AI agents is a growing term in GenAI. But what are they? How will we use them?

NinjaTech AI’s CEO, Babak Pahlavan, sheds light on this burgeoning phenomenon of AI Agents. Offering a glimpse into the intricacies and implications of reshaping our daily interactions in this tech-driven world.

There’s been accelerating publicity lately around “AI agents.” In November 2023, Bill Gates wrote on his blog about how AI agents will completely change how we use computers. In February 2023, CNBC reported that in the last several weeks, big tech CEOs such as Satya Nadella of Microsoft, Mark Zuckerberg of Meta, and Sundar Pichai of Alphabet discussed their companies’ increasing focus and investment in AI agents numerous times in their respective earnings calls.

We’ve also recently seen well-funded AI agent startups emerge from stealth operations, as Brian Sozzi of Yahoo Finance reported in February 2024, including Bret Taylor’s company Sierra, which focuses on building AI agents for customer support. AI agents’ investment, focus, and deployment are on a rapid trajectory – but what exactly are AI agents?      

The Evolution Beyond Assistants and Co-pilots

While similar in their intent (i.e., to help users be more productive), AI agents differ from assistants and co-pilots because they can operate asynchronously or autonomously to complete tasks. The next-gen technology behind AI agents enables them to dynamically take on chaining problems, which allows more complex reasoning.

In non-technical terms, this means that AI agents can save each user more time and more money and make them more productive by taking over entire tasks for the user rather than helping them synchronously while they work. In many ways, AI agents are akin to having a personal or executive assistant to solve specific problems.

Let’s look at the example of a sales development representative (SDR) by using three common components of an SDR’s workflow:

  1. Researching and selecting prospects to sell into.
  2. Conducting email outreach to those prospects.
  3. Booking time for a call or demo with interested prospects

Co-pilots are designed to help the pilot (i.e., the SDR) while going through this workflow. A co-pilot could synchronously help the SDR at certain stages by researching a prospective company in real-time or drafting an email for outreach as they prepare to send it.

With an AI agent, this workflow can be put on autopilot by working autonomously on the entire chain or big pieces. The SDR could give an AI agent instructions, such as researching the top 10 prospects, drafting the emails uniquely for each prospect, notifying the SDR when the analysis and emails are ready, sending the emails once the SDR has approved them, and booking the meetings for me with the prospects that respond to the emails. Both the co-pilot and autopilot examples result in time savings and productivity enhancement.

The primary difference is that the co-pilot makes the user more productive while they work, while the autonomous AI agent can take over the work for the SDR and provides more time savings than the co-pilot. To be clear, co-pilots and autopilots will both be needed as there are inherently synchronous tasks and others that will become asynchronous.

See More: How Nadella’s AI Vision Can Reshape The Future of Developers

The List of AI Agent Tasks Is Growing Rapidly

2024 and 2025 will see explosive AI agent growth. A wave of AI agent startups is emerging as venture funding pours rapidly into this space; new market maps are regularly drawn. Perhaps most encouraging is that startups find elastic alternatives to the scarce and expensive custom chips required for the training and inference needed to build these artificial “niche” intelligence agents.

We’ll undoubtedly see smaller, custom large language models (LLMs) and incredible advances in self-learning and verification that offer increasingly compelling economics for AI agent companies and, in turn, their users. Anticipating this coming tectonic shift, we’ll likely see a large custom chip provider take center stage with an elastic computing model that enables AI agents to thrive.

Over the next two years, we’ll see a rapidly increasing spectrum of tasks that AI agents can tackle as they emerge as a powerful force for consumer and business productivity. In the near term, the tasks we’ll see AI agents autonomously tackle will most likely create the highest administrative burden for personal and business users.

Looking at recent McKinsey and Goldman studies, the problem of admin tasks is material, representing almost 30% of an average employee’s time, with an estimated 50% of these tasks able to be addressed by AI. The most immediate AI agents’ use cases will include scheduling meetings, conducting research, booking travel, shopping, and customer support.  

Overcoming Hurdles To AI Agent Adoption

The future is imminent; there are a few hurdles the AI agent adoption wave will need to overcome to deliver on its potential successfully:


Analogous to autonomous vehicles, where not every rider is comfortable with a self-driving vehicle, not every user will be immediately comfortable with an AI agent taking over components of their workflow autonomously. Trust will need to be built through repetitive successful tasks; AI agent companies must aim for incredible accuracy in their task delivery. Fortunately, the self-verification technology emerging in this field drives incredible advances and success ratios that will build trust quickly.

Over selection

With so many AI agents likely to emerge in the near future, customers could be overwhelmed with a tsunami of options. Like digital streaming services, users could quickly find themselves with numerous subscriptions. Users must be equipped to evaluate all the options for each type of task and select a conglomerate of agents. We’ll likely see a wave of AI agent expansion and eventual consolidation as the winners emerge in each category.

Onboarding and training

AI agents are novel and unique technologies that tackle such a historic productivity challenge. It will take time to fully utilize all the ways an AI agent can tackle admin tasks, whether by job function or vertical. In the SDR example above, we are likely just scratching the surface of what is possible. Enabling users to explore all these possibilities will be critical to achieving AI agents’ full potential.

Whether you’re a senior executive, a solo entrepreneur, a customer calling into a support center, or a personal user booking your next trip, AI agents will be embedded within your tech experiences imminently. Understanding their importance and being prepared to maximize their utility will result in immense productivity savings and next-gen user experiences.

How can you prepare for AI agent adoption hurdles and implement strategies to leverage their growth potential? Let us know on Facebook, X, and LinkedIn. We’d love to hear from you!