Get ready for Agentic AI (wave three of AI) as it will 'change everything'

[editor]

Another new job-swallowing AI technology is on its way and in this article it's the folks in marketing that are getting worried...

Agentic AI, wave three of AI, has officially washed ashore. Agentic doesn’t just predict or create, it acts. It’s a to-do list of sorts that crosses off tasks on its own. Tell it to plan a beach vacation and it doesn’t just give a list of options – it books the flights, reserves the room, builds the itinerary, and mixes you up a margarita. Maybe not the cocktail, yet. But agentic AI works like a hyper-efficient assistant that never sleeps.

If predictive AI is a calculator and generative AI is a storyteller, agentic AI is your everyday handyman that “figures it out” on its own. A take-action tool that breaks the mold of normal human-computer interaction. It’s fascinating, albeit slightly unnerving.

Imagine your refrigerator starting a group chat with your grocery app and your checking account. Based on your purchase history and current budget, it orders your oat milk. Extra creamy and store brand, this time.

Is your ice box wrong to go over your head? Maybe? But you can’t function without a splash in your morning mushroom coffee, so do you even care?

This shift in intelligence raises some big questions for marketers. If AIs are making decisions on behalf of humans, brands quickly need to determine how to capture the attention of the robots.

Agentic AI is action with minimal oversight. It’s wanting that new sofa but needing weeks to scroll through the endless options, compared to AI selecting the perfect one, purchasing it, and prepping for next-day delivery. For now, you can sit back on your futon and binge-watch reruns until tomorrow. AI will ultimately know you better than you know you.

For marketers, turning the consumer decision-making process from days into seconds completely changes the game. If your brand isn’t already in position when AI makes its decision, you’re probably out of the race. In a world where autonomous cars are the future, how do we gain the attention of the engine vs. the driver, so to speak?

And - God help us - autonomous AI marketing campaigns:

It could also facilitate autonomous campaign management. AI systems that can fully manage marketing campaigns by monitoring performance metrics, adjusting game plans on the fly, and reallocating budgets, all without human input. A level of autonomy that frees up the team to focus on creativity and big-picture strategy.

Agentic AI will change everything – including marketing

Agentic AI may soon make decisions on human’s behalf. Piper Dolan at Team Lewis explains how this change might affect the marketing industry.

www.thedrum.com

 

[Voley]

Oh fucking hell, this sounds awful.

 

[Santino]

It sounds unlikely.

 

[May Kasahara]

Can't we just switch capitalism off at the wall?

 

[Shippou-Sensei]

Might not be but a lot of the does everything AI solutions have been this in

 

[zedr]

The billions being spent on this is why we can't all have nice things. I'm sick of working in underfunded schools whilst people waste money on these things which we don't really need.

 

[May Kasahara]

Exactly. I like making my own decisions, thanks, and as boring as going to the shops is, I'd still rather do that than funnel the world's finite resources into making something that does everything for me while I sit on the sofa ingesting content like a lobotomised farm animal.

 

[mentalchik]

Exactly. I like making my own decisions, thanks, and as boring as going to the shops is, I'd still rather do that than funnel the world's finite resources into making something that does everything for me while I sit on the sofa ingesting content like a lobotomised farm animal.

I once upset some of my younger colleagues at work after listening to them all prattling on about getting home delivered take aways etc etc by suggesting that in the future we'll end up as legless blobs with huge thumbs sitting on a sofa not having to move at all hooked up to catheters and iv's watching life on a screen

 

[Miss-Shelf]

will it be able to mark my essays?

 

[Miss-Shelf]

will we be able to marry it? it sounds like it's the wife I've always wanted

 

[RevolutionSound]

This is like the dot com bubble.

 

[Rob Ray]

Is your ice box wrong to go over your head? Maybe? But you can’t function without a splash in your morning mushroom coffee, so do you even care?

I wonder if anyone reading the marketing spiel before it went public thought for a moment about the mountain of dystopian sci-fi writing this line so efficiently summarises. Probably not. But I could very easily see this being submitted for one of those 'one sentence fiction' competitions.

