NASHVILLE, Tenn.–When it comes to artificial intelligence, it’s likely you and your credit union feel behind. That’s OK, but it doesn’t mean anyone can afford to “sit this one out,” according to one expert, who offered some ideas on what’s coming and where to start. 

Steve Brown, a former executive with Intel and Google’s Deep Mind project, and the author of “The Innovation Ultimatum,” told the Lending Tech Live event hosted by Origence that he believes that due to the rapid advancement in AI, “we may see more technologically driven change in the next five years than we have experienced in the last 50. Things could take off very, very fast.”

A ’New Age’

Brown said it’s his view the world is on the precipice of a new age.

“What we are about to see is the emergence of capable, powerful intelligence where the cost is near zero,” Brown said. “What that means for every business on the planet is now you have intelligence on demand.”

While prior ages were about creating “mechanical muscle,” the new age is about creating what Brown called “electrical minds” that he said “will help us achieve more.”

Machine learning has created scale when it comes to training AI models and that, in turn, has led to the creation of numerous new capabilities, he said, including the expanding ability to recognize images and objects. 

The Next Waves of AI

Brown noted the two types of AI most often used today and familiar to just about everyone at this point are:

• Traditional Machine Learning & Artificial Neural Networks. This is the predictive AI that can do classification and clustering of data, such as spam filtering, recommendation engines and portfolio optimization. 
• Multimodal Language Models. This is the generative AI that does content creation. “Generative AI is the translator between different spaces, such as text to images,” said Brown. It is primarily for answering questions, doing deep research and ideation, code generation, video generation and more.

Pending Breakthroughs

All manner of amazing breakthroughs are coming, according to Brown, including new protein structures, scientific papers, new alloys, synthetic organisms, and even novel inventions AI creates on its own.

For credit unions, Brown said AI will make many employees much more productive. He observed that while CUs and other companies have incredible amounts of data, the challenge is the average information worker spends 32 days per year just looking for the information they need to do their jobs.

“What if you could surface that information…and make it easier for them,” asked Brown. “One of the ways you can unlock that data and make it available to the right people at the right time in the right places is to create what’s called an enterprise intelligence agent, where you connect your company data, your product data, your customer data, your original data, into what’s called a vector database that puts it in the same kind of format. You then use AI to make it accessible through these agents through a chat bot or something similar.”

Agentic AI

Brown predicted 2025 will be remembered as the year of “agentic AI.”

“One of the key things that is important about these new AIs is they have the ability to reason,” Brown said, outlining the differences between what’s known as System One Thinking and System Two Thinking, and explaining that AI is now moving into System Two Thinking. 

“An agent is a smart piece of software but it has the ability to have this sort of–think ChatGPT Gemini at the core, or Claude–and then wrapped around it you have a perception module so the AI can understand a bit about the environment,” Brown explained. “So, it might be the ability to see the screen on your laptop so they can help you with tasks. Then it has a memory module so that not only can it learn your preferences and your needs, but it also can learn from its own mistakes. It will make mistakes but it will get better over time at tailoring what it does to serve your needs.

‘Reasoning Capability’

“Importantly, it has this planning module, so it uses that reasoning capability to break problems down and reason its way through them to then build a plan on how to solve a problem or to perform a complex task, and then to execute on that task,” Brown continued. “It has an action module. It can use tools that can actually do things, and that’s why it’s called an agent: it has agency in the world.”

Brown predicted the world is moving toward having a trillion such agents in place working in different realms.

“There will be agents to do everything and they will be there at your beck and call ,” he said.

Three Functions

Brown predicted there will be three functions of agents:

• Offload tasks (agent will do the boring stuff)
• Elevate the Performance. “We work alongside them to do things neither of you could do alone,” Brown said.
• Extend. “Allows you to do something you could never do alone.”

Brown said credit union leaders are going to need to prepare for having digital coworkers and for their flaws.

“They are going to be super smart, but naïve. They will make mistakes,” he said. “By its nature it will probably always have imperfections and need your oversight.”

The agents will initially be chatbots, then become voice agents, and then become agents “that you can see. They will have synthetic digital faces, and you can make it whatever you want, I suppose,” he forecast. “Your next coworker might be a machine and that could be good for you. These digital agents work for electrons. There will be lots of competition.”

Brown noted such agents are already being seen in HR departments in prescreening candidates. “The AI even listens in on the interview, takes notes and then provides an assessment,” said Brown.

When it comes to deploying agents, Brown cautioned his audience to not  just focus on productivity, but to focus on performance to boost quality, efficiency, capacity, abd creativity. 

What it all means is everyone becomes a manager, according to Brown, who said there are already tools in the market to build agents with build-and-drop interfaces, such as MindStudio.

Spatial AI and World Models

While robots will be less relevant to credit unions, he said leaders should nonetheless understand what’s coming. He outlined the technology robots need in order to operate effectively, key to which is spatial AI and the ability to perceive 3D space and time, rather than depending on written instructions.

Brown shared a video from Project Astra as part of Google’s DeepMind. A demo can be found here.

According to Brown, the first new device to launch since the introduction of the iPhone will be smart glasses, which should emerge later this year.

“You take agentic AI and spatial AI and put them in a robot and you get physical AI,” Brown said, predicting robots/automation will fill the labor shortage being seen in physical jobs. “Robots are going to free us from robotic work.”

Process Orchestration

When it comes to using AI, Brown urged credit union leaders to map out the many processes inside their operations and to ask which of these is still best done by a human, which can be done by an agent, and which can be done by a robot. 

He offered the graphic below as a means of mapping out those processes.

Brown said companies that have used this approach have seen ROI across multiple categories.

The Next Stop: Human Level Intelligence

In response to the question on everyone’s minds, when will AI become as smart as humans, Brown offered, “The smart money on 2027 or 2028. My money is on 2028. You may have an AI good enough to do AI research so it can develop a new AI and develop a new AI, and eventually you could have super intelligence.

“You have no opportunity to sit this out,” he cautioned. “To stay relevant you have to invest in AI. You need to make time to learn and experiment with AI tools.”

The Resources

Brown urged CU leaders to explore Perplexity and Eleven Labs to learn more, as well as his own website, which offers numerous resources at www.stevebrown.ai.