As artificial intelligence moves from experimentation to enterprise-wide adoption, many organisations are discovering that automation, while powerful, is far from a silver bullet. In this interview with TechBullion, David Antony, President and COO of Flatworld Solutions, shares a grounded view from the front lines of global operations. With more than two decades of experience spanning customer-facing and back-end functions, Antony explains why scaling AI often exposes, rather than solves, deep-rooted inefficiencies. He outlines how forward-thinking businesses are shifting from tool-led experimentation to outcome-driven operational design, balancing innovation with governance, and building integrated systems that can deliver both efficiency and resilience at scale.
Please tell us more about yourself.
A: My name is David Antony, and I am the President and COO at Flatworld Solutions. We have over two decades of experience in helping businesses grow their success by optimizing their customer-facing and back-end operations.
Flatworld Solutions is a company that specializes in different domains in different industries, such as call centers, which are customer-facing operations, and back-end operations such as healthcare, finance, software, mortgage, and so on, and we serve over 100 countries worldwide.
Our focus has changed and is now at the intersection of business process outsourcing, technology, AI, and automation to drive efficiently and scalability
Automation promises simplicity, yet many firms face new operational complexity. Why does scaling AI often create fragmented workflows and integration challenges across enterprise systems?
A: When automation is layered over an already broken process, it usually makes the problem worse rather than better. Scaling AI and automation do not cause fragmentation; it reveals the inefficiencies that were already in the system.
For example, the marketing team might be working with a chatbot to better engage with customers, the sales team might be working with an AI-based platform for outreach and automation, and the finance team might be working with tools for invoice data extraction. Individually, each initiative delivers value, but they are often pursued in isolation.
When any organization decides to deploy automation in silos, each of these tools will resolve only the local problems but will result in fragment workflows, duplication, inconsistent data and lack of a single source of truth. Integration becomes more complex and challenging when more systems are layered on. To scale effectively, organizations need a holistic, enterprise-wide approach rather than isolated implementations.
From your perspective at Flatworld Solutions, why are some automation initiatives adding layers of operational oversight rather than simplifying business processes?
A: The problem usually begins with impractical expectations about automation. Many businesses believe that when AI and automation is implemented, it will work instantly, correctly and at scale. However, the truth is it does not work that way.
By their nature, AI systems need to contend with variables like data quality issues, model drift, exceptions, and even changes in regulation. Even well-tested systems can encounter edge cases once deployed in real-world environments. As a result, organizations often find themselves adding layers of monitoring, quality control, and manual oversight after implementation.
This is why automation initiatives can result in increased complexity rather than reducing it because they were designed with the assumption of complete automation from day one. A better approach, however, would be to plan for these realities from up front.
Instead of planning for complete automation, organizations should design workflows that include human validation, audit trails, and clear checkpoints in the process from the beginning. If automation is carried out with this approach in mind, it could be more reliable.
How can organisations design automation strategies that reduce operational complexity rather than introduce disconnected tools and fragmented decision-making frameworks?
A: When developing an automation strategy, it is not recommended that the focus be on tools but on business outcomes and process architecture. It is imperative that organizations evaluate its state of workflow and focus on simplifying, standardizing, and optimizing the process. Automating an inefficient process only scales inefficiency.
Another critical aspect is that an organization should not focus on individual tools, but rather a unified workflow. It is easy for an organization to get caught up in the idea of deploying different solutions for different functions, which will eventually lead to disconnected systems. Instead, an automated solution should be designed as an integrated, enterprise-wide capability with seamless coordination across functions.
Clarity in decision-making is also important. Organizations need to be clear at the onset about which decisions should be made by AI, where human intervention is necessary, and how exceptions will be managed.
At the same time, an effective data and integration platform is also necessary to ensure that all systems are operating on an effective platform and that fragmented data is not created. Finally, when organizations adopt this approach, then automation is not simply technology; it is an operating capability that enables efficiency, scalability, and value.
As automation spreads across finance and operational functions, what governance structures are needed to ensure transparency, accountability and risk control?
A: As organisations implement AI and automation, these technologies are no longer in silos; they cut across functions. In this context, management and compliance cannot be approached in a narrow or fragmented way. This makes governance and accountability critical, as their impact extends far beyond a single process or team.
For a governance framework to be effective, it should operate on various levels. For example, an effective governance framework should include model governance, which covers the monitoring of AI systems to address performance issues such as model drift It also requires process governance, where workflows are regularly evaluated and refined, and data governance to control access, maintain data quality, and mitigate risk.
At the same time, the risk landscape is expanding. With AI, exposure is higher – not just in terms of data, but also in how systems can be exploited. AI-driven attacks and advanced social engineering techniques are becoming more sophisticated, making it essential for organizations to proactively manage security and access controls.
It is also noteworthy that accountability cannot be left to AI alone and that there is still an essential role to be played by humans in terms of quality, compliance, and decision-making processes. Organizations need to ensure that AI-driven processes are transparent, auditable, and explainable.
Ultimately, effective governance is about in financial services combines strong oversight with clear accountability, ensuring that AI operates as a controlled, reliable, and trustworthy part of the enterprise rather than an unmanaged black box.
Many enterprises struggle to integrate AI tools with legacy infrastructure. What practical steps can leaders take to manage integration complexity while maintaining operational efficiency?