 

[Pickman's model]

I wonder if anyone reading the marketing spiel before it went public thought for a moment about the mountain of dystopian sci-fi writing this sentence so efficiently summarises. Probably not.

sure it passed under the camera eyes of some ai approver

 

[xenon]

Looks like a zeitgheisty variant on the old internet connected fridge trope again.

As much as I can't be arsed shopping I don't want my fridge deciding what I'm having for breakfast.

 

[Rob Ray]

You Are Having Cereal Xenon. You Always Have Cereal. Today You Are Trying Honey Nut Cheerios Thanks To The Good Folks At General Mills Who Own The Auto-Choice Rights In This Machine. Praise General Mills. Praise Britain.

Jaunty Jingle plays: "I'm Backing Britain, You're Backing Britain, We're All Backing Britain Today ... This Message Brought to You By The Patriots At General Mills."

 

[editor]

This is like the dot com bubble.

Sure. Loads of these ventures will fail. But the internet is absolutely everywhere now, and I think it'll be the same for AI.

 

[RevolutionSound]

Sure. Loads of these ventures will fail. But the internet is absolutely everywhere now, and I think it'll be the same for AI.

Yes, I think that AI is overhyped at the moment, and the bubble will burst. This will then mean slower but steady development of AI over the coming decades.

 

[SpookyFrank]

The use case described in that blurb would only require pretty basic computing surely? I'm sure companies use stock control software that can do all of that, and it must have been around for years.

 

[SpookyFrank]

I'm also getting pretty sick of this idea that basic executive function is a terrible burden that needs to be lifted from our shoulders.

The world seems to be targeting all its problem-solving skills towards the people with the fewest actual problems.

 

[Fez909]

The billions being spent on this is why we can't all have nice things. I'm sick of working in underfunded schools whilst people waste money on these things which we don't really need.

If handled right, this could help with some of the issues of underfunding. I'd long argued with my partner (psychologist in the NHS) that AI could probably do the assessment screening at, or above, the level of the current professionals who are doing it.

Here's how it works currently:

Go to doctor, explain your mental health symptoms.
Doctor thinks it sounds like ADHD/autism/whatever
Gives you a screening questionairre
You fill it in, you meet the criteria for an assessment

The local MH trust, who are also underfunded/understaffed/overworked get these screening assessment and have to read through them to decide if they agree with the doctor, and whether it seems likely that this person might have ADHD or whatever. It's similar to the CPS in that respect, where the police give the evidence and then CPS decide whether there's enough evidence to make a case.

So, every day, psychologists and psychiatrists, and other staff, are reviewing screening assessments to decide whether they should give this person a full assessment. All the time they're doing this, they're not doing assessments on those who have been waiting for literally years to get a diagnosis.

If we trained AI on this, we could free up the MH staff to actually help people, rather than just triage them.

And all of this is 'just' classification AI, not generative, or agentic.

I could see a similar use case for agentic AI in the process, too, though. Sometimes the assessment criteria that they are deciding on doesn't have enough information for the MH staff to make a decision either way. So they need to request info from parents, partners, etc., to further determine whether this person needs assessing, or whether they get a diagnosis.

If an AI agent was doing the screening, they could pro-actively request this information to again free up time for the professionals to work through the waiting lists. The agent could combine the supplementary info with the screening tool responses and then determine if they meet the criteria, or perhaps more info is required.

Now, there's lots of risks involved in this kind of thing, and it would require extensive testing and initially human supervision. But in the long run, I can't see how this would be a bad thing. Service users would get decisions faster, get seen by professionals sooner, and the doctors and other staff would spend less time on 'admin' and more on formulation, therapy and treatment.

I'm sure there are many other organizations which have similar processes which could likewise be improved, including teaching. I just know more about the NHS/healthcare as I've had two girlfriends who work in this area, so hear all their complaints. And it's related to what I studied at uni.

 

[Fez909]

The use case described in that blurb would only require pretty basic computing surely? I'm sure companies use stock control software that can do all of that, and it must have been around for years.

Some of it is basic, some quite complicated.

Booking a flight? Checking your bank account? Dead easy...if you know the bank and travel company in advance, and could build something to interact with their APIs. But in theory*, an agentic AI could work with any system without having seen it before. It would be able to browse the web and find cheap deals from any travel agent, and match the transfers with the bus/ferries companies at either end, and so on.

