Artificial intelligence agents can transform a small staff into a productive workforce that can tackle tasks typically handled by larger teams.
While most people are familiar with AI chatbots like ChatGPT, Perplexity AI or Claude, AI agents are the next level for productivity at work. They provide information like chatbots, and they do the task.
They can also manage other AI agents in more complex work.
Companies like Salesforce and ServiceNow, and platforms like Agent.ai, offer AI-powered agents that can support customer service, data management, workflow automation and sales processes. Cloud providers Amazon Web Services, Microsoft Azure and Google Cloud also offer agents on their platforms.
“The impact for business is going to be profound,” David Johnston, adviser to investment firm DLTx, told PYMNTS. “Having expert-level AI available to anyone is going to accelerate all types of work that involve writing, analyzing, designing and digitally creating just about anything,”
“Soon it will act as a multiplier with every worker being able to achieve 10 times and then 100 times as much productivity per day,” he added. “2025 is going to be remembered as the year this trend became mainstream.”
Business leaders without a technical background can assemble their own AI agent teams using plain language — and learning to do it takes minutes. They can also create their own AI agents without needing to know how to code.
Examples of AI agents include:
The only publicly available collection of AI agents thus far — free or for hire — is offered by Agent.ai. Its creator, HubSpot founder and Chief Technology Officer Dharmesh Shah, compared it to the Apple App Store. Agent.ai is an app store for AI agents that will “eventually” move to HubSpot, he told PYMNTS.
The idea for Agent.ai started when his wife wanted to create a video out of short clips on YouTube, “something that would have taken several steps, tools and hours of manual work,” Shah said. So, he built an agent to do the video compilation, which it did in minutes.
Since launching in May, Agent.ai has amassed 500,000 users, with nearly 12,000 builders signing up, Shah said.
“Agents on the network today can do everything from company research to building ideal customer profiles, call analysis, YouTube transcript generation, and so much more,” he said.
A sales representative typically spends two hours a day meeting with clients and the rest of the time researching their companies, for example.
On Agent.ai, the Conversion Rate Optimizer AI agent can do the same task in minutes. Put in the client website and it will generate a report that includes sector, industry, number of employees, location, company type, founding year, specialties, founders or executive team, competitors, products, news articles, web traffic and other data.
The agent also generates an FAQ, such as “Do they offer a subscription product?” Users can ask their own questions and then add them to the FAQ so the agent will remember to ask them about another company. Users can also make the AI agent track new developments coming from the client — and notify them.
To create a team by adding another agent to work with the Conversion Rate Optimizer agent, one possible candidate is the Industry Analysis AI agent. This agent does a deep dive into the client’s industry to spot trends and opportunities. It can help deepen the sales representative’s understanding of the marketplace dynamics affecting the client.
“Growing businesses have the most to gain from this level of transformative productivity, and that’s exactly what we’re aiming to do,” Shah said.
Since many AI agents use large language models, especially those used to generate text, users should always double-check the content in case they hallucinate or make things up.
Even multimodal models — those that can accept and generate text, images, video or audio — may also hallucinate, according to a paper published by Amazon researchers. For example, the generated text response may not match the generated image because it describes missing objects.
Shah said Agent.ai can access OpenAI’s GPT-4o or o1 models, Claude 3.5 Sonnet, Gemini Pro, Perplexity and others. It also can use image generators including DALL-E 3, Stable Diffusion and Ideogram.
“Agents are like very complicated Lego blocks,” Johnny Gabriele, head analyst of blockchain economics and AI integration at The Lifted Initiative, told PYMNTS. “There really is no limit to how you can put these things together. But at the end of the day, they are agents, and agents are going to come with limitations.”
Human Voice Media CEO Bob Hutchins told PYMNTS that using AI can lead to “erosion of leadership and decision-making skills over time as there becomes an overreliance on AI guidance.”
“Additionally, there is a question of the psychological issues that may arise,” he added. “Would employees or customers trust the decisions” made by the agent? “The trust gap may have to be bridged by human-AI mediators.”
For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.
