Orby AI: ‘AI agents don’t just do the work – they understand what you’re doing’

By 2030, 60% of US workers could have 30% of their work automated, according to McKinsey research. 

When broken down, that’s roughly 72 billion hours of labor annually in the US alone – equivalent to around 100,000 people’s lifetimes each year.

This is just one of the statistics that inspired Will Lu to Co-Found Orby AI – an AI agent platform giving enterprise employees the tools to scale their work, such as delegating repetitive tasks to computers.

As an expert in the field of automation and agentic AI, UNLEASH spoke exclusively to Lu to discuss the current AI landscape – as well as what we have to look forward to in the future. 

The reality of AI systems

Generative AI is commonly believed to have PhD-level intelligence, with the idea that Artificial General Intelligence (AGI) is just around the corner. 

But Lu highlights that although Gen AI is capable of solving some incredible problems, the real challenge for widespread enterprise adoption is that it’s often “hit or miss”.

He explains: “Most of the time, it [Gen AI] does solve the problem, but there are instances where it makes silly mistakes. If you look at the benchmarks, no matter how impressive they seem, the best performance typically hovers around 90%, even for relatively simple tasks like high school-level math problems.”

He expands on this, stating that when Gen AI is used extensively, users will find that sometimes it doesn’t always follow very specific instructions or it misses the mark. 

This therefore means users should proceed with caution, using it for daily tasks such as drafting materials or doing a touch-up – but remembering to go back to check it.

However, Lu warns that when leaders look to reliably apply Gen AI in an enterprise environment, like handling financial processes, it becomes problematic.

Every single dollar mistake could lead to millions of dollars in losses, whether it’s due to compliance fines or making incorrect payments to vendors, which is costly to resolve,” he says.

“That’s why we’re developing our own AI agent model at Orby, to address this problem and to make AI scalable and reliable for enterprises.”

There are also a number of challenges that AI agents face when addressing HR scenarios – particularly those that involve complex human judgment, such as hiring decisions or performance evaluations. 

Lu gives the example of HR policies that require candidates to have at least five years of experience. 

“That’s a very simple compliance or policy requirement,” he notes, “but agents may not always be able to follow those requirements. For example, they might decide that four years is enough to meet the five-year requirement – things like that can happen”.

Another issue Lu raises is that agents can miss important information. For instance, if someone comes from a small country where their school or major is less known, the agent may not reliably understand that information and could make random decisions based on what it has seen in other resumes.

“I think it’s crucial to be able to define policies in a clear, symbolic way so that you can trust the agent to follow those policies and make reliable decisions,” he adds. 

“It’s also important to have an efficient human review process built in so humans can review the results.”

This is why through Orby, Lu is offering insights based on a large volume of results – such as a manager makes 100 hiring decisions, Orby will provide tools that allow you to analyze the process, identify outliers, recognize potential risks, or spot opportunities to improve efficiency. 

How AI agents will shape the future

Although there are some nuances to AI agents, Lu is hopeful of the impact they will have in the future. 

In fact, he states that the biggest opportunity with AI agents is that they will become widely available to employees, “providing unparalleled agility and scalability”. 

To highlight how this could happen, Lu uses the example of finance – a sector where there are often business seasons with unpredictable, high volumes of work that come in during specific periods. 

“Usually, companies can’t hire enough people to handle that high volume upfront, so employees are expected to work long hours during busy seasons and then take extended vacations when things slow down,” he says. 

With AI agents, businesses can scale up quickly and handle a larger volume of tasks. Employees can manage these tasks on-demand and more efficiently.”

This agility means that even though employees may know how to do their work, it’s still time consuming. For instance, clicking a button, opening a screen, and submitting receipts can take time. 

But now, with machines understanding the tasks and doing them reliably and quickly, companies can respond to customers, bosses, and colleagues faster. This level of agility and scalability becomes critical.

The second point is something that Lu believes is often overlooked by most AI agents and vendors. He says: “AI agents don’t just do the work for you – they also understand what you’re doing. 

“As employees go through their daily work, the agent can track what they’ve done, summarize the tasks they’ve completed, and compare it with other users,” Lu highlights. 

“It can identify areas for optimization, things that can be improved, or even highlight potential risks. It can also alert employees if they’re not following policies correctly. 

“Having agents that can help review and compare work in this way becomes a powerful tool for managing daily tasks.”

So with this in mind, when will your business be ready to embrace AI agents?