The Next Evolution: Agentic AI, Autonomy, and the Balance of Human Judgement

Will we see creative uplift unlocked by this kind of automation? Or will it have an impact purely on speed?

Patrick Hann: I think it’s mostly speed, but there are also certain links that you won’t see as a human because you’re not looking for them, because you might have a preset idea of what you’re looking for.

There may be a combination of things that you would never think about that agentic AI can point out to you. But there could be huge failures of people trying out things that the AI would suggest. Again, I think it’s all part of what will come out of this period of AI actually being let loose in the wider community and seeing what works and what doesn’t.

I think there is a big opportunity for pinpointing things that have never been made clear, in terms of campaign insight. Dynamic creative optimisation has been around for a while, but as AI improves, we’re in this big experimental stage where people will be trying more things out. For me, though, the most attractive thing if I were a buyer would be the speed of set up.

David McMurtrie: I would argue that these two elements are interdependent. Despite programmatic advertising being around for many years now, campaign setup, evaluation, and reporting remain time-consuming and inefficient. If AI can enhance the speed, transparency, and quality of these processes, more time can be dedicated to creative input and analysis, ultimately improving overall advertising effectiveness.

This shift will also enable agencies to conduct more efficient testing programs and gain a deeper understanding of the relationship between advertising and audiences. Former CEO of WPP, Mark Read summed this up well in a recent interview: “It is too early to say whether these [AI] moves will lead to more or fewer jobs in the sector. It’s much easier to see all the jobs that AI will disrupt than all the jobs it will create. But it’s going to make creative people even more important because it can level the functional playing field and make the idea itself even more critical.”

Cameron Walters: Creative optimisation is an area that we have identified as one which warrants further exploration for our product roadmap. Our R&D team have already demonstrated the potential of AI generating branded creatives based on client landing pages, reducing the input required from an end user.

Our clients already have access to detailed insight from their designated account manager, but AI could elevate this analysis by offering new perspectives and more immediate insights. Not only would these developments offer more value to clients, but they would leave account managers more time to do what only humans can do, and what humans do best.

Do you see any other areas in the contextual programmatic workflow, like bid optimisation and media planning, that agentic AI could impact?

Patrick Hann: Media planning is the one that everyone’s a bit wary of at the moment in terms of thinking “a robot can’t do my job”, because there’s a lot of nuance to it. Speed of analysis of things, for sure will be an important use. Finding links between data and insights will be a big thing.

AI is in this static kind of world where it has loads of time to think of things, which we don’t, which gives it a far greater level of analysis.

David McMurtrie: There’s always a balance between delivering optimal results and differentiating your media strategy from that of any other advertiser. AI will do the first part well, but it currently falls short in delivering the unique "WOW factor" that sets a strategy apart.

Some advertisers will be quite happy to run campaigns planned and optimised largely by AI. For others, the desire for a truly distinctive approach will always necessitate human creativity.

Cameron Walters: The multi-stakeholder nature of media planning suggests to me that there could be a slower uptake of innovative technologies such as AI. When there are so many different parties involved in the process of curating a media strategy, often with conflicting KPIs and priorities, an element of subjectivity and nuance is helpful. If you were to try and go down the AI route, could you get these different stakeholders to agree on a model they are happy to trust and use?

Our bidding algorithms already use AI models to determine if inventory is worth buying based on contextual relevance and brand safety suitability within a fraction of a second. However, these models are continually under development to ensure that other factors such as time of day, day of week, inventory levels, delivery so far and existing win rates, are used to determine the probability of placing a bid. These additional data points, along with our existing contextual and brand safety considerations, result in our most optimised bidding to date, and ultimately better results for our clients.

Given the power of Agentic AI, where do we draw the line? How should we define the guardrails and level of autonomy we grant AI within a live campaign environment?

