AI agents are already beginning to do more than answer questions, with emerging tools helping users book travel, make purchases and automate everyday tasks. But as these systems become more capable, the next challenge is ensuring they can work together reliably, securely and with clear accountability.
For Enzo Amur, creator of the Global Agentic Network and Mockingjay Network at Soika Labs, the opportunity lies in moving beyond individual AI assistants towards a connected environment where specialised agents can coordinate tasks and collaborate across organisational boundaries.
“We believe the next phase of AI is about moving from assistance to execution,” Amur said.
“The real opportunity is not another interface where you ask AI a question. It is building an environment where AI can help coordinate and complete real work.”
From individual assistants to collaborating AI agents
The first wave of generative AI largely focused on helping people write, research, analyse information and make decisions faster. Agentic AI takes that further by enabling systems to act across workflows, coordinate different tasks and work towards outcomes within defined permissions.
Amur believes the next major phase of adoption will be driven by the ability of AI agents to complete entire workflows rather than simply assist employees with individual tasks.
“The next major ROI phase will come when AI moves from copilots that help employees work to agents that can actually complete workflows,” he said.
“The real value will not simply be measured by how much faster an employee can complete an individual task, but by how effectively AI agents can support or complete entire workflows, reduce operational friction and allow people to focus on areas where human judgement and expertise matter most.”
Enzo Amur
For consumers, multiple AI agents could eventually coordinate tasks across travel, shopping, customer service and personal finance. Rather than relying on one assistant for every step, users could delegate different parts of a task to specialised agents.
Amur uses travel planning as an example of how this could work.
“Instead of asking an assistant to recommend flights and then manually booking everything yourself, you could have one agent understand your preferences and budget, another evaluate flights, another identify suitable accommodation and another coordinate the itinerary,” he said.
“The important development is not simply that these agents can find information, but that they can potentially work together to move the task towards completion within permissions set by the user.”
He believes the same principle could apply across other everyday services, with agents understanding what a person needs, comparing options, interacting with specialised services and helping execute the next step.
“That is why I find agentic AI so exciting. We are moving from AI generating an answer to AI potentially coordinating an outcome,” Amur said.
Why security and accountability matter
As AI agents gain greater access to business systems and information, their ability to take action creates new questions around privacy, security and accountability.
For Amur, these concerns are central to the future of agentic AI, particularly as businesses consider allowing AI systems to handle increasingly complex operations.
“With generative AI, we worried primarily about what the model might say. With agentic AI, we increasingly have to worry about what the model might do,” he said.
Businesses need to know what an agent is authorised to access, which actions it can take, where data is being processed and what happens when something goes wrong.
“My view is very simple: AI can execute a decision, but accountability cannot be outsourced to an algorithm,” Amur said.
He believes human oversight, permissions, traceability and clear governance must develop alongside the technology. Sovereignty is also important, particularly for governments and enterprises handling sensitive information.
Organisations, he said, need greater clarity and control over their data and AI infrastructure rather than handing critical processes to systems they cannot adequately govern.
“The companies that solve agentic AI responsibly will not just be the ones that build the most capable agents. They will be the ones that make those capabilities trustworthy enough to deploy in the real world,” he said.
The UAE’s opportunity in the next phase of AI
The UAE is well positioned to benefit from the next stage of AI adoption, according to Amur, pointing to the country’s investment in AI infrastructure and willingness to deploy emerging technologies.
“Dubai gives us something extremely valuable as an AI company: an environment where there is both ambition and a willingness to deploy technology,” he said.
However, he believes that building a competitive AI economy requires more than infrastructure and investment. The next challenge is translating these foundations into scalable companies, intellectual property, commercially viable products and measurable productivity gains.
“Compute is essential, but compute alone does not create an AI economy,” Amur said.
For companies such as Soika Labs, the opportunity is to build from the UAE while addressing challenges that extend beyond its borders.
“If we can demonstrate that AI developed here can operate securely, deliver measurable business value and scale across markets, that strengthens not only individual companies but the wider technology ecosystem,” he said.
How AI agents could reshape the workplace
As AI agents become more capable, their impact on employment may initially be felt in how jobs are structured rather than in entire roles disappearing, Amur said.
Many jobs involve a combination of repetitive processes, coordination, decision-making and human judgement. AI agents could increasingly take on predictable and time-consuming tasks, allowing employees to focus on problem-solving, relationships and strategic decisions.
Functions involving the movement of information between systems, routine requests, report preparation, administrative processes and established workflows are likely to be among the first to change.
However, adopting agentic AI requires more than introducing another software tool.
“Running a successful AI pilot is relatively easy. Integrating AI into a company’s real operating environment securely, reliably and economically is much harder,” Amur said.
Businesses need to rethink their workflows, identify which responsibilities can be delegated to AI agents and establish where human oversight remains essential.
“I see this as a transition towards a different way of working, where people increasingly manage and collaborate with AI agents rather than manually carrying out every step of a process,” he said.
Building a network for AI collaboration
At Soika Labs, the focus is on developing Mockingjay Network as part of a broader vision for AI agents to collaborate within businesses and, over time, across organisational boundaries.
The company describes Mockingjay Network as a platform for publishing and hosting AI agents. Its Mockingjay Network App is designed as a corporate-level agentic operating system for internal business automation and smart business governance.
The broader ambition is to enable businesses to connect specialised agents and coordinate workflows while maintaining the security, permissions and accountability required for collaboration between organisations.
“The gap we saw was the problem between companies to cooperate safely using AI agents, so now we handled that challenge with Mockingjay full fledged solution,” Amur said.
The company’s immediate focus is on enterprise customers, where it sees clear opportunities to demonstrate measurable value through automation and workflow execution. Consumer-facing applications are also part of its longer-term vision as people become increasingly comfortable delegating complex tasks to AI.
“Our immediate focus is on businesses because that is where we see some of the clearest opportunities to demonstrate measurable value from agentic AI,” he said.
For Amur, whether the user is a business or an individual, the underlying principle remains the same: AI becomes more valuable when it can move beyond recommendations and help complete tasks safely.
That shift will depend not only on the capabilities of individual agents, but also on whether they can work together within systems that businesses and consumers can trust.
As agentic AI develops, the challenge will be to make collaboration between these systems as reliable, secure and accountable as the tasks they are designed to perform.
