Artificial intelligence is changing the economics of technology development across capital markets, making it faster and cheaper for firms to build their own applications. But while AI is lowering the barriers to software development, it is also raising questions about long-term ownership costs, the pace of vendor innovation and where firms should focus their technology investments.
These issues were explored during The New Buy/Build Calculus, a panel discussion at A-Team Group’s recent Buy AND Build event in London.
Audience polling suggested that AI had shifted the balance towards building, with prototyping and coding identified as the areas experiencing the greatest change. But the discussion revealed a more nuanced picture. While AI is making software development easier, the decision to build rather than buy still depends on competitive differentiation, operational requirements and the ability to sustain investment over time.The Changing Economics of Building
AI-assisted development is allowing business users to create prototypes and applications without relying entirely on traditional technology teams, reducing development bottlenecks and accelerating experimentation. However, generating working code is only one part of the process. Turning prototypes into production-ready applications requires engineering expertise, security testing, integration and ongoing maintenance. These costs can outweigh the savings achieved during initial development.The accelerating pace of vendor innovation adds another consideration. A firm investing significant resources in developing a proprietary application could find that an existing provider delivers equivalent functionality before the internal project is completed. Even successful internal builds require continued investment to remain competitive. Vendors can spread development costs across multiple clients while continually enhancing their products.
This reinforces the distinction between technology that delivers genuine competitive advantage and functionality that primarily supports operational efficiency. The former may justify proprietary development, while the latter can often be sourced more economically from established providers.
As we explored in our recent coverage of the changing economics of equities trading infrastructure, firms are reassessing the cost and flexibility of their technology architectures as market requirements evolve.
Vendors Face a Higher Bar
As software development becomes easier, established functionality becomes less defensible as a source of differentiation. Vendors must increasingly demonstrate value through domain expertise, architectural flexibility and operational reliability, including their understanding of trading workflows and regulatory requirements.
Rather than purchasing complete applications with limited scope for customisation, firms are looking for dependable technological foundations that support internally developed capabilities. Core functions such as market connectivity, routing and transaction processing can be sourced externally, while proprietary workflows, analytics and AI-enabled tools are developed around them.
This requires vendors to provide flexible APIs, accessible documentation and architectures capable of accommodating successive generations of AI models. Providers must support clients’ evolving technology strategies rather than locking them into proprietary AI capabilities that may become outdated or conflict with enterprise-wide standards.
Governance and Control
As AI-assisted development expands, governance is becoming an increasingly important consideration in buy/build decisions.
Responsibility for trading and investment decisions remains with individuals and their firms, regardless of whether AI contributes to the decision-making process. The introduction of autonomous agents raises additional questions around permissioning, accountability and auditability.
Technology architectures must ensure that AI applications operate within established security and data governance frameworks, controlling access to information, restricting permissible actions and maintaining records of activity.
The discussion highlighted growing interest in centralised AI model management. Rather than allowing individual vendors to embed separate models within their applications, firms are exploring enterprise-wide model routing and usage monitoring. This creates additional requirements for vendors, including model agnosticism, integration with existing governance systems and detailed reporting of token consumption and associated costs.
Regulatory compliance and third-party risk management are also becoming more demanding. Requirements under the EU’s Digital Operational Resilience Act (DORA), for example, are increasing scrutiny of technology providers’ operational resilience and contractual arrangements.
The panel also questioned whether AI should be used for every automation task. Where a process can be executed through conventional rules-based software, using AI continuously may introduce unnecessary cost and complexity. AI can help develop such processes without necessarily remaining involved in their execution.
Who Owns the Decision?
As business users gain greater development capabilities, they can take more direct responsibility for addressing operational problems. However, this doesn’t remove the need for centralised technology oversight.
The discussion drew a distinction between ownership of business requirements and responsibility for technology decisions. Business teams define the problems and desired outcomes, while technology functions assess architectural compatibility, security, vendor risk and long-term sustainability.
The proliferation of AI-generated applications may increase the importance of technology governance, particularly where independently developed solutions create duplication, security vulnerabilities or maintenance obligations.
Finding the Right Balance
The economics of buying and building vary across asset classes and business functions. In equities, where firms may have greater scope to differentiate through proprietary trading technology, internal development can offer significant advantages. In fixed income, established electronic trading platforms and their continuing investment in functionality can strengthen the case for buying.
The focus is also shifting towards demonstrating measurable returns from AI investment. Faster development alone doesn’t establish a compelling business case, particularly when experimentation consumes resources and production deployment remains challenging.
The emerging model combines externally sourced infrastructure with internally developed capabilities, allowing firms to concentrate resources where proprietary technology delivers meaningful differentiation.
AI is certainly changing the buy/build calculus, but the fundamental considerations remain: which capabilities firms need to own, which they can source more effectively from vendors and how to maintain flexibility as technology evolves.
Subscribe to our newsletter
