Augusto Martinez Reyes, COO EMEA and President of Multilingual Hubs at TP Group, discusses human oversight, Arabic localisation and the coordination of AI agents, enterprise data, and people to improve customer experiences
AI agents are reshaping customer operations, bringing new demands for oversight, accountability and service design. Across the Middle East, successful deployment also depends on understanding Arabic dialects, cultural nuances and customer expectations, while connecting automation with enterprise data and human expertise.
Augusto Martinez Reyes, COO EMEA and President of Multilingual Hubs at TP Group, spoke to TahawulTech about managing AI agents, preparing employees to work alongside them and preserving trust as customer interactions become increasingly autonomous. Reyes explains why businesses must rethink the entire customer journey, with clear safeguards and human judgement guiding decisions that affect people.
Interview excerpts
How should organisations manage AI agents, measure their performance and maintain effective human oversight?
Effective oversight starts before an AI agent is deployed. Organisations need to establish what it is authorised to do, which decisions require human approval and who is accountable for its actions. Data protection is a fundamental part of those guardrails. Protecting employee and client data has always been fundamental to our business. AI does not change that responsibility. Organisations need to understand how data is managed and used, the regulations across their markets, and the boundaries around identity management, risk and compliance. There also needs to be a balance between regulation and innovation. We should protect people and their privacy while giving organisations room to test technology that can solve meaningful problems. On performance, I would look beyond cost reduction to whether the process works better for the people using it. In customer service, for example, an AI agent might resolve a routine enquiry more quickly, but speed alone does not tell us whether it has done a good job. We should also measure whether the answer was accurate, whether the issue was resolved and whether the customer had to contact us again. Where human support is needed, the handover should be straightforward.
Customer and employee feedback should help organisations identify what needs to change, rather than assuming that more automation necessarily means a better outcome.
What challenges do Arabic dialects, cultural context and customer behaviour present when deploying localised AI across the Middle East?
Arabic dialects vary across the Middle East, so deploying localised AI requires more than translating a response into Arabic. The system needs to understand local expressions, tone and the context behind what a customer is saying. We face similar challenges with Spanish across different countries, or with Portuguese in Brazil and Portugal. Understanding those differences is essential if the service is going to feel natural to the person using it. Cultural context and customer expectations also shape the interaction. Recognising the words is only part of the task. The AI needs to respond appropriately to how someone expresses a concern, asks for help or signals frustration. These distinctions are not always well understood outside the Arabic-speaking world, and I believe they need more attention as organisations develop services for the region. At our Egypt hub, interpreters from LanguageLine Solutions, a TP company handling tens of millions of interpretation interactions a year across more than 270 languages, help us fine-tune AI responses for different Arabic dialects. Their contribution goes beyond translating words. They bring an understanding of how language is used in context and what the speaker intends to communicate.
The harder part is enabling AI to recognise and adapt to a person’s dialect and communication style through the conversation itself, without needing to ask where they are from. That is the experience we are working towards, and I am encouraged by the progress I have seen.
How will customer AI agents interacting directly with corporate AI agents transform customer service, transactions and brand relationships?
When customers can ask their own AI agents to compare offers, make purchases or resolve service issues, businesses will have to earn their preference in a different way. A customer could ask an agent to find a service that meets a particular budget, check the conditions and arrange the purchase within agreed limits. That could remove much of the effort people currently spend moving between websites, comparing terms and contacting support. For brands, this raises an interesting question, how do you maintain a relationship with a customer when their agent handles much of the interaction? If the agent is comparing providers against the customer’s priorities, businesses may face greater pressure to demonstrate why their offer is worth choosing. Clear pricing, reliable service and keeping promises could become even more influential in winning repeat business. Trust and identity will be central. A business needs to establish that an agent is authorised to act for a customer and understand the limits of that authority. Permission to compare prices, for example, should not automatically mean permission to make a purchase or accept a new contract. Customers also need visibility over what their agents have agreed to and a straightforward way to intervene. The opportunity is to make routine transactions almost effortless. But when something goes wrong, the customer will still expect the brand to take responsibility. The companies that handle those moments well could build stronger relationships, even if customers speak to them less often.
Why will coordinating AI agents, enterprise data and human teams become a critical competitive advantage?
Orchestration is critical because automating individual tasks does not automatically make the overall operation more productive. When you take pieces of work away from a human, you can create gaps and inefficiencies unless you redesign how the whole journey fits together. Consider a customer calling about a refund who also has two other issues. You cannot keep moving that person backwards and forwards between AI and a human. Once the interaction reaches a person, that person needs to be able to deal with the complete requirement. The advantage comes from knowing where AI adds value, where humans belong and how to coordinate both productively. But coordination on its own is not enough. AI agents have to be integrated into the workflows that already run the business – the systems, data and handoffs behind each interaction- and in many cases that means redesigning the journey itself rather than layering automation on top of it.
Organisations that treat AI as a reason to rethink how work flows end to end, not as a tool bolted onto existing processes, are the ones that will turn it into a durable advantage.
Where should businesses draw the line between AI-assisted decisions and fully autonomous decision-making?
Pilots and testing are essential to understanding what the technology can and cannot do. But the limits are ultimately set by humans, including the customers and employees expected to use it. Think about navigation apps. They help us reach a destination more efficiently without constantly criticising every wrong turn. If an application continually told people how badly they were performing, they might stop using it. The same consideration applies to AI in the workplace. In our recruitment process, AI supports assessments and automates documentation, while people remain involved in onboarding and building a connection with candidates. That illustrates the balance: AI can support the process, but decisions about a person’s suitability for a role should remain subject to human judgement. Candidates should also be able to speak to someone who can understand their circumstances and answer their questions.
The aim is to improve the hiring experience while retaining accountability for decisions that affect people’s careers.
How could AI reshape the contact centre and the wider architecture of customer operations?
AI is accelerating change in customer experience, not starting it. Unlike earlier technologies that moved existing processes into different channels, it requires us to rethink the journey itself. Humans bring context from their experience. AI needs that context supplied, alongside clear boundaries on what it should and should not do. Data management, protection and identity safeguards become increasingly important as threats evolve. We also need to identify the moments when customers require emotional intelligence and prepare people for new roles in AI deployment and quality. This is not hypothetical for us, TP employees completed over 56 million hours of training in a single year, and an increasing share of that is preparing people to work alongside AI rather than be displaced by it. I am optimistic, provided we combine the technology with safeguards, skills and sound human judgement across the entire customer journey. Otherwise, we have a very fast car without knowing how to drive it.
Source: Tahawul Tech
