Expert details how agents are better than chatbots, and how it can help workers

For employers, the allure of generative artificial intelligence (GenAI) may soon be replaced by a newer trend: agentic AI.
“Agents” is “the next level of maturation or evolution of the digital assistants,” says Keith Bigelow, chief product officer at Visier, in talking with Canadian HR Reporter.
While digital assistants like ChatGPT remembers instructions for the session that you're currently having, agentic AI is designed to actively think, act, and improve over time, he says.
To use agents, users should have it go through a process similar to onboarding a worker, he says.
“It needs an onboarding plan that you write.”
Agents store the plan to their memory and are also able to pull new data from outside sources.
Agents also go through a process of “reflection” to make sure that the output they are providing “is of quality,” he says.
“Those four things are all different from a chatbot, because a chatbot is just a conversation back and forth. But this is something where I can give you a task that's very clear.”
Over nine in 10 (92 per cent) of small and medium-sized business (SMBs) leaders are confident in their company's growth prospects over the next three years, and they are banking on AI and automation, according to a previous report from KPMG.
Concerns about agentic AI
Despite its potential, agentic AI presents challenges, especially around data security and model reliability.
Bigelow notes: “One concern is security – like, am I asking questions that I shouldn’t know the answer to?”
Another issue is hallucinations, where agents generate incorrect or fabricated responses, he says.
“If they have sparse data, models may hallucinate... We mitigate this by using techniques like retrieval-augmented generation (RAG), which ensures the agent fetches accurate, relevant data.”
While hallucinations won’t disappear overnight, companies can build safeguards to minimize their impact.
“Double-check outputs because it’s probabilistic rather than perfect,” says Bigelow.
And over-reliance on GenAI is the top ethical concern for workers, according to a previous report from GetApp.
How to implement AI successfully
Before jumping over to a more advanced version of GenAI, it’s important that employers understand their reasoning for turning to the technology.
“There's a category of companies that we see that has this fear of missing out, runs in and is hoping to not miss the opportunity, but may or may not actually have all of the rigours in place to check and manage” if they are receiving real benefits from AI, says Bigelow.
While these companies have mandated the use of GenAI for their workers, “they haven't set out a business case for why they're doing it,” he says.
This reflects a recent IBM Canada report that found that while the majority of companies are making progress when it comes to implementing their GenAI strategy, far fewer IT decision-makers (ITDMs) are seeing positive results.
The companies that are implementing GenAI best are “absolutely engaging employees early,” says Bigelow.
These companies are “surveying the employees, asking them what kind of training they need, and explaining how it's going to be used and what the goals are”.
Half of Canadian employers think workers are not prepared to use GenAI, according to a previous Deloitte report.
For employers planning to tap into the powers of agentic AI, Bigelow has this recommendations:
- Write a business case.
- Specify what you want to achieve.
- Set up metrics to track progress.
- Engage your workers in the deployment.
- Track what’s actually happening as compared with the plan.