Reported by Simon Daniel Yusuph l Journalist at Weng Global
Business and technology leaders have called on organisations to adopt agentic artificial intelligence responsibly, with clear governance, defined decision-making boundaries and sustained human oversight as AI systems increasingly move from assisting employees to executing tasks and making decisions within approved parameters.
The call was made during a virtual webinar organised by the EY Nigeria Alumni Association as part of EY Global Alumni Week. The session, themed “Leading with Confidence in the Age of Agentic AI,” examined how increasingly autonomous AI systems could affect business operations, decision-making, customer service and the workforce.
The webinar featured Wilfred Mamah, Partner and Digital Transformation Services Leader at EY West Africa; Tomiwa Adefokun, Senior Manager, Technology Consulting at EY; and Ashish Bakhshi, Partner and Head of Markets at EY West Africa. The discussion was moderated by Bisi Onasanya, former Group Managing Director of FirstBank.
Agentic AI Moves Beyond Conventional Automation
Mamah described agentic AI as a significant change in how organisations interact with software.
Rather than simply generating content or responding to instructions, agentic systems can be designed to pursue defined goals, plan tasks, use digital tools and execute approved actions.
According to Mamah, the defining change is the transfer of specific tasks, decisions and actions from people to software.
He argued that organisations should therefore approach agentic AI as a transfer of clearly defined decision rights rather than simply viewing the technology as a more advanced chatbot.
The distinction is important because conventional AI applications generally provide information or assistance to users, while agentic systems can operate across multiple steps of a workflow and, depending on their design, take actions with varying degrees of human involvement.
EY’s own research and technology materials similarly describe agentic AI as a way of combining human insight with intelligent agents while maintaining trust, accountability and governance. (EY)
Banking, Insurance and Public Services Among Potential Uses
Mamah identified banking, insurance and public services as areas where agentic AI could have practical applications.
Potential uses discussed during the webinar included credit and service orchestration, claims triage and settlement support, and citizen-service fulfilment.
AI agents could, for example, gather information, coordinate approvals, identify exceptions and communicate subsequent steps within a defined workflow.
Such capabilities could change how organisations distribute work.
Mamah said the impact of agentic AI could extend beyond individual tasks to the structure of organisations themselves, with work potentially moving away from departmental silos towards business outcomes.
Routine execution could increasingly give way to exception management, while decision-making could become more continuous and responsive.
The implications would extend beyond technology departments because the introduction of autonomous or semi-autonomous systems could affect employees, managers, customers, regulators and other stakeholders.
Experts Stress Human Oversight
While highlighting the potential benefits of agentic AI, Adefokun stressed that organisations should not give every AI system the same degree of autonomy.
He said the appropriate level of independence should depend on factors including risk, reversibility, confidence and potential business impact.
Under such an approach, an AI agent could have authority to recommend an action, prepare an action for human approval, execute a low-risk action while notifying a human, or escalate an uncertain or high-impact matter for intervention.
This creates a spectrum of autonomy rather than a simple choice between fully manual and fully autonomous operations.
Adefokun said organisations should clearly establish what an AI agent is expected to achieve, which decisions it can make independently, when it must escalate an issue, who remains accountable for its actions and which data and systems it is authorised to access.
He also stressed that the future of AI should involve collaboration between people and AI systems rather than the removal of humans from important decisions.
Governance Must Develop Alongside AI
The discussion placed governance at the centre of responsible agentic AI adoption.
As AI systems gain greater access to organisational data, applications and decision-making processes, the potential consequences of errors or unauthorised actions can increase.
The experts therefore identified several measures organisations should establish before allowing AI agents to operate with significant autonomy.
These include assigning a named business owner to an AI agent, defining its authority and data-access limits, maintaining human oversight over sensitive actions, monitoring its activities and ensuring that its operations can be traced and audited.
Organisations should also have procedures for approving, updating, suspending and eventually retiring AI agents when they are no longer appropriate.
Such controls are intended to ensure that an organisation knows not only what its AI systems can do, but also who is responsible when something goes wrong.
EY’s broader agentic AI framework similarly places emphasis on AI-ready data, human-agent capabilities and confident agent orchestration and governance.
AI Adoption Could Change How Work Is Organised
The webinar also examined the potential effect of agentic AI on organisational structures and employees.
Rather than simply replacing individual tasks, autonomous systems could alter how work is divided between people, departments and software.
Routine processes that currently require employees to carry out multiple administrative steps could increasingly be handled by AI agents, allowing human workers to focus more heavily on exceptions, judgement and activities requiring contextual understanding.
However, that transition would require organisations to rethink job responsibilities, accountability structures and workforce skills.
The change could also create new responsibilities around supervising AI systems, reviewing their decisions, managing exceptions and ensuring that automated processes comply with organisational policies and applicable regulations.
