If you want to see the most egregious examples of hyperbole on the Internet today, Google the term agentic AI. ‘Revolutionary’, ‘epoch defining’, ‘transcendental’ are just a few of the highly strung descriptions that’ll assail your eyes.
Is all this hyperbole justified? Is agentic AI about to be the thing that makes artificial intelligence absolutely indispensable in every aspect of our lives?
I set out to find out.
Here’s what I discovered.
What is agentic AI?
Developments are now arising so quickly in the field of artificial intelligence that even avid followers of the field like myself are struggling to keep up.
So, you can be more than forgiven for not knowing what the term agentic AI refers to.
What, then, is agentic AI?
The clue lies in that prefix, ‘agentic’ - derived from the Latin agentia ‘doing’.
Agentic AI refers to AI models that have the freedom to autonomously handle tasks.
Instead of having to be continually fed prompts - as is the case with the current generation of generative AI models like ChatGPT or Grok AI - agentic AI models are able to complete tasks under their own volition.
Agentic AI models are able - under their own terms - to decide how much time they spend on tasks, the tools and information they need to complete those tasks, and even (and this is quite remarkable), the ability to leave a task incomplete if they decide they don’t have the capability to complete it1.
Let me distil everything down here.
Agentic AI models are truly transformative in the sense that they can act and complete tasks independently.
Pause for a moment and consider the implications of this.
If agentic AI models are really able to execute and complete tasks independently, then entire swathes of the economy could (theoretically) be automated.
Perhaps you can now see why talk of agentic AI is so maculate in hyperbole.
The (very short) rise and fall of prompt engineering
Within mere weeks of the launch of ChatGPT in November 2022, a whole new profession arose - prompt engineering.
Defined as ‘the process of creating instructions for AI models to create the best possible outputs’, it looked like it was going to be the job of the future.
And, whilst it’s still a very useful skill set to possess, agentic AI looks set to put this nascent profession into an early grave.
It comes down to what’s known as the productivity paradox.
As I’ve written previously, generative AI runs up against the work of Robert Solow.
Solow’s Paradox can be summed up as follows: “you can see the impact of the computer age everywhere but in the productivity statistics”.
I contend that this applies to generative AI.
If you actually care about the end result, the process of creating and entering endless prompts and quality checking the outputs largely (but not entirely) eliminates the potential productivity gains to be had from using generative AI in the first place.
(As an aside, I’ve seen practically no one make this argument about generative AI in the media - but, this could simply be because I’m completely wrong).
Agentic AI potentially eliminates this problem.
If agentic AI can be trained to operate in a fashion that is genuinely autonomous, then it could open up truly staggering gains in productivity.
We’re talking sci-fi levels of abundance and productivity here…
Already, the likes of Deloitte are predicting that ‘in 2025, 25% of companies that use gen AI will launch agentic AI pilots or proofs of concept, growing to 50% in 2027’.
Another agentic AI authority, Mike Finley of enterprise AI firm AnswerRocket has stated that ‘rather than simply making jobs easier as many AI tools have promised, this tech has the potential to entirely automate job functions’. He’s also been quoted as saying that, “This thing can literally replace the way that you do business”.
Agentic AI in the wild
The agentic AI era is already here. Take the tech behemoth that is Google. In December, the company released its Gemini 2.0 model - which it positioned as being the ‘AI model for the agentic era’.
Tech hardware giant NVIDIA has released its own platform specifically for building AI agents called Blueprints.
Salesforce has joined the agentic AI party with the release of Agentforce - a platform that allows users to build and deploy AI agents for multiple use cases.
Interestingly, Salesforce is marketing its agentic AI platform as being ‘digital labour’ and the ‘digital labour platform’.
Which brings us on to the next possible scenario for agentic AI…
Agentic AI: your new digital workforce
Agentic AI models are set to be incredibly valuable.
If a company like Salesforce is able to create and deploy AI agents that excel at completing routine business tasks, then they’ll be worth billions - perhaps even more than that.
In fact, as agentic AI models progress, improve and become more affordable (from a computation and energy point of view), we are likely to see agencies offering ‘agentic employees for hire’.
Imagine if you could ‘genuinely’ outsource tasks like resource management and allocation, compliance checks, the production of statements of work etc, how much would you - as a business - be willing to pay?
This is the big bet that AI companies are making.
The likes of Google and Microsoft have shelled out billions on the data centres necessary to power their AI models.
Now’s the time to make a return and ‘sweat’ those assets to the max. Agentic AI - in the way that generative AI has failed to do - looks like big tech’s big revenue making opportunity.
Let’s be honest. Generative AI is a fun tool to play around with - you can use it to generate some text or a weird-looking image. But, it’s not really that useful (not when you consider the sheer amount of money that has gone into developing and maintaining it).
Agentic AI, on the other hand, represents a genuinely utilitarian application of artificial intelligence - one that could upend the world if it lives up to expectations.
