Jonathan Lalou's Blog

Lecturer

Luke Higgins serves as the Chief Architect for global asset and automation deployment at Accenture, where he focuses on integrating artificial intelligence and automation into service delivery frameworks. With over twenty years at Accenture, Luke has contributed to numerous innovations, including award-winning projects in generative AI applications. Kishor Panth leads global asset engineering at Accenture, overseeing the development of software assets, tools, and automations for client solutions. With more than twenty years in the firm, Kishor specializes in applying AI and automation to software development and platform management.

Abstract

This comprehensive analysis delves into the integration of generative artificial intelligence within service delivery platforms, drawing from Accenture’s experiences with its proprietary GenWizard system. It explores the contextual evolution from traditional automation to AI-driven workflows, methodological approaches to embedding Amazon Q and foundation models, and the broader implications for operational efficiency, decision-making, and innovation across technology lifecycles. By examining real-world applications in software engineering, application management, and platform optimization, the article highlights how these technologies foster accelerated project timelines and customized client outcomes.

Evolution from Automation to Generative AI in Service Delivery

The journey toward incorporating generative AI in technology services reflects a shift from rule-based automation to intelligent, adaptive systems. Initially, efforts focused on streamlining processes through predefined scripts and AI models for tasks like anomaly detection and predictive maintenance. However, the advent of large language models introduced capabilities for natural language processing and code generation, transforming how organizations approach software development and operations.

At Accenture, this evolution culminated in the GenWizard platform, designed to enhance service areas by leveraging AWS services such as Amazon Q. The platform addresses challenges in managing complex technology lifecycles, where traditional methods often led to inefficiencies in code migration, application rationalization, and incident resolution. By infusing generative AI, GenWizard enables forward and reverse engineering, allowing for rapid analysis of legacy systems and generation of modern equivalents.

This transition was driven by the need to handle vast codebases—often millions of lines—across diverse languages like COBOL, Java, and .NET. Reverse engineering, for instance, involves creating visual representations of code structures to identify dependencies and inefficiencies, while forward engineering automates the creation of new code based on specifications. The integration of Amazon Q facilitates natural language queries, making these processes accessible to non-experts and accelerating timelines from months to days.

Methodological Integration of AWS Services in GenWizard

GenWizard’s architecture employs a multi-agent framework powered by AWS foundation models, where agents specialize in tasks such as code analysis, generation, and testing. This methodology draws from software development best practices, incorporating continuous integration and deployment loops to ensure reliability.

A key component is the use of Amazon Q for contextual understanding and response generation. For example, in code migration, agents analyze source code, infer intent, and produce target language equivalents, followed by automated testing against predefined criteria. This reduces human error and enhances consistency, as demonstrated in projects converting legacy mainframe applications to cloud-native formats.

In application management, the platform’s event operations module rationalizes incident tickets by identifying duplicates and correlating issues with configuration management databases. This involves clustering related events and suggesting resolutions from a knowledge base, significantly cutting resolution times.

Platform engineering benefits from predictive analytics, where AI models forecast resource needs and optimize configurations. The methodology emphasizes data-driven insights, using metadata from development workflows to inform decisions.

Code sample illustrating a basic agentic workflow in Python, simulating code generation and testing:

“`
def generate_code(specification):