The rise of AI has given Digital Media Buyers more options when executing their Media Campaigns than ever before. While the initial rhythm of AI does not feel new in digital due to the always present machine learning mechanisms, significant developments in buy side application have changed the space. In this article we will be focusing on AI that is utilized to optimize live campaigns.
Marketers can use a combination of Embedded and Applied AI when delivering campaigns. The evolution of Embedded AI, built into platforms’ own technologies, and the increasing number of companies offering Applied AI, which works independent of platforms, offers new opportunities, but also complicates the space. It is important for buyers to understand both and find the balance between adoption and over-reliance on AI.
The Evolution of Embedded AI In Digital media buying, Embedded AI allows us to use platform algorithms to mass process data and execute media campaigns whilst focusing on the campaign goals. For example, The Trade Desk’s ‘KOA’ tool, or DV360’s proprietary algorithms are examples of Embedded AI within a buying platform.
Embedded AI has traditionally been utilized in bidding processes by Demand Side Platforms (DSPs), which analyse each bid request from Supply Side Platforms (SSPs) and adjust bids accordingly. Proprietary algorithms determine the bid price based on the likelihood of achieving the targeted outcome, and teams can also leverage custom bidding solutions to create more complex methods for assessing the value of each impression.
In 2025, DSPs have advanced to offer more sophisticated buying methods, such as PMAX in Google Ads, Demand Gen in DV360, and Performance+ in Amazon. These features enable buyers to input assets, copy, and targeting parameters into the platform, along with the KPIs. AI is then used to determine the optimal combination of assets to drive the most efficient results. Amazon Performance+ goes further by generating ad copy and cycling through variations to drive incremental performance.
The Introduction of Applied AI
Applied artificial intelligence (AI) is the use of AI to solve real-world problems. Companies such as SciBids, 59A, and Chalice, specialise in the mass analysis and processing of DSP bid stream data (as well as external data sources) irrespective of platform, to help buyers
execute their media campaigns. In principle, this offer increased flexibility and customisation, but there are also considerations, outlined below.
Pros of Embedded and Applied AI
Embedded
- Proprietary algorithms are available in platform and incredibly easy to adopt.
- All buy types are subjected to the same bidding algorithms, consolidating the buying approach.
- DSPs are at the forefront of algorithmic development, and are constantly releasing updated iterations, to be implemented quickly.
Applied
- Real world outcomes can be brought into the bidding algorithm for real time optimisation, increasing customisation and allowing optimisations against metrics that matter to clients.
- A higher volume of data points analyzed at impression level will increase campaign performance if applied correctly.
- The opportunity to create bespoke client solutions, and a unified optimization approach across platforms. Dentsu have partnered with multiple vendors to create custom solutions, leveraging external and proprietary tools.
Cons of Embedded and Applied AI
Embedded
- While each buying platform has its own AI solution, they are not always the most effective. For example, the algorithms may use multiple data points but often do not take advantage of all the available data.
- DSPs use the same methodology for all Advertiser using the platform, creating increased competition. Buyers can only optimize to in-platform metrics, which often leads to lower quality inventory.
- Products such as PMAX and Demand Gen lack reporting transparency, and a clear picture as to where ads are being served. This is being addressed, but slowly.
Applied
- Agencies need to Ensure that Cost Per Outcome incorporates the suppliers’ incremental fee.
- The inclusion of additional out of platform metrics, and complexity, can lead to longer lead times for activation.
- Bringing in additional data points for optimization decreases the interoperability of the scripts, which may lead to delays.
Leveraging The Best of Embedded & Applied AI
It is the responsibility of Digital teams to assess when to rely on AI and how much optimization control to give up. In some instances, where clients are running a single channel, activated on a single platform, leaning more heavily on Embedded AI to execute campaigns can make sense for efficiency purposes.
However, it is not good practice to solely rely on AI decision making. Embedded AI cannot make large, strategic decisions to your campaigns, and only operates within the boundaries set by a digital trader. A combination of AI usage, and trader governance, is required to deliver the best results. Key considerations for both Embedded and Applied AI are outlined below:
Embedded AI: Since embedded AI can be useful to maximize in-platform KPIs, buyers can run tests to understand which platform KPIs are a proxy to business outcomes (e.g. which bidding algorithm generates the highest incremental Brand Lift). Once validated, setting the AI rules to deliver towards those proxy metrics can be beneficial.
We have seen success on Meta by utilizing ADV+, where Dentsu’s ASC Decoder helps buyers understand the true impact of their campaigns, and optimize accordingly (+29% Revenue, +17% ROAS).
Applied AI: These solutions incur a fee, so curate an A/B test with a cost per outcome criteria to showcase the incremental value driven. Additionally. ensure these tests are connected to a real business or behavioral outcome KPIs – the key differentiator of Applied AI is the ability to optimize to metrics outside of the DSP.
Finally, there are some logical steps that can be taken to set campaigns up for success:
1. Understand what KPI can drive real world results for the brand and define which AI models are best placed to optimize towards this.
2. Test the model and assess that it delivers value (inclusive of cost).
3. Periodically retest your solution against new providers or update versions to maximize effectiveness.
To conclude, AI adoption in digital media buying is not a one-size-fits-all approach. Embedded AI offers efficiency and ease of use, while Applied AI enables deeper customization and optimization. To drive the best results, buyers must strike the right balance—leveraging Embedded AI for streamlined execution while integrating Applied AI where additional performance gains justify the investment. Success lies in continuously testing, refining, and aligning AI strategies with real-world business outcomes