Salesforce Explores a Leaner Model for Software Development
Salesforce is testing whether artificial intelligence can change one of the fundamental equations of software development: how many people are required to build and maintain applications.
The enterprise software company is experimenting with AI coding agents as it investigates whether smaller teams can handle a greater volume of software development work. The initiative comes as businesses across the technology sector increasingly explore generative and agentic AI for programming, testing, automation and other parts of the development lifecycle.
Rather than viewing AI solely as a tool for generating individual pieces of code, the experiment points toward a potentially broader change in engineering productivity. If AI systems can reliably take over repetitive development tasks, engineers could spend more of their time on architecture, product decisions, complex debugging and other work requiring greater human judgment.
Why Smaller Development Teams Matter
Software development has traditionally required teams covering multiple responsibilities, from writing code and reviewing changes to testing, documentation and deployment.
AI coding tools are beginning to challenge that model.
An AI agent capable of completing portions of these activities could potentially increase the amount of work that an individual engineer or a small group can manage. That does not necessarily mean companies will eliminate engineering teams. Instead, businesses may reconsider how those teams are structured and where human expertise delivers the greatest value.
The wider industry is already debating this possibility. Recent evidence from the software sector suggests that lower development costs could actually expand the number of people building technology rather than simply reducing developer demand. Atlassian, for example, has argued that AI is making software creation accessible to a wider range of workers and increasing overall participation in development.
Salesforce's Broader AI Strategy
The development experiment fits within Salesforce's larger push to incorporate artificial intelligence throughout business workflows.
Salesforce already promotes AI across areas such as customer relationship management, predictive analytics, automation and AI-powered agents. Its AI strategy increasingly focuses on helping businesses automate work while using company data and existing enterprise systems as context.
Applying similar technology internally to software engineering therefore represents another test of how far AI-driven productivity can extend.
For smaller companies and development organizations, the implications could be significant. Hiring large engineering teams is expensive, and limited technical resources can slow the development of internal applications and new products. If AI meaningfully increases each developer's capacity, organizations could potentially undertake projects that previously required larger budgets or teams.
AI Coding Still Brings Risks
Higher development speed, however, does not automatically produce better software.
AI-generated code can contain security weaknesses, incorrect assumptions and difficult-to-detect errors. Recent research into AI-generated software has highlighted that models may produce code that falls short of security or compliance requirements unless appropriate instructions, review processes and safeguards are applied.
That makes human oversight particularly important for enterprise applications dealing with customer information, authentication, payments or other sensitive systems.
Another challenge involves maintenance. AI could make creating new internal applications dramatically easier, but every application may eventually require security updates, monitoring, integrations and ongoing support.
Businesses therefore need to consider the complete lifecycle of AI-generated software rather than focusing exclusively on how quickly the initial code can be produced.
Could AI Change the Economics of Software?
Salesforce's experiment addresses a larger question facing the technology industry: if AI substantially lowers the cost of creating software, will companies simply operate with fewer developers, or will they use the additional capacity to build considerably more technology?
The second possibility could transform the market.
Smaller teams might create specialized internal tools that previously would not have justified a dedicated development project. Departments outside traditional engineering organizations could also participate more directly in application development as natural-language interfaces lower technical barriers.
At the same time, commercially available software will continue to compete with internally developed applications. Building an application more cheaply does not eliminate expenses associated with maintenance, security, reliability and technical talent. Recent analysis of the build-versus-buy question suggests those longer-term costs can still make purchased software more economical for many important business systems.
Balanced Analysis
Salesforce's test should therefore be viewed as an experiment in productivity rather than proof that traditional software teams are becoming unnecessary.
If coding agents mature, engineering organizations may be able to accomplish more with fewer people assigned to individual projects. That could accelerate product development and make custom software accessible to organizations with limited technical resources.
But greater productivity also creates new responsibilities. Companies will need strong testing, security controls, code review and governance to ensure that increased development speed does not generate a corresponding increase in technical debt.
The important question may ultimately be less about whether AI replaces software developers and more about whether it changes the amount of software each developer—and each organization—is capable of creating.
This article is based on reporting published by Financial Express.
