Testing and obtaining customer input for traditional software are relatively straightforward. For example, a “print feature” prints the file. It starts with a dialogue box that has the printing options. The logic and feature are the same for everyone.
When it comes to developing applications that use AI and machine learning, however, fundamental changes for providing input to the dev team are required. With AI-based software, each time a specific user interacts with AI algorithms, their query gets a new and different answer from the algorithm. The answer changes each time depending on that query or input. So how do you test all the potential questions to see if the AI algorithm provides good answers each time? This is the problem Adobe Technology Previews solves. Providing potential users with access to the algorithm to see if it gives good answers (and using AI tools to identify the quality of the algorithm’s output) is a game changer. Technology Previews is a new development approach that gathers the input necessary to optimize the algorithm and ensures that it works effectively in all cases.
Technology Previews, announced today in Adobe Analytics, bridges agile development and AI processes by letting end users test-drive new AI-enabled features or capabilities during the development cycle, delivering critical input in a timely manner that supports the agile dev team. Code updates are made quickly, allowing the changes to be evaluated before delivering the first production release, consistent with agile development. Without this new solution for obtaining customer input, the utility and accuracy of algorithms are hard to forecast, and dev teams would have to hope for the best.
As John Bates, director of Product Management at Adobe, noted, “In legacy applications, testing focused on logic and feature operation, both of which are quite finite. However, with AI-powered applications, the testing process must fundamentally change. Optimizing algorithms requires that they be tested by many users to ensure they work across many different scenarios.”
Technology Previews in Adobe Analytics uses AI to evaluate customer input and identify how algorithms need to be enhanced or modified. In many ways, customers become virtual product managers, delivering input and insight that they are uniquely able to provide. This “fast feedback” approach connects customers and the development team in ways that were never possible with legacy software development. The result is software that is far more production-ready when released and that customers can have more confidence in.
“We allow our customers access to a new AI- or machine learning-enhanced feature for 90 days, and this interaction speeds the delivery of this innovation to customers,” said Bates. “We can now deliver AI-powered solutions in a much shorter period of time, with certainty that the customer will get exactly what they need.”
The inability to effectively gather and incorporate user feedback is why some AI-enabled solutions must be updated soon after launch. This new technology for gathering customer input, pioneered by Adobe, is removing that risk. Technology Previews is a leading example of new development paradigms for intelligent applications that are necessary as AI and machine learning become commonplace.
As Bates observed, “The development process for next-generation AI- or machine learning-enhanced software had to evolve to include the ability to test and gather input in a manner that would consistently deliver outstanding solutions. We have moved from the waterfall process into agile processes. With Technology Previews, we can now use agile processes for AI-powered solutions.”