While artificial intelligence is today a key accelerator for enterprise transformation, it is also at the same time exposing the complexities built up inside technology environments over time.
How can we better manage the ensuing complexities that comes from AI?
In an interview with iTNews Asia, Alvaro Garrido, COO, Technology and Operations & Chief Information Officer for Information Security and Data of Standard Chartered sharing best practice tips and learnings from the bank’s current and past AI initiatives, said we first need to be aware that AI will always bring fragmented processes, duplicated data, hidden dependencies and slow execution into sharper focus.
To help resolve this, Garrido advised that AI and modernisation should not be treated as competing priorities. Organisations must address both in parallel if they want to scale AI effectively. “AI is bringing a lot of this operational complexity that was embedded to light,” he said.
He added further that to be able to integrate AI into the organisation effectively, organisations must first address existing fragmented processes, such as duplicated data, hidden dependencies, or factors causing slower execution.”
AI cannot simply be layered over technical debt
Garrido cautioned that enterprises risk making their technology environments more complicated if they add AI to fragmented systems without addressing the underlying problems. “The last thing you want to do is throw another layer of AI on top of a fragmented and technical debt space,” he said.
He explained that some early AI initiatives were built primarily around experimentation or without a clearly defined business outcome. Others attempted to replicate approaches used elsewhere without sufficiently considering the organisation's existing technology environment.
As organisations move beyond experimentation, Garrido said they need to change how they approach AI investments.
He also warned against simply copying AI strategies used by competitors without considering an organisation's existing technology environment. “You have to be careful on how to retrofit all those recipes that work for others into your existing bank,” he said.
AI can provide greater observability and help identify areas where technical debt could be addressed, particularly in back-office operations and infrastructure, but the technology itself does not eliminate that debt, he explained
Modernisation and innovation must move together
Rather than deciding between modernisation and innovation, Garrido said organisations should identify where technology changes can create the greatest downstream impact.
For a large technology estate, that means using operational data and telemetry to identify bottlenecks, then prioritising foundational workflows and platforms that are consumed across the organisation. “Normally they go hand in hand,” he said.
For example, he said that Standard Chartered is directing some of its AI efforts towards standardising core platforms. “Once those foundations are simplified and technical debt reduced, automation and AI can be applied more effectively,” Garrido said.
This approach, he adds, creates a cycle in which improvements to foundational platforms can increase the impact of AI across processes that depend on them.
Data emerges as the next frontier
Of the major technology layers, Garrido identified data as the area where addressing complexity could deliver the greatest benefit. While infrastructure has historically been viewed as a difficult area to modernise, Garrido said the level of discipline and expertise around infrastructure has developed significantly.
Data for us is the next frontier. Data is complex. Data is everywhere. Data is the alpha and the omega.
- Alvaro Garrido, COO, Technology and Operations & CIO for Information Security and Data, Standard Chartered.
He said the challenge is particularly amplified for financial institutions because data management involves multiple dimensions of risk, as well as jurisdictional legislation, restrictions and regulations.
AI models, regardless of whether organisations use closed models or experiment with open-weight models, ultimately depend on the quality of the underlying data.
“At the end of the day, you're drinking from the data,” Garrido said, stressing the importance of reliable, quality, curated and properly contextualised data.
He added that Standard Chartered is now working on a global data platform designed to support AI requirements, with standard APIs, reusable capabilities and governance intended to allow production, reporting and AI systems to work from integrated data.
Garrido said this approach was chosen by the bank so that when an AI use case reaches the data layer, improvements to the underlying data infrastructure can happen at the same time.
AI could also create a new layer of technology debt
As enterprises introduce copilots and agentic AI, Garrido said the next challenge could come from the decentralised use of AI itself. He compared this with the long-standing problem of end-user computing, where important business logic can sit inside individual spreadsheets rather than controlled enterprise systems.
Agentic AI and copilots could expand that risk if employees begin using AI to support decisions and business processes without adequate controls.
He recommends that organisations develop guardrails, audit trails and visibility into how AI is being used across data, networks, models and consumption. Without those controls, AI could create another layer of unmanaged technology that eventually becomes technical debt.
Complexity has a measurable cost
Technical debt eventually shows up in the day-to-day performance of an organisation.
Garrido said leaders should look beyond the size of their technology estate and examine how much variation exists within it from different operating systems and database versions to multiple versions of the same platform.
The more technology variations an organisation has to maintain, the greater the operational complexity. “Complexity drives two things - cost and downtime,” he said.
That complexity can also slow delivery. Garrido challenges the assumption that giving teams more technology choices automatically make them faster, arguing that standardised components can help teams move more quickly while maintaining the necessary controls.
He compares the approach to building with standard Lego blocks, teams can still create different solutions, but they work from a common set of components.
Standard Chartered has been applying this principle across its technology environment, standardising infrastructure and moving into areas such as data and payments. The bank aims to reduce the number of technology variations while allowing specialised applications to consume common capabilities through APIs.
Garrido said the process has already significantly reduced the bank’s technology footprint over the past five years, although some complexity remains and will require continued effort.
Technical debt is also a security problem
Technical debt can also create cybersecurity risks when ageing technology platforms fall out of support. Garrido noted that once platforms are no longer supported, organisations may no longer receive new security patches, increasing their exposure to potential exploitation.
For banks and other organisations operating critical systems, he said modernising older platforms, reducing the technology footprint and keeping systems up to date are therefore important not only for efficiency, but also for maintaining security.
The future will be measured beyond AI deployment
For Garrido, the transition to an AI-driven organisation will ultimately need to show up in business and workforce outcomes.
“One is client satisfaction and how clients interact with the bank. The other is talent. If I'm able to attract top talent at the same time that I retrain or prepare my workforce for that era, it means that we are an attractive place to work,” he said.
He pointed to employee access to AI tools and training as another indicator of organisational readiness. The broader shift, however, is already changing how technology leaders need to think about AI. Rather than retrofitting existing processes with an AI layer, Garrido said for AI success, organisations ultimately need to consider what the future operating model should look like.
