News

November 21, 2026

Why software engineering needs AI engineer team model – Expert

Why software engineering needs AI engineer team model – Expert

Technology leader and Chief Technology Officer, Olukotun Olanrewaju James, has called for a shift in how organisations integrate artificial intelligence into software development, advocating a coordinated AI Engineer Team as a Service (AETaaS) model to improve software delivery.

James said while AI coding assistants had made it easier for developers to generate functions, explain code, identify errors and write software tests, faster coding alone did not guarantee efficient and reliable software delivery.

In an article examining the changing relationship between artificial intelligence and software engineering, he argued that software development involved more than writing code, requiring business analysis, requirements definition, system architecture, quality assurance, governance and operational accountability.

According to him, organisations that introduce AI only at the coding stage risk accelerating existing weaknesses in their development processes rather than improving overall performance.

“When AI is introduced only at the coding stage, it inherits the weaknesses of everything that comes before it. Although it can generate code quickly, it cannot compensate for a poorly understood business problem,” he said.

He explained that AI could suggest architectural solutions and generate software tests, but engineers must evaluate the recommendations against security, scalability, integration requirements and intended business outcomes.

James said the AETaaS model offered an alternative by integrating skilled engineers, AI-enabled workflows and structured delivery practices into a coordinated engineering system.

Under the approach, artificial intelligence would support different stages of software development, including requirements analysis, planning, design, implementation, code review, testing and documentation, while human engineers would retain responsibility for decisions, validation and delivery.

He stressed that effective AI integration should begin with understanding business problems rather than relying solely on prompts to generate code.

Although AI tools could help identify gaps and omissions in project requirements, he said stakeholders and engineers must validate the findings to ensure that the actual needs of the business were properly understood.

He also emphasised the importance of architectural planning before implementation, noting that AI could assist in comparing different technical approaches, but engineers must assess the implications and trade-offs of each decision.

James recommended that software projects be developed incrementally, with systems divided into manageable modules supported by clearly defined acceptance criteria.

He said this would allow development teams to receive feedback earlier, identify problems more easily and validate each stage before proceeding to the next.

On quality assurance, he maintained that testing should remain an essential part of the development process rather than an activity reserved for the final stage.

“AI-generated code still requires the same level of testing and review as code written by human developers,” he noted.

The technology leader further warned against treating AI-generated outputs as final decisions, insisting that human accountability must remain central to software engineering.

He advocated the use of version-controlled documents, recorded design decisions and review processes to ensure that technical responsibilities and decisions remained traceable.

Comparing traditional software development practices with an integrated AI engineering approach, James said the latter placed greater emphasis on coordinated workflows, structured business analysis, continuous quality validation and clearly assigned responsibilities.

He argued that organisations should measure the success of AI adoption by improvements in overall engineering performance rather than the frequency with which developers use AI tools.

James identified the DORA delivery metrics, including lead time for changes, deployment frequency, change failure rate and recovery time from failed deployments, as useful indicators for assessing the impact of AI on software delivery.

He, however, said organisations should also monitor requirements quality and review load to obtain a more comprehensive assessment.

According to him, requirements quality measures omissions, ambiguity, change requests and rework resulting from poorly defined project expectations, while review load examines the amount of senior engineering time required to assess AI-generated work.

He warned that without proper monitoring, the time saved through AI-assisted coding could be lost through increased rework and additional review responsibilities.

James advised organisations to establish performance baselines before introducing AI-powered engineering processes and compare subsequent results to determine where the technology was delivering measurable value.

He maintained that the future of software engineering would depend on collaboration between human expertise and artificial intelligence rather than competition between the two.

“The strongest engineering organisations will be those that combine human knowledge, AI capabilities, structured processes and measurable accountability. The real advantage will not come simply from adopting more AI tools, but from improving how software engineering teams work together,” he said.