We’re all celebs in the agentic AI universe
So far, so good. Businesses could potentially unlock never before seen productivity gains with agentic AI.
But, what benefit will agentic AI have for the average consumer?
Well, I’m going to ask you to use your imagination again.
I’m also going to speak rather bluntly.
Speaking as a middle-aged man, booking a restaurant reservation is - quite frankly - a pain in the backside.
The same goes for booking train or plane tickets for a long-distance journey.
As for shopping for clothes or food - well, I don’t want to plunge into the depths of expletives.
I’d much rather be spending my time doing something enjoyable like reading a book or foraging for mushrooms (your idea of enjoyment is likely to differ from mine).
Imagine if you had an agent - rather like Hollywood celebs do - who could take care of these tasks on your behalf.
That’s the sparkly promise of agentic AI.
Agentic AIs - especially if they become embedded within consumer electronic appliances and devices - could be used to shop on your behalf, book services for you, set up meetings and arrange calls with no intervention from yourself.
Yes, you’d have your own personal agent (or butler, if you prefer).
Can you see why technologists are so excited about agentic AI?
In the shadow of Schumpeter
If agentic AI lives up to the promises of its creators. If it’s able to truly act autonomously as a human being would, then we’re about to enter into a time of massive societal and economic upheaval.
It’s this aspect of agentic AI that you’re going to read about in the media.
The spectre of massive job losses, automation of vast swathes of the economy.
It’s these topics that’ll dominate the news headlines.
But, is it all doom and gloom?
Not if you ascribe to the theories of Joseph Schumpeter.
Who is he?
Schumpeter was a relatively little-known, but hugely influential economist - primarily for his formulation of the concept of ‘creative destruction’.
Also known as ‘Schumpeter’s gale’, this theory suggests that technological development engenders a process of destruction and creation within the labour market.
As new technologies emerge, older technologies become obsolete, thus destroying jobs in the process. Yet, as the new technologies become established, whole new types of job roles are created - creative destruction.
My own job role is a case in point.
Look back 30 years ago, and the idea of an ‘SEO-content writer’ didn’t even exist.
When the Internet emerged and search engines gained traction, traditional writing-based jobs such as newspaper journalism began to wither on the vine - destruction.
But, as this was occurring a whole new type of job emerged - the writing of content to improve the rankings of websites on search engines - creation.
We could be about to see creative destruction on a scale we’ve never seen before (again, with the caveat that agentic AI actually lives up to the hype).
Implications for brands and marketers
Naturally, if you’re reading this, then you’re probably wondering what agentic AI means for your brand (or your job as a marketer).
It’s something I’ve been pondering for a few weeks; and I’ve arrived at what some may seem is a counterintuitive conclusion.
Not a lot.
Wait, what?
I’ve just spent over a thousand words describing how potentially transformative agentic AI could be - and I’m now saying it won’t change things for brands or marketers?
That’s right. Let me explain…
If agentic AI reaches its full potential - and is able to operate in a similar fashion to a human - then brands simply need to adhere to digital marketing best practices.
I’ll try and provide an example.
A human consumer picks up their phone and asks their AI agent to find them the ‘best laptop with an RTX 4090 graphics card’.
What does the AI agent then do?
It’ll navigate to a search engine and type that query into it. It will then fetch the top link that appears in the results and return it to the user.
If you’re already running a technically sound website that ranks well in search, then you’re going to be found by AI agents.
Again, this contention hinges on the notion that AI agents are going to be able to operate in a similar fashion to how a human would.
If that contention is true - then brands and marketers need to continue what they’re already doing - optimising their websites. It’s just that they’ll have a new audience in addition to their existing one - AI agents.
If, however, it is discovered that agentic AI has certain quirks or algorithms that can be gamed - then all bets are off.
Advice to brands and marketers
It’s such early days regarding agentic AI, that it’s somewhat foolhardy to profer any solid advice.
But, as I’ve contended above - if AI agents end up being deployed to consumers at large, then these agents are likely to use existing marketing channels in order to provide actions on behalf of their users.
So, you need to make sure you are as visible as possible on those marketing channels. This means that you:
-
Need a technically sound, fast website that won’t put up any barriers to AI agents.
-
That website needs to be easy (for both humans and AI agents) to navigate through and make purchases on.
-
You need to be the first brand an AI agent encounters on search engines in both organic and paid contexts.
Guess what? Velstar can help you achieve all those things and more so you’re ready for the drawing of the agentic AI age.
Want to find out more? Let’s chat…
Footnotes
1. This has an important implication. Within the current generative AI models, said models are compelled to provide an answer - which often leads to the problem of ‘hallucinations’, where the AI makes up information out of whole cloth, simply so it can ‘achieve’ a completed response to a prompt. Ironically, providing AI models with the agency to decide as to whether or not they can complete a task is likely to reduce this hallucination problem. It will, however, increase the likelihood of AI models stating that they cannot complete a task - how human!