*Agentic AI doesn't actually exist yet. At least not in the way described. There's a few things I've seen/heard of, the most famous of which is Devin, which claims to be an entire software development team. It's not very good, though.

 

[RevolutionSound]

If handled right, this could help with some of the issues of underfunding. I'd long argued with my partner (psychologist in the NHS) that AI could probably do the assessment screening at, or above, the level of the current professionals who are doing it.

Here's how it works currently:

Go to doctor, explain your mental health symptoms.
Doctor thinks it sounds like ADHD/autism/whatever
Gives you a screening questionairre
You fill it in, you meet the criteria for an assessment

The local MH trust, who are also underfunded/understaffed/overworked get these screening assessment and have to read through them to decide if they agree with the doctor, and whether it seems likely that this person might have ADHD or whatever. It's similar to the CPS in that respect, where the police give the evidence and then CPS decide whether there's enough evidence to make a case.

So, every day, psychologists and psychiatrists, and other staff, are reviewing screening assessments to decide whether they should give this person a full assessment. All the time they're doing this, they're not doing assessments on those who have been waiting for literally years to get a diagnosis.

If we trained AI on this, we could free up the MH staff to actually help people, rather than just triage them.

And all of this is 'just' classification AI, not generative, or agentic.

I could see a similar use case for agentic AI in the process, too, though. Sometimes the assessment criteria that they are deciding on doesn't have enough information for the MH staff to make a decision either way. So they need to request info from parents, partners, etc., to further determine whether this person needs assessing, or whether they get a diagnosis.

If an AI agent was doing the screening, they could pro-actively request this information to again free up time for the professionals to work through the waiting lists. The agent could combine the supplementary info with the screening tool responses and then determine if they meet the criteria, or perhaps more info is required.

Now, there's lots of risks involved in this kind of thing, and it would require extensive testing and initially human supervision. But in the long run, I can't see how this would be a bad thing. Service users would get decisions faster, get seen by professionals sooner, and the doctors and other staff would spend less time on 'admin' and more on formulation, therapy and treatment.

I'm sure there are many other organizations which have similar processes which could likewise be improved, including teaching. I just know more about the NHS/healthcare as I've had two girlfriends who work in this area, so hear all their complaints. And it's related to what I studied at uni.

The problem with this is that the screening questionnaires are based on very outdated ideas of autism and ADHD, with arbitrary cut offs on the scoring. My worry would be that if this is taken over by AI, there would be no flexibility to challenge decisions. Computer says no.

 

[ska invita]

The problem with this is that the screening questionnaires are based on very outdated ideas of autism and ADHD, with arbitrary cut offs on the scoring. My worry would be that if this is taken over by AI, there would be no flexibility to challenge decisions. Computer says no.

This is the danger particularly with "benefits" processes.
Hopefully with health it will lead to a next referal by a human to dig deeper
Frankly I trust AI to get it right more than a GP, but my opinion of GPs is very very low.

 

[Voley]

You Are Having Cereal Xenon. You Always Have Cereal. Today You Are Trying Honey Nut Cheerios Thanks To The Good Folks At General Mills Who Own The Auto-Choice Rights In This Machine. Praise General Mills. Praise Britain.

Jaunty Jingle plays: "I'm Backing Britain, You're Backing Britain, We're All Backing Britain Today ... This Message Brought to You By The Patriots At General Mills."

Hi Rob Ray,

It's your friendly AI General Mills overlord here.

We detected a hint of sarcasm in your last message so we'd just like to remind you that Siberia exists and, despite our best efforts, is still pretty empty. Let's keep it that way, eh?

Love, General Mills x

 

[Fez909]

This is the danger particularly with "benefits" processes.
Hopefully with health it will lead to a next referal by a human to dig deeper
Frankly I trust AI to get it right more than a GP, but my opinion of GPs is very very low.

Yep, lots of GPs don't even believe in autism or ADHD. Lots that do are not particularly knowledgable, either. So the referels are poor, and take up bags of time in the screening process I described.