Patrick Hann: Giving the keys to a robot is probably quite risky. I’ve worked at managed service DSPs, and you would even be wary of humans going and doing things incorrectly on platforms. So I think the idea of handing over all responsibility to AI can be quite terrifying. The more you kind of synthesise a campaign, the more every part of the internet could potentially become synthetic. Your AI could be delivering ads to other AIs doing other human jobs.

I think the idea of fully delegating to an AI feels intrinsically wrong, but is potentially something that could happen. But a human still has to pick all this up at some point. All this comes back to how comfortable we are letting these decisions be made for us. It goes beyond advertising, but advertising is a very interesting microcosm of it.

David McMurtrie: The question is not whether autonomy is useful but whether the quality of the data used to train models is robust enough to achieve the desired results and avoid the potential risks. There are still significant challenges in the campaign management process regarding the use of data and integration across various platforms. These challenges need to be fully addressed for AI to achieve complete autonomy and dynamism.

Cameron Walters: I believe there are aspects of campaign management that cannot be automated - building genuine client rapport and applying emotional intelligence to navigate complex client scenarios for example - which reassures me that there will still be value in Account Managers. If we are not yet in a position where AI autonomy is possible, but we accept that there is also value in AI, the role and necessary skillset of Account Managers will have to develop.

The key will become defining the parameters and guardrails that AI is able to operate within, as well as providing the final sign-off, until we are comfortable with allowing for more AI autonomy. If this is done correctly, I do not see any reason why it cannot benefit our team.

Do you think people are uncomfortable with it from what you’ve seen at the moment, across advertising companies?

Patrick Hann: I think this industry in general is more experimental with it. Partly because we’ve been exposed to a lot of platforms using AI.

David McMurtrie: There are many different groups within the advertising industry with different motivations, interests and concerns regarding the introduction of new technologies. Agency Management may prioritise financial performance, assessing how new solutions can reduce costs or deliver a competitive advantage, especially against the consulting firms that have entered the market.

AdTech companies are thinking about how they can use AI to improve client outcomes, establish a USP and diversify their business. They are also concerned about failing to innovate, leading to irrelevance in the market.

Employees across all companies face a shift towards specialisation. The automation of routine tasks will reduce entry-level positions, necessitating higher skill levels. This has been happening for a while in other industries, such as automotive manufacturing. A key challenge for employers will be recruiting highly skilled workers if the ability to gain these skills through entry-level positions is greatly reduced.

Cameron Walters: The recently published whitepaper ‘The Big AI Secret’ outlined that 40-45% of businesses that use AI are still at the earliest stages of experimentation, and have not reached the structured implementation phase. AI can oftentimes be banded around as a marketing material buzzword, and the adoption of technologies is more of a knee-jerk reaction to the AI wave than as a result of a serious need. In this scenario, discomfort is perhaps understandable.

On the flip side, because AI and LLM technology have been a fundamental aspect of the SmartMatch™ product and business model for several years, I have not experienced client discomfort when leveraging our technology. There is a trust in our expertise in our proprietary technology, which has been designed for a very specific function, and how this actively benefits their media planning.

The evolution of our product is ultimately driven by the needs and feedback of our partners. To this end, I do not think that there is a discomfort around the use of AI from our clients, but rather a cautious enthusiasm. Conversations with almost all of our partners, whether that is advertisers, agencies or publishers, suggest that there is broad acceptance that creative approaches to campaigns are inevitable, and AI plays a huge role in that. We are receiving more and more briefs that are asking for innovative approaches to advertising, so it feels as though if you do not lean into this, you could be left behind. When you combine this with our engineer’s enthusiasm to innovate, the future of our product is very exciting.

Do you think people are more uncomfortable when they know more about AI than when they know less?

Patrick Hann: I would say no. I think people who don’t know much about it just don’t care to engage with it. I understand the discomfort with it. But again, I think my perception of how familiar people are with it is probably based on the fact that I use it, so I probably feel like everyone’s using it, but actually I don’t think that’s true. I sat in a presentation in 2024 and they claimed 25% of all adults in the UK have tried ChatGPT. That’s increased hugely to now. At IAB UK we’ve built our own internal bot which has been interesting because then you really understand how these things work. You start to see the capabilities of it: it’s not just prompting, you can do so much deep analysis; you can create flow charts for things that would take a long time to work out. It makes life a lot easier. But we still have to have that human element of checking in with it.

