issues? Ultimately, this team provides perspectives not just from “AI technologists” but also from domain, legal and security experts, as well as product managers and analysts.
2. Harness your data.
In today’s digital-first world, adopting a data-driven approach is critical. Generative AI makes harnessing and leveraging this necessary data easier than ever, but it’s only as good as the data that’s fed into these models.
For example, quality engineering tools generate a plethora of data, such as test results, defect reports and performance and usage metrics. Generative AI can harness this information to provide unique insights specific to that company’s unique context, tests and results. AI can also be implemented in continuous monitoring mechanisms to capture real-time data on system performance, usage patterns, log analysis, anomalies, user feedback and more.
Generative AI enables teams to better synthesize and leverage information, proactively identify issues, iterate and drive continuous improvements, and spot failure before it occurs. For quality engineering specifically, generative AI can be used to make test case recommendations and generate tests in tooling to significantly improve development efficiency. However, it’s important to note that this process will still require human oversight, as context matters in the development of test cases and the evaluation of test results.
3. Be “privacy alert.”
Harnessing your data can unlock unique competitive advantages, but it must remain your data. There are already high-profile lawsuits addressing the concerns of generative AI learning from data that the AI company doesn’t own. Once data is absorbed into the AI, it’s very difficult, if not impossible, to disentangle them without the wholesale reversion of the models. If organizations embark on custom training regimes, they should be aware of how those models learn data and consider using the model-per-customer approach or other similar techniques to guard against data leakage.
With the proliferation of GPT-enabled services, companies need to use caution when deploying services that use inputs as future training data. With the power of these networks to draw connections that aren’t explicit, it becomes extremely hard to anonymize data. Organizations should be vigilant about reviewing terms and conditions to validate data privacy, deletion and storage mechanisms. In some cases, adversarial testing may be required to validate the claims made.
4. Develop a rigorous verification and testing strategy.
Generative AI presents a unique challenge that’s evident in its name: It “generates,” meaning we expect it to create, not just regurgitate. This means that there will be times when the AI confidently says something that’s completely incorrect. Without a rigorous testing, verification and detection strategy, these inaccuracies may produce incorrect results or invalid representations, which are then passed on to customers.
As generative AI is used more frequently, it’s critical that companies implement solid strategies for testing and verification to assess reliability, accuracy and performance. A comprehensive testing framework includes synthetic scenarios used to proactively monitor services and detect issues, as well as adversarial testing, in which people are assigned to second-guess AI conclusions. It also includes defensive measures, such as bias prevention and ethical criteria.
Additionally, it’s imperative humans drive part of this testing strategy because much of what AI churns out is missing context or nuance. By utilizing human critical thinking, common sense and experiences, companies can help ensure that everything put into production fits existing systems and works as anticipated. Reinforcement learning from human feedback (RLFH) is what makes generative AI—and ChatGPT in particular—especially unique. Companies can better use generative AI by having stakeholders weigh in or provide sentiment before blindly trusting AI answers or assuming a false sense of completeness. This interaction can further improve the AI model over time along with the quality of its answers.
Although there are many advancements in this space, we must remember that the new workforce isn’t AI-driven but, rather, AI-augmented. Our vision converges the best of human and AI intelligence, and we firmly believe that the combination builds trust and confidence.
A Generative AI Future, Guaranteed
It’s practically a guarantee we’ll continue to see more generative AI in the modern workplace. Not only does this mean more innovation, but it can also lead to better-quality software and increased collaboration.
However, although it’s exciting to live in a time of fast innovation, companies should be careful when implementing AI. These best practices—putting together a cross-functional team, harnessing your data, maintaining privacy and developing a rigorous verification and testing strategy—can help companies navigate their ever-evolving landscape, stay nimble with innovative technology and drive both employee and company success.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
Read the full article here
