A recent report by the tech consultancy firm Thoughtworks highlights the rapid expansion of artificial intelligence (AI) tools and techniques within the software industry, as organisations increasingly seek to optimise large language models for practical purposes. Despite the potential benefits, the report also underscores the challenges posed by the improper use of these tools.
In its latest edition of the Technology Radar, Thoughtworks finds that approximately 40% of the 105 identified items, which include tools, techniques, platforms, languages, and frameworks deemed “interesting,” are related to AI. Sarah Taraporewalla, who leads Thoughtworks Australia's Enterprise Modernisation, Platforms, and Cloud practice, has noted that AI solutions are proving their worth beyond the initial market hype.
During an interview with TechRepublic, Taraporewalla explained that for a tool or technique to be included in the Technology Radar, Thoughtworks' teams must be actively using it. This utilisation enables them to generate informed opinions on its effectiveness. About 40% of the AI-related features identified in the Radar are reportedly derived from ongoing project work across Thoughtworks' global offices.
The Technology Radar serves as a tool for tracking emerging trends within the software engineering domain. It provides guidance to technology buyers by categorising these trends into four ratings — "adopt," "trial," "assess," and "hold."
The report advises companies to “adopt” retrieval-augmented generation techniques, signalling these as the preferred methods for enhancing the quality of responses generated by large language models. Techniques like using one large language model to evaluate another are given a “trial” status, meaning they are ready for use but not yet fully proven. The GCP Vertex AI Agent Builder, which facilitates the creation of AI agents, also falls into the “trial” category.
Taraporewalla mentioned that for a technique to be marked for “trial” status, it must have already been moved into production and have demonstrated success in practical applications. She remarked that the rapid development and deployment of AI tools reflect a significant shift in client expectations, particularly within the Asia-Pacific region.
The swift adoption of AI technology is also creating antipatterns — counterproductive patterns — that could lead to adverse results for organisations. A notable issue identified involves an over-reliance on AI-driven coding suggestions without sufficient human oversight, which could degrade code quality.
The report strongly cautions against replacing traditional pair programming with AI models, urging companies to maintain robust engineering practices to ensure code integrity and maintainability.
Addressing the shift in the AI toolscape, Taraporewalla advises organisations to prioritise understanding the problem they face before selecting the most appropriate technological solutions. Emphasising AI governance, she recommends establishing dedicated teams to define standards and continuously monitor developments within the AI landscape.
Moreover, Taraporewalla suggests that organisations can enhance their operations by embedding governance features into their AI platforms, allowing for structured experimentation while mitigating risks.
Despite the need for careful management, experimenting with AI tools could yield lasting benefits. Taraporewalla notes that such experimentation can bolster an organisation's platform and capabilities over time, emphasising that a testing mindset can ensure that AI investments yield returns over the long term.
This comprehensive analysis from Thoughtworks provides valuable insights for organisations navigating the fast-evolving landscape of AI technologies, outlining both the opportunities and challenges inherent in this burgeoning sector.
Source: Noah Wire Services