A: Legacy systems have been developed over a long period of time, with businesses still using them due to their reliability. In many cases, the mindset remains: if it isn’t broken, don’t fix it. The issue, however, is how to incorporate new AI and automation technologies with legacy systems without adding to the complexity of the system.
A practical starting point is to focus on a few high-value workflows rather than attempting large-scale transformation all at once. By prioritizing targeted use cases, organizations can implement, test, and refine automation in a controlled way before scaling it across other areas.
Another critical consideration is how AI interacts with legacy systems. Often, these systems lack modern API access, which leads organisations to embed automation directly into specific parts of the application. While this may work in the short term, it can create significant maintenance challenges over time, as multiple tightly coupled components become difficult to manage and update.
A more sustainable approach is to either leverage available APIs where possible or introduce a middleware or data layer that acts as a bridge between legacy systems and new AI capabilities. This creates a more flexible architecture, making it easier to scale, maintain, and evolve automation initiatives.
It’s also important to recognise that integrating AI into legacy environments is not a one-step exercise but a phased journey. Defining clear business outcomes for each phase while continuously monitoring progress helps organisations stay on track, identify issues early, and avoid committing excessive time and resources to large, uncertain implementations.
How is Flatworld Solutions helping clients rethink operational design so that automation supports scalable, sustainable enterprise workflows?
A: One of our key advantages as an organisation is our deep experience across both business processes and technology. We’ve spent years executing and managing processes for clients across multiple industries, which gives us a strong understanding not just of the technology landscape, but of how real-world operations function on the ground.
This fundamentally shapes our approach. We don’t start with the intent of deploying technology for its own sake. The objective is not to “implement AI” or “increase automation.” If that becomes the goal, it’s difficult to measure success in any meaningful way.
You may end up with AI-driven operations on paper, but without any tangible impact on business outcomes. Instead, we begin with the business outcome. We take a step back and evaluate the client’s existing workflows end-to-end.
Importantly, we recognise that not every part of a process should be automated. In some cases, certain steps are better handled by humans especially when they are highly critical, exception-driven, or relatively small in scope. Automating such tasks may introduce inefficiencies if additional oversight is required. The key is to be deliberate and selective about where automation adds value.
In your experience, what operational mistakes cause automation programmes to become overly complex or difficult to manage at scale?
A: One of the most common mistakes is trying to automate a process without fully understanding it. In such cases, organisations tend to automate the process “as is.” The problem with this approach is that the underlying process itself may be inefficient, fragmented, or poorly designed.
In some cases, there may not even be clear documentation, meaning important nuances and dependencies are missed. When you automate under these conditions, you don’t solve the problem – you amplify it.
Another challenge is adopting multiple tools without a unified underlying framework. This creates disconnected systems and adds to operational complexity rather than reducing it.
Change management is another area that is often underestimated.
Operational teams are used to working in a certain way, and simply introducing a new system does not guarantee adoption. Organisations need to actively plan for how people will transition to new ways of working.
For instance, long before the AI era even in basic digitisation efforts teams continued using old methods out of habit such as completing tasks manually on paper and then filling it in the system, ultimately reducing efficiency instead of improving it.
Finally, many programmes are designed for pilot success rather than enterprise-scale deployment. Pilots are often executed under controlled, ideal conditions, which may not reflect real-world complexity. As a result, what works well in a pilot can break down when scaled across the organisation.
As tech firms expand automation in compliance, payments and analytics, how should leaders balance innovation with the need for stronger operational oversight?
A: The objective is not to slow down innovation, but to ensure that it is implemented in a way that is both trustworthy and sustainable. A key part of this is clearly separating experimentation from production.
Organisations need dedicated environments and teams to continuously explore new tools, technologies, and ideas, while being equally disciplined about when and how these innovations are moved into live operations. This transition should only happen after thorough testing, validation, and refinement to ensure reliability in real-world conditions.
At the same time, many AI tools especially advanced models operate as highly sophisticated, efficient “black boxes,” where the internal logic is not always transparent. This makes governance critical, particularly in regulated industries.
For example, in sectors like mortgage or insurance, it’s not enough to arrive at a decision such as declining a loan application; organisations must be able to clearly explain the reasoning behind that decision for each individual case. Explainability, therefore, becomes non-negotiable not just for compliance, but for accountability and trust. Additionally, the cost of errors in such environments can be extremely high, making strong oversight essential.
This requires a robust governance framework that includes auditability, traceability, continuous testing, clear approval mechanisms, human-in-the-loop controls where necessary, and defined escalation paths. Combined with effective risk management, these elements ensure that innovation can scale responsibly without compromising operational integrity.
Looking ahead, how will businesses evolve their operating models to manage increasingly complex AI-driven processes while still delivering efficiency and strategic value?
A: The future would be about intelligent operations, as opposed to task-oriented or outcome-oriented automation. The focus would be on orchestrating workflows where decisions are made regarding which processes are better handled by humans, AI, or software systems.
Adaptability will be another key factor in the future. For example, Amazon had to continuously change and improve its original business model to remain at the forefront, whereas traditional competitors such as Barnes and Noble had to rethink their approach entirely.
Similarly, BlackBerry was once at the forefront in mobile innovation, but now it is nowhere to be seen in the market due to its failure to adapt to changing market dynamics. Ultimately, the most successful organizations will not be those that automate the most, but those that operationalize AI effectively while remaining agile and responsive to change.