For businesses, the challenge would therefore extend beyond acquiring AI technology. It would involve redesigning processes around the capabilities and limitations of the systems being introduced.
Business Leaders Urged to Focus on Value
Bakhshi challenged organisations to move from discussions about AI’s potential to practical implementation.
He argued that successful adoption should not be measured simply by how many AI agents an organisation deploys.
Instead, organisations should focus on responsibly redesigning valuable work while maintaining the confidence of employees, customers and other stakeholders.
This approach places business objectives ahead of technology deployment.
An organisation could theoretically deploy numerous AI agents without achieving meaningful improvements if the systems are poorly integrated, given inappropriate authority or used for processes that do not justify automation.
The experts’ emphasis on purposeful deployment therefore points to a broader question for businesses: not simply where can AI be used, but where can AI be used responsibly to create measurable value?
Risks Increase as AI Gains More Authority
Greater autonomy also introduces new risks.
The webinar highlighted conerns that can arise when AI agents receive broad permissions or operate across multiple systems.
Potential problems include excessive system access, manipulated instructions, unauthorised actions, leakage of confidential information, compromised third-party integrations, weak audit trails and uncontrolled growth in the number of AI systems operating within an organisation.
These risks make governance particularly important where AI agents interact with sensitive financial, customer, employee or government information.
An agent that is permitted to take action rather than merely provide a recommendation can potentially create consequences more quickly than a conventional software tool.
That makes controls such as permission management, monitoring, escalation mechanisms and auditability important components of responsible deployment.
Human Accountability Remains Central
A central message from the webinar was that increasing AI autonomy should not mean removing human accountability.
Organisations still need people who understand what their AI systems are authorised to do, what data they can access and when human intervention is required.
This is particularly important for decisions involving significant financial, legal, employment, security or customer consequences.
Adefokun’s emphasis on risk, reversibility and business impact provides one framework for determining how much independence an AI agent should receive.
Low-risk and reversible activities may be suitable for greater automation, while decisions involving significant or irreversible consequences may require stronger human oversight.
The distinction also reinforces the need for organisations to understand their own processes before automating them.
Implications for Nigerian Businesses
For Nigerian businesses, the discussion comes as organisations across sectors increasingly explore artificial intelligence for customer service, financial services, operational efficiency and digital transformation.
The potential applications identified during the EY webinar are particularly relevant to industries such as banking, insurance and public services, where organisations routinely process large volumes of information and coordinate multiple stages of service delivery.
However, the same sectors also handle sensitive information and high-impact decisions.
This means AI adoption cannot be separated from questions about data access, cybersecurity, accountability and human oversight.
For businesses operating in Nigeria and across Africa, the challenge will be to capture the efficiency and scale offered by AI while ensuring that technological adoption does not outpace organisational controls.
The approach outlined by the experts suggests that businesses should begin with clearly defined use cases and authority limits rather than granting AI systems unrestricted access to important operations.
From AI Experimentation to Responsible Deployment
The discussion reflects a wider shift in the AI conversation.
Earlier applications of generative AI largely focused on assisting users with tasks such as drafting text, summarising information, generating ideas and analysing material.
Agentic AI expands the concept by allowing systems to plan and execute sequences of actions toward specified objectives.
That additional capability creates new opportunities but also raises more significant questions about responsibility.
As AI systems become more capable of acting on behalf of organisations, businesses will need to determine where human decision-making remains essential and where automated execution is appropriate.
EY’s current materials similarly describe agentic AI as a developing approach that combines intelligent agents with human insight and places governance alongside technological capability.
What Happens Next
The next stage for organisations considering agentic AI is likely to involve moving from experimentation towards controlled deployment.
That means identifying appropriate use cases, defining decision rights, setting access permissions, establishing accountability and continuously monitoring performance.
Organisations will also need mechanisms for updating or shutting down AI agents when their behaviour, performance or business purpose no longer meets organisational requirements.
The EY alumni webinar’s central message was that AI adoption should be driven by purpose and accompanied by governance.
As businesses give AI systems greater responsibility for executing tasks and influencing decisions, the ability to maintain human oversight, trace actions and establish clear accountability will become increasingly important.
For Nigerian and African organisations, the opportunity lies not simply in adopting more AI, but in building systems that combine technological capability with responsible human judgement.
Why It Matters
Agentic AI could change how organisations perform routine work, coordinate services and make operational decisions. But its growing ability to act rather than merely assist also creates new governance and accountability requirements.
The EY experts’ message is that organisations need to define the authority given to AI agents, limit their access to sensitive systems and maintain appropriate human oversight.
The technology may continue to evolve rapidly, but responsible adoption will depend on organisations establishing clear boundaries around what AI systems are permitted to do and who remains accountable for the results
Weng Global – stories beyond borders
Sources
EY Nigeria
EY
ThisDay