David McMurtrie: Google’s founders, Larry Page and Sergey Brin originated the philosophy of being ‘Uncomfortably Excited’. This refers to taking on large challenges which, through a combination of fear and excitement, lead to breakthrough thinking and innovation. This is the philosophy that everyone needs to adopt when thinking about the development of AI.

Cameron Walters: AI has been filtering our spam emails, suggesting new shows to watch, and automatically vacuuming our homes for years, so 99% of people have already adopted AI without realising. I believe that the more people know about how AI already influences our lives, the prospect of future uses of AI becomes less daunting, and the discomfort will dissipate. This is also the case when it comes to advertisers incorporating AI technology, Agentic or otherwise, into their media strategies - the more awareness and exposure we have to this tech, the less there is to be uncomfortable about.

In your view, have there been any promises or claims that feel particularly overstated?

Patrick Hann: Not yet, just because I don’t think anyone has made a hugely bold claim. I’ve seen companies set up a sample campaign quickly at least in a demo environment, and I presume that it’s replicable. But the reality, I don’t know. Until we start to see it being used more widely, it’s difficult to say.

I don’t think anyone's claimed it’s going to completely change everything yet, but it could be that people start to make those claims. I think more of the bold claims are coming at macro level, across all industries. And I have no reason not to believe that a lot of this stuff will happen. I think the rate of progress is already insane.

David McMurtrie: Of course, but then our industry thrives on hyperbole. Looking at the rise of mobile devices, the immediate claims were grossly overstated in the short term but nobody could have predicted their use and ubiquity today. The same is true of the benefits of programmatic advertising and not all of the claims about its efficacy have materialised.

When discussing AI, there’s a tendency to focus on its potential to bring about massive shifts, particularly when thinking about the impact on work practices and employment. This overlooks the less dramatic, but no less profound, day to day changes it will bring to people’s lives.

Cameron Walters: There are not any specific promises or claims that come to mind as being overstated, rather I think there are confusions about the speed and scale at which AI will be adopted and the impact this will have.

David touches on an interesting point regarding the less dramatic day-to-day changes that AI will bring - this is where I see the biggest scope for progress for Account Management. Rather than expecting an immediate and complete overhaul in how campaigns are managed, developments are much more likely to be steady and small, gradually improving the efficiency of minor tasks, fundamentally contributing to an overall improvement.

Do you think that any use cases haven’t been in the spotlight, but that you think have a lot of potential within the industry?

Patrick Hann: We talked a lot about campaign set-up, but I could also see this being very key to campaign maintenance. I think that kind of more elaborate decision making, tinkering around with things, could be well maintained by AI.

David McMurtrie: Following on from the previous question, I think it’s the smaller, more immediate changes that we should be thinking about.

In the advertising industry, consider the implications of integrated systems that seamlessly combine data from every available source throughout the funnel. Another example would be the ability to employ AI to target based on a wide range of contextual signals in real time, something that we’re already doing with SmartMatch™. Reducing transactional friction should be a priority across the industry.

Implementing several small changes across the advertising process, many of which would be insignificant on their own, will ultimately bring about more meaningful change than some of the larger predicted use cases.

Cameron Walters: Again, it is the dramatic changes and use cases that grab the spotlight, but the smaller, ‘less glamorous’ features have already helped our team. One example that springs to mind is the advancement of AI-driven note-taking and document summarising features.

Not only do these features allow you to actively engage with the person you are speaking to because the workload of note-taking has been alleviated, but this also allows for standardised team documents, streamlining communication and planning. When dealing with several accounts concurrently, and interacting with so many internal and external stakeholders, keeping clear and concise records of conversations is crucial to effective campaign management.