AI has rapidly become a major investment priority for financial services. With the potential to improve how institutions operate and develop new services, it is easy to see why leaders are keen to move quickly. Freshworks research involving more than 12,000 IT decision-makers globally found that 90% of financial services organisations are actively using AI. Furthermore, nearly half plan to increase investment significantly over the next 12 to 24 months, the highest rate of any industry surveyed.
Finance is also more likely than any other sector to develop AI in-house, with some 43% stating building as their primary approach, compared with 34% across the full survey. And as the cost and effort involved in creating software features continue to fall, there is no better time to take a deeper look at what should be built, what can be standardised and where long-term advantage will come from.
Why the decision to build matters
A financial institution cannot approach AI in the same way as every other organisation. The sector is responsible for other people’s money, holds scores of sensitive data and operates within demanding regulatory requirements. Naturally, leaders want confidence that any AI capability reflects the controls and risk appetite of their business.
Building in-house can provide that confidence when the capability is closely tied to the way an institution operates. It gives the organisation greater influence over how the technology is designed and how it develops over time. Yet, there is often a false assumption that an internal build is automatically the safest or most controlled option. If teams create separate capabilities without a common approach, the organisation can end up with a growing collection of systems that are difficult to connect, oversee or maintain. This is a key reason why being able to build more does not always mean a business should.
Building AI the right way
There are some defining questions financial leaders should consider before committing to an internal build. The first is whether custom development will create something meaningfully different for the business, or simply reproduce a capability that is already widely available.
Technology that reflects a distinctive business model or a highly specific regulatory need may well warrant investment. Elsewhere, choosing a standardised capability can free internal teams to concentrate on the areas where their expertise adds more value.
The second question is what the new capability will add to the wider technology estate. Organisations lose an average of 25% of their AI budgets to complexity overhead, according to the research. Within financial services, 40% describe AI-related complexity as very significant, which is 11 percentage points above the overall average. These figures show why the cost of development cannot be judged by the build itself. Leaders also need to account for the work involved in connecting, governing and supporting it once it is in use.
Keeping complexity under control
Mass, non-standardised development can make each new project harder to absorb. Different teams may make different decisions about how AI is introduced, leaving employees to work around the gaps between systems. To avoid this, companies need a clear view of the AI capabilities already being used across the business. Responsibility for each build should be understood, while documentation and oversight need to remain consistent enough for leaders to see how the overall estate is developing.
This does not remove the flexibility that makes internal development attractive. In fact, a common approach to governance can make it easier for teams to test new ideas because the basic expectations are already understood. At the same time, organisations also need to understand that a successful build is one that people can use confidently as part of their work, without creating substantial rework elsewhere. If employees need to spend more time checking outputs or reconciling information between systems, the organisation may have introduced technical capability without improving the way the business operates.
The growing value of trusted distribution
AI has dramatically lowered the effort required to create new software features. As a result, features that once helped one provider stand apart can be developed more quickly by competitors. This changes where long-term advantage sits. For financial organisations, trusted distribution and regulatory maturity are becoming more valuable, because they determine whether a new capability can reach customers and operate responsibly at scale.
Internal development should strengthen those advantages rather than distract from them. An organisation that devotes too much time to recreating common technology may have less capacity to improve the services, relationships and specialist processes that customers recognise. Therefore, the build-versus-buy decision should not be treated as a verdict on technical ambition. It is a choice about where the institution wants to place its time, skills and investment.
Pressure should not dictate architecture
The pressure to show returns can make these decisions more difficult. Across the research, 71% of executives expect AI ROI within eight months, even though more than half of deployments take six to 12 months simply to go live.
This mismatch can encourage companies to focus on a visible launch rather than the less visible work required to make the technology sustainable. A quick succession of internal pilots may demonstrate momentum, but it can also leave the business with an AI estate that becomes more expensive to manage as it grows.
Leaders should therefore consider the likely cost of a build over its full life, rather than concentrating only on the speed of its initial release. If the organisation knows how a capability will be governed, supported and connected from the outset, it is more likely to generate value beyond the pilot stage.
Looking ahead to the future
The future looks bright for financial services institutions investing in AI, especially those able to combine innovation with sound judgement about where customisation is genuinely required. Building in-house will remain the right choice in many cases. At the same time, standardising technology that does not create a distinctive advantage will help prevent investment from being consumed by avoidable complexity.
For IT leaders in financial services, the direction is clear. Build where control or differentiation matters, and avoid treating every internal requirement as a reason to start again. When that balance is right, AI becomes easier to scale and the institution can focus on the advantages that are much harder for competitors to copy.
