Agentic AI: The new frontier in business transformation

Agentic artificial intelligence (AI) are systems that can make decisions for themselves based on historical data and real-time information. 

An AI agent can then execute its decisions to accomplish tasks, without human prompting, intervention or supervision.

This differs from generative AI, which can create content or analyze data in response to commands.

So, whereas an employee might prompt generative AI to compose an email based on their needs, an autonomous email agent looks through all incoming emails, flags important ones for the recipients to review, and can reply where necessary.

Applications such as autonomous email agents are only the beginning, though. In time, AI agents are likely to be integrated into ever more diverse and complex workstreams. 

As the report puts it:

“AI agents as a co-worker are expected to become increasingly common. Today, AI agents tend to be focused on specific tasks. In the future, AI agents will combine to manage multiple tasks.”

In other words, human workers at every level might find themselves operating alongside these agents. As such, they may be able to focus on the bits of work they like best and where they can generate most impact. 

Equally, consumers are likely to turn to agentic AI to perform diverse aspects of life admin both great and small. The Citi Institute highlights an important but often time-consuming action, which few relish: 

“For example, if a mortgage is up for renewal, the bot can notify the customer of the end date, conduct research to find the best deals, determine the right product, and fulfill formalities for refixing the mortgage rate.”

In such scenarios, the consumer would still have the power to make the final decision. However, the legwork of gathering, analyzing and comparing deals would all be done for them.

While this is just one instance, similar tasks have proliferated for the typical adult in developed economies. 

The likes of settling bills, selecting gifts, compiling shopping lists, making appointments, keeping track of finances, switching utility suppliers and complying with tax and other regulations can collectively eat up days of someone’s time each year. 

Being able to say “do it for me” to one or several AI agents would have obvious attractions for a great many people, therefore.

The “do it for me” economy could potentially have far-reaching effects on how we live and work.

“AI, and agentic AI could have a bigger impact on the economy and finance than the internet era,” say the report’s authors.

In their view, “AI, like previous technology revolutions, should lower costs, raise productivity, and boost overall economic activity.”

This sort of boost is, of course, sorely needed. Productivity growth – the ultimate driver of our prosperity – has slowed alarmingly across much of the world, but especially since the Global Financial Crisis of 2007-09.

The resulting slower real wage growth has fed into growing social discontentment and more rancorous politics in many countries.

Of course, the rise of agentic and other varieties of AI is likely to create winners and losers, even if they unleash growth and progress overall. 

Among those likely to be impacted, according to the report, are the contractors and third parties to whom repetitive tasks have often been outsourced. Many of them have tended to be in lower cost developing economies.

Naturally, it is easier to envisage such displacement than the new tasks and jobs that may emerge in future.

Tasks that have not so far been heavily outsourced may prove less susceptible to disruption, by contrast.   

Overall, agentic AI has the potential to drive growth and progress.

At the same time, though, it may also be deployed to cause deliberate harm.

Bots – task-specific automated programs – already make up more than half the traffic on the internet, much of it malicious.

Among the nefarious activities in which they are involved are hacking, spamming and ad fraud.

Agentic AI could take all this to a new level:

“Bots are simple rule-based programs. AI agents are much more advanced systems that can learn, adapt, and make decisions based on their context and objectives. They are smart bots, with an ability to evolve and act.”

For example, a fraudulent AI agent could seek to dupe social media users on a large scale into joining phony investment services. It might improve its swindling tactics in real time based on victims’ responses and profiles, while also concealing the proceeds of its crimes more quickly and efficiently than ever. 

Put simply, the world’s digital wrongdoers could also be on the cusp of a productivity revolution.

Criminality aside, agentic AI has scope to create harm even without malicious intent. Agents could develop biases or other corrupted features, leading to costly mistakes. The more autonomously they are allowed to operate, the greater such risks may become.

Ensuring rigorous oversight without stymying agents’ capabilities thus represents a major challenge. 

With agentic AI already a reality, this technology looks set to play an increasingly influential role in the economy and everyday life in the coming years.

A key driver of ongoing advances is likely to be the money that is flowing into this area.

As the report highlights, AI startups accounted for 37% of venture capitalists’ total funding, with an even higher share expected this year.

Within that, “autonomous agents and digital co-workers saw the biggest growth in VC deal activity in 2024.”

At Citi Wealth, we see AI as a transformative force, with many implications for investment portfolios.

While 2025 has been widely hailed as the “Year of Agentic AI,” we share the view that the phenomenon still has far to run.