From Chatbots to Doers: The Next Wave of AI

Generative AI, also known as conversational AI, changed the way we interact with technology. It can understand what we say and create human-like text and images.

However, AI agents take things a step further.

Think of it this way: conversational AI talks, while AI agents actually get things done. They can observe what's happening, understand what needs to happen next, make a plan, and take action.

That ability to act is what makes AI agents so powerful. Here are four things they can do (at the moment):

1. Plan and reason. AI agents can take a big, complicated goal and break it down into smaller steps. Then they can work out how to complete those steps.

2. Take action on their own. AI agents can make decisions and get things done without someone guiding every move. Of course, people can decide how much freedom an agent gets. It might work completely on its own, or it might stop and ask for approval before taking an important action. As people gain confidence in the agent, they can give it more freedom.

3. Remember what happened before. AI agents can keep track of information and past interactions and use that context to make better decisions. This makes them useful for tasks that involve many steps or happen over a longer period of time.

4. Work with existing tools and systems. AI agents can connect to the software, databases, and other tools a company already uses. So instead of just telling someone what to do, an agent can actually do the work.

From Saving Time to Changing the Game

Companies have been experimenting with AI agents for the past few years, and the results are starting to get interesting.

The biggest opportunity isn't simply doing existing tasks faster. It's using agents to rethink how work gets done in the first place.

Here are a few examples:

• R&D: A shipbuilding company used AI agents to handle parts of the design process. It cut the engineering resources needed by 45% and reduced the time needed to design each ship deck by 80%.

• Sales: A global logistics company used agents to automate its response to requests for proposals. The result: efficiency improved by 30% to 50%.

• Sales and marketing: A large Southeast Asian bank gave relationship managers AI-powered, real-time suggestions for creating personalized customer offers. Assets under management increased by 5% to 10%, while customer conversions increased four- to sixfold.

• Customer engagement: A global cosmetics company created a GenAI-powered beauty assistant. The new experience increased conversions by five- to tenfold compared with traditional digital channels.

• Supply chain: A leading industrial company used an AI agent to run supply chain simulations, spot potential risks, and suggest ways to address them. The company increased its EBIT margin by 3 to 10 percentage points.

So, what's the common thread?

These companies aren't just using AI to automate individual tasks. They're using AI agents to rethink entire workflows.

Agents can connect work that used to happen in separate departments, remove unnecessary steps, and help work move more smoothly across the organization.

But getting this kind of value doesn't happen by simply adding an AI agent to an existing process. Companies need to start with the business problem they are trying to solve and then figure out where agents can make the biggest difference.

This is where the (BCG) 10–20–70 rule comes in. Roughly speaking, 10% of the effort goes into algorithms, 20% into technology and data, and 70% into people and processes.

That last 70% matters a lot.

And as AI agents start making decisions and taking actions, responsible AI becomes even more important. Companies need to make sure agents operate safely, follow the right rules, and don't create new risks.

New Possibilities, New Challenges

AI agents can do a lot, but they're not magic.

Getting them to work well means taking a hard look at the way work gets done today. In many cases, companies will need to redesign their processes before they automate them.

Here are some of the biggest things to think about.

1. Fix the process first

A bad process doesn't become a good process just because you automate it.

Before handing a workflow to an AI agent, companies should step back and ask: Do we actually need all these steps? Who really needs to be involved? Can some tasks be combined or removed?

The goal is simple: fix the process before you automate it.

2. Get the knowledge out of people's heads

AI agents need to know how a company wants a process handled.

That's not always easy. In many businesses, people rely on experience, judgment, and unwritten rules to get things done. The "right way" of doing something may never have been formally documented.

With AI agents, companies need to make that knowledge much more explicit.

3. Clean up the data and systems

AI agents are only as good as the information and systems they can access.

An AI agent might be able to pull information from five different systems, but if the data is outdated, incomplete, or inconsistent, the agent's answer won't be very useful.

The same goes for the models and other tools the agent relies on to make decisions.

Good AI agents need good data and reliable systems.

4. Take risk and testing seriously

A traditional AI model might answer a question. An AI agent can take a series of actions to accomplish a goal—and may use several different models and tools along the way.

That's part of what makes agents so useful. But it also makes them harder to test and control.

Companies need ways to catch mistakes, understand why an agent made a decision, and step in when something goes wrong. They also need clear safeguards and ways to monitor what agents are doing.

The more agents a company deploys, the more important these controls become.

5. Make agents fit the business

There is no one-size-fits-all AI agent.

Companies will need to adapt AI agents to their own processes, goals, data, and ways of working. Fortunately, they don't have to build everything from scratch. Off-the-shelf AI agent platforms can provide ready-made tools and workflows that help companies get started faster and keep costs down. Companies can then customize those AI agents for the areas that matter most to their business.

The idea is to buy the basics and build where it really matters.

Making Agents Part of the Business

Bringing AI agents into a company isn't really about adopting another piece of technology.

It's about changing how the business works.

Companies shouldn't think of agents as standalone tools that sit on the side. They should think of them as part of the way the business operates: helping people work more efficiently, make better decisions, innovate faster, and ultimately grow.

AI agents probably won't be a unique competitive advantage forever. As the technology becomes more common, having AI agents will become more like table stakes.

The real advantage will come from how a company uses these AI agents. Companies that have strong processes, high-quality data, proprietary knowledge, and well-integrated AI agents will be in a much better position to create lasting value.

In the end, it's not about having the most AI agents. It's about knowing where they can make a real difference—and redesigning the business to take advantage of these AI agents.