The problem with this is that the screening questionnaires are based on very outdated ideas of autism and ADHD, with arbitrary cut offs on the scoring. My worry would be that if this is taken over by AI, there would be no flexibility to challenge decisions. Computer says no.

If this was done at scale, you could refine the process with the data gained from the referels, cross-checked with the diagnostic outcomes. Spot patterns, tweak the questions, and design the forms to better capture the required information.

And again, that's not agentic AI. That's just data analysis.

In theory, if the sytem was good enough, you could skip the GP step completely and have AI do the initial screening assessment. You could do it in your own time, at any hour of the day, and it could take as long as neccesary. The agent could rephrase the question or clarify the meaning when asked, rather than just do a box-ticking exercise like what happens now.

Some NHS trusts have already run trials with chatbots for minor admin tasks, and the feedback from service users is good. I would love it if i could book an appointment at my doctors via an AI agent, vs trying to get through to them on the phone, or being told that "appointments aren't being taken right now".

I imagine lots of queries could be handled completely by agents, without the need for a doctor at all.

Again, some personal anecdotes: my girlfriend has an under-active thyroid and needs thyroxin, but has had some issues recently so they reduced her dose and she has regular blood tests. She then needs to go back to the doctors who tell her, "your thyroxin is too low so we're going to put you on a higher dose. Here's a prescription, pick it up and try this for a month then come back for more tests".

Why couldn't that be done by an agent? It's so simple, I could do it, and I have no medical training. But it still takes time out of the GP's day, and my partner has to go to the doctors to hear it. She misses work, it takes longer, and it's time that could be better spent doing other things for both of them. And she forgets what they told her because she doesn't write it down and has a poor memory. If it was an agent, she could look back at the previous appointment and see the transcript.

Obviously some people will not want this. I get that. But for those that do - and I think there would be lots - it would be a great option.

 

[ska invita]

Yep, lots of GPs don't even believe in autism or ADHD. Lots that do are not particularly knowledgable, either. So the referels are poor, and take up bags of time in the screening process I described.



If this was done at scale, you could refine the process with the data gained from the referels, cross-checked with the diagnostic outcomes. Spot patterns, tweak the questions, and design the forms to better capture the required information.

And again, that's not agentic AI. That's just data analysis.

In theory, if the sytem was good enough, you could skip the GP step completely and have AI do the initial screening assessment. You could do it in your own time, at any hour of the day, and it could take as long as neccesary. The agent could rephrase the question or clarify the meaning when asked, rather than just do a box-ticking exercise like what happens now.

Some NHS trusts have already run trials with chatbots for minor admin tasks, and the feedback from service users is good. I would love it if i could book an appointment at my doctors via an AI agent, vs trying to get through to them on the phone, or being told that "appointments aren't being taken right now".

I imagine lots of queries could be handled completely by agents, without the need for a doctor at all.

Again, some personal anecdotes: my girlfriend has an under-active thyroid and needs thyroxin, but has had some issues recently so they reduced her dose and she has regular blood tests. She then needs to go back to the doctors who tell her, "your thyroxin is too low so we're going to put you on a higher dose. Here's a prescription, pick it up and try this for a month then come back for more tests".

Why couldn't that be done by an agent? It's so simple, I could do it, and I have no medical training. But it still takes time out of the GP's day, and my partner has to go to the doctors to hear it. She misses work, it takes longer, and it's time that could be better spent doing other things for both of them. And she forgets what they told her because she doesn't write it down and has a poor memory. If it was an agent, she could look back at the previous appointment and see the transcript.

Obviously some people will not want this. I get that. But for those that do - and I think there would be lots - it would be a great option.

Yes agree. thats what happens when you call 111 - someone leads you through questions - I dont think its particularly skilled
And then with diagnosis theres the cases of AI having read all the latest papers and being able to spot things doctors dont
Health is one area I think AI has the potential to be great in.
The problem as always is tech companies and profit motives

(as you say this is nothing to do with agentic ai)

 

[_Russ_]

I'm very anti - AI. it pisses me off the way young Americans in particular have really embraced it for basically trivial tasks.
I know it has some great uses in medicine and such but imo far outweighed by the accelerated push its giving to the general decline of society into a feckless majority lorded over by a few powerfull cunts