The excitement around AI agents is feverish. In 2025, declared by Forbes the year of agentic AI, Deloitte predicts that 25% of large enterprises will implement agent-driven schemes. Futurum forecasts that AI agents will contribute a staggering $6 trillion to the world economy by 2028.
What Agentic AI means is a machine deciding what steps it should take to achieve a goal and then executing those steps autonomously. It’s the difference between AI suggesting a travel itinerary and going ahead and booking flights, making restaurant reservations and negotiating an overdraft with your bank. It’s the leap from using AI to write blog posts to letting it run your entire marketing strategy - planning, executing, analysing, and optimising with no human in the loop.
What could go wrong when you hand AI the reins? Plenty. Empowering millions of intelligent agents to pursue goals independently poses staggering societal and security risks. Unintended consequences aren’t just likely - they’re guaranteed. Some will be disruptive; others could be seismic. These developments raise urgent questions about accountability, control, and oversight - questions we’ve currently no idea how to answer.
But the potential benefits are equally vast. To see what autonomous AI systems could do for you, consider how Phuket marketer Gael Ovide-Etienne has been able to automate marketing for his wife’s property agency.
Every few hours, his AI system springs to life, identifying the top questions potential Phuket property buyers are asking. The most compelling becomes a headline, which is spun into a blog post and published on the agency’s website. That blog is then transformed into a podcast, voiced by two lifelike digital personalities, and converted into a video clip. The system produces four new videos a day which are pushed out across multiple platforms - YouTube, Facebook, TikTok and anywhere else that supports reels. The agency’s YouTube channel alone gained tens of thousands of subscribers in just a few weeks and business grew tenfold. Gael’s daily input? Zero.
“Once in a while, I check our feeds - and I’m always amazed by the quality,” says Gael. The entire operation costs his company, Turquoise Digital, as little as $10 a day in API compute costs. Replicating the same output manually - either by doing it himself or hiring a designer and account manager - would cost fifty to a hundred times more.
In fact, though, what Gael is using is not true agentic AI, but a precursor: AI workflows. AI workflows are an intermediate step towards machines that pursue goals independently, automating tasks through a predefined sequence of operations that are still set by humans.
Gael explains: “AI workflows are linear operations. There’s a set of tools and chains of commands you dictate at the outset. With AI agents, you just specify the goal, and the agent pics the tools and determines the steps.”
Gael’s method employs Open AI’s API, Midjourney and Topaz for generating and enhancing images, and an automation platform like Zapier for stringing them together.
And though he’s avidly following news about Agentic AI, for now Gael prefers AI workflows. “They make me less nervous. With agents, you never know what they’ll come back with - like human staff.”
The era of AI agents is here and poised to reshape our world. By adopting AI workflows now, you can harness the transformative power of automation while mitigating the risks of fully autonomous agents.
Joe Smith is Founder of the AI consultancy 2Sigma Consultants. He studied AI at Imperial College Business School and is researching AI’s effects on cognition at Lancaster University. He is author of The Optimized Marketer, a book on how to use AI to promote your business and yourself. Contact joe@2Sigmaconsultants.com.