By Peo Persson, Co-Founder & Chief Commercial Officer, DanAds

This year at CES, six of us from DanAds were on the ground in Las Vegas – meeting with broadcasters, streamers, platform partners, and agencies from morning to night. And if I had to sum up the CTV conversation in one sentence, it would be this: agentic buying is coming fast, but the industry is still struggling with the fundamentals. 

There was a lot of excitement on the floor – some of it warranted, some of it aspirational. What became clear very quickly is that while AI is accelerating planning and execution, fragmentation, signal loss, and measurement challenges haven’t magically disappeared. In fact, in many cases, they’re becoming more visible. 

That tension – between promise and practicality – is exactly where DanAds sits today.

Digital AI is table stakes. Execution is the real story.

Digital AI? That’s so 2025. The convention floor at the 2026 Consumer Electronics Show (CES) was all about “physical AI” – in the form of robots. A highlight was LG’s CLOiD, an AI-enhanced humanoid robot housekeeper that demonstrated its (leisurely paced) cooking, fetching and laundry-folding capabilities for event attendees. 

While suddenly Rosey from The Jetsons doesn’t seem so far off, it’s also hard to imagine CLOiDs appearing (ominously?) in every household window later this year. The same cannot be said for AI agents in digital advertising, which are proliferating so fast that real-time bidding specs are getting AI upgrades (ARTF) and new protocol standards are being put through their paces. 

Physical AI may be the next frontier, but the digital media and advertising world was deep in discussion about digital AI execution during CES, particularly with CTV. At the same time, issues with fragmentation, audience targeting, and measurement finally came to a head, with industry buyers and sellers increasingly embracing context.

AI everywhere, all the time (especially in CTV)

CES featured no shortage of adtech AI announcements, and a high percentage of them focused on CTV advertising:

  • NBCUniversal debuted a seller agent for both digital and linear inventory.
  • Samsung announced it has packaged AI capabilities into its entire suite of ad products to enhance contextual relevance and improve incremental reach.
  • Disney debuted new agentic planning tools that can reportedly turn RFPs into CTV media plans within hours, as well as generative video editing tools to quickly build a variety of CTV ad assets.
  • Powered by the new Lattice Brain AI, DSP Viant introduced the Outcomes autonomous real-time campaign optimization tool.
  • Yahoo introduced its “Yours, Mine, and Ours” framework, demonstrating how brands and agencies can easily integrate data sources and custom third-party agents with its own native agents.
  • Following its successful pilot CTV campaign with agency Butler/Till through the Ad Context Protocol (AdCP), SSP PubMatic formally launched its AgenticOS cross-channel platform for facilitating agent-to-agent planning, transactions, and executions.
  • SSP Magnite also introduced an AdCP-ready seller agent integrated into its SpringServe digital video ad server.
  • Verification provider and campaign optimizer DoubleVerify unveiled Authentic Streaming TV™, a unified workflow for AI-powered content discovery, reporting, analytics, and, yes, ad performance optimization in CTV. 
  • Reddit introduced AI-powered Max Campaigns, leveraging Reddit Community Intelligence™ to automate buying optimization via real-time impression value predictions.
  • Several agencies introduced or announced upgrades of AI “operating layers,” which effectively sit on top of major LLMs for planning and creativity support. Of note: Omicom announced an updated version of Omni; Havas Media launched the AVA global portal; and WPP launched Agent Hub.
  • IAB Tech Lab introduced the Agentic Roadmap detailing how agentic buying and selling can scale on top of existing adtech standards like VAST and OpenRTB.

What stood out to me and to many of the conversations we had at DanAds, wasn’t just how impressive these announcements were. It was how clearly they reflect a shared industry goal: reduce friction between planning, activation, and reporting