Samira Bakhtiar, GM of Media & Entertainment, Amazon Web Services (AWS), spoke to tipsheet.ai yesterday about the broad range of AI-related, AWS offerings for advertising platform clients as well as momentum she’s seeing.
Next week, AWS will be highlighting its capabilities at the Cannes Lions ad festival. More information is here.
interview tips include..
- the generative AI product “cake”
- chip sets
- language models
- assistants
- use cases for ad platform clients
- publishers: workflow
- everyone: personalization
- marketers: workflow, creative, targeting
- virtual product placement
- generative AI momentum
- for publishers
- with creative
- video semantic understanding
- five years from now
This interview has been lightly edited for clarity.
how generative AI drives performance for ad platform clients
Ms. Bakhtiar: There are multiple differentiators that AWS can provide to agencies, publishers and marketers who [each] work with us.
Relative to generative AI at AWS, we think of it as three “layers of a cake.”
Cake Layer 1 – Chip Sets
At the bottom layer, we have custom built chip sets that do things like training and inference. These are for our most advanced customers who want to build their own foundational models. We recognize that not everybody wants to do that, or has the capital to do that, so we point them to the middle layer of the cake, if you will, and that’s a fully managed service called Amazon Bedrock.
Cake Layer 2 – AI model access
Amazon Bedrock does many things. One primary attribute is the ability to leverage different first-party and third-party foundational [AI] models.
Regarding our first-party foundational models – which go by the name of Amazon Nova – there are seven different types, many of which we’ll be showing at Cannes.
And then there are third-party foundational models -this is important [to AWS] because we don’t think that there’s going to be one foundational model that’s going to rule them all. We believe customers are going to want to leverage different models for different tasks. This is important, for example, for an agency that cares about the full stack of what they can do to benefit their clients.”
AWS clients are going to want to leverage different models at different times such as Meta’s Llama, Stability AI, DeepSeek, Anthropic, AI 21, Mistral AI, Twelve Labs (to be made available), just to name a few – so that’s the middle of the stack.
Cake Layer 3 – Generative AI Assistants
And then the top level of the stack is a series of generative AI assistants that are either focused on business needs/outcomes or developer needs. For the [ad platform] audience, they care about business outcomes.
gen AI use cases for ad platform clients
Publishers – RFPs
Ms. Bakhtiar: If you’re a publisher and you want to be able to automate RFP responses and make sure that the sales organizations within the “ad org” are being responsive, we can now help with the creation of agentic workflows that will help with the ingestion of the RFP and suggested responses to the RFP, pulling from third-party partners such as salesforce.com without having to have so much human interaction.
The RFP example has been one that many of the publishers I’ve been working with are pretty excited about.
Publishers, Marketers, Agencies – Personalization
If you look at the future of advertising and marketing, I think everybody would agree that the future is going to be hyper-personalized. It’s going to be multi-channel, and the companies who are the most data-driven are going to have the most advantage.
AWS’ ability to help create purpose-built services that can accelerate personalization all the way at the edge is something many customers are focused on.
Marketers – Workflow, Creative, Targeting
A few things.
- We create impact areas for marketers around enhanced retail media solutions and allow for those agentic AI workflows to help really streamline operations and reduce costs.
- We allow for advanced, multimodal advertising, based off of [our] services.
- And, we can help marketing customers work with agencies and really focus on more nuanced audience targeting.
To that end in Cannes this year, we’re bringing back “A’Maison: The House of Amazon” at Cannes Lions festival where all the different lines of businesses across Amazon will be available in one experience: AWS, Twitch, Amazon Advertising, Wondery, Prime Video -just to name a few. (See more Cannes activation details.)
Virtual Product Placement
Vast opportunity for virtual product placement across eSports – animated or computer generated content – is also buoyed by AI, and also across live feeds. So the ability to now have a seamless ‘look’ like it was meant to be across a racetrack, for example, and not look like an overlay is very helpful.
We’re also seeing that virtual product placement being leveraged across different social media channels with different influencers.
There are a series of AWS partners who’ve actually built their own virtual product placement services. And our customers seem to be very delighted with that, and AI is helping create that engagement.
Gen AI momentum: scale, automation, creative
Ms. Bakhtiar: All of these outcomes that I’ve been describing, like virtual product placement, pause ads, playable ads – I think all of these are accelerating scale, automation and creative optimization as a result of AI.
Depending on the business persona that I’m working with, whether it’s at an agency, marketer or publisher, it varies.
Publisher:
If I’m speaking to a president of ad sales, they care deeply about being able to automate sales fulfillment measurement tasks that have historically been manual and asynchronously managed. There’s a big benefit from AI in that capacity.
Creative:
On creative, the ability now is to be able to customize content from the end user perspective and drive engagement.
Video Semantic Understanding:
I think one other thing that reaches across scale, automation and creative optimization is video semantic understanding, i.e. the ability to be able to understand all the metadata associated to live feeds that are being ingested or archival content that has historically been very difficult to do. Previously, you needed a human being to watch the content and tag everything that was being said.
With our AI services – and with some of the foundational and third-party models – our customers can rapidly source the metadata at time of content ingestion, and that allows for so many different things.
Specific to an advertiser, what this can do is [ull out the IAB and Global Alliance for Responsible Media (GARM) taxonomy associated to the content and its suitability for advertisements in certain regions of the world. This reduces quality control and it’s automated. It becomes scalable, because now that content can be determined for suitability in different regions of the world.
And, then there’s the ability for it to become much more contextually aware of what the content is.
So with a brand as an example, you can determine where ads should and should not be shown across both linear and digital platforms with the level of preciseness that didn’t exist before.
five years from now
Ms. Bakhtiar: I can’t tell you exactly what’s going to happen, except that the pace of innovation, the convergence that we’re seeing, and the availability of scalable privacy compliance, cloud services, Amazon DSP services [and so on] – if you’re an agency or a marketer, you’ve got a lot of choice relative to what you can create.
Possibilities become more infinite.
You’re also in a position where you can start to see many of these customers increase their reach geographically as well as across different demographics.
The future is one in which they are much more relevant to the demographics that they’re talking to in a very precise way – and they are able to streamline their operations, reduce costs, and hopefully be in a position to create new new revenue sources.
The future is multi-dimensional and one in which it will no longer going to be sufficient to be good at one one thing. A lot of these companies have to get comfortable reinventing themselves and do a multitude of different things. That’s where AWS and the AI services we provide are going to help.