In the tech and finance industries, all eyes are on generative AI.
Companies have poured billions of dollars into genAI tools like ChatGPT and GitHub Copilot, and a Bloomberg Intelligence report projected that the total market for genAI will reach $1.3 trillion by 2032.
Yet despite the massive investment, the wheels are starting to come off the AI hype machine. An influential Goldman Sachs report makes the case that genAI isn’t delivering enough value to justify current and projected spending levels. It could be that we’ve reached the “trough of disillusionment” phase of the hype cycle, or it could be that genAI tools like ChatGPT never deliver the value that the market once predicted.
We can argue about whether or not the hype around genAI is real. While the technology may not be replacing knowledge work to the degree and quality that people predicted, it is undoubtedly leading to changes in hiring and organizational structures.
But as the debate rages on around genAI, we’re misunderstanding the technology that is actually going to disrupt markets and industries: AI agents. To thrive in a future shaped by AI, companies and employees need to recognize where AI agents will have the biggest impact and prepare accordingly.
Why GenAI Is Falling Flat
What differentiates genAI agents is how they gather and apply knowledge. Some AI agents, including sprawling LLMs like ChatGPT, take a bottom-up approach to gathering information: when a user asks a question, the LLM draws on all of the information it could find on how to answer the question.
Businesses need excellent decision-making to solve complex problems. You would think that a massive volume of information increases the likelihood that the LLM provides a good answer; counterintuitively, it’s precisely that volume of information that leads to poor performance. A bottoms-up system like ChatGPT is going to give the average answer of all people who attempted the endeavor. The problem? The average answer is bad.
Growth and innovation aren’t the result of average performance – 10,000 mediocre employees will not be able to reach the same achievement as a few truly brilliant minds.
What does that mean for AI solutions? The truly disruptive AI agents will be those that find answers by searching through the top 10% of relevant information — marrying bottom-up knowledge gathering with top-down expertise. For an AI agent to deliver truly transformative value, it must move beyond retrieval knowledge — finding a fact in your existing data — to achieve procedural knowledge — knowing what ought to be done in a given circumstance.
The Disruptive Potential of AI Agents
It’s no surprise why tools like ChatGPT captured the public imagination. They bring together information on an overwhelming scale, and they can perform an amazingly diverse set of tasks: from translation to poetry to code debugging. On the other hand, AI agents are highly specific — built to accomplish a narrowly defined set of tasks and business goals. An AI agent that does nothing but trade municipal bonds isn’t particularly exciting, even if it has the potential to generate consistent profits.
But while ChatGPT may not write such good poetry that it will make human poets obsolete, effective AI agents specifically automate many of a given role’s most time-consuming tasks. A logo design agent can create dozens of options in the time it would take a human to design one; a sales engineering agent can automate everything from designing solutions to creating demonstrations; a quality assurance agent can handle the workload of dozens of QA engineers. Human employees will move from doing the work to reviewing the work.
AI agents have the most potential to disrupt highly specialized roles – those that have emerged either to bridge the gap between two core business pillars (like sales engineers between the sales and product functions) or to tackle a challenge of abundance. In the case of a logo designer, a company’s graphic design unit might have had such a large workload that they were able to allocate one employee to focus entirely on logo design. With AI agents now able to automate the bulk of a logo designer’s workload, those employees are now left with little to do – and a lack of experience in adjacent areas.
How to Respond to the AI Invasion
What should you do if your role is threatened by an AI agent? How should business leaders think about applying these tools in their current workflows?
For employees, one option is to double down on expertise. Sales engineers won’t be wiped out entirely by AI, and those with exceptional abilities and technical knowledge will still be able to deliver value. But there are limits to this approach: only the best of the best will be able to stay on in their current positions, and even the world’s greatest telephone switchboard operator ended up losing their job to digital switching platforms.
The more durable approach is to focus on using the AI tools themselves – learning how to work with AI agents instead of fighting against them. The core pillars of a business – sales, product, engineering, marketing – will always be there. Employees who can use AI agents to deliver new value to one of those pillars will be essential in the coming years.
For business leaders, there’s no time to waste: we need to be integrating automation and AI into our operations as soon as possible. Weaving AI into the fabric of your company offers two distinct benefits: first, it provides leverage to maximize the productivity you get out of each employee and asset; but perhaps more importantly, embedding AI means that your systems will improve in tandem with the AI models themselves. If an AI vendor updates the LLM you’re using and delivers significant performance enhancements, your AI agent is going to improve overnight without requiring as much work from your team. Investing now in AI agents puts you on the road to perpetual progress.
The rise of AI agents represents a once-in-a-generation moment of opportunity. The ongoing emergence of AI agents has placed us all on a level playing field, and those who can master and apply these tools the fastest will race ahead of the pack.
Look past the genAI disillusionment and you’ll see that the moment of AI disruption is already here. Will you be controlling the moment, or letting the moment control you?
Author
Get our newsletter and digital focus reports
Stay current on learning and development trends, best practices, research, new products and technologies, case studies and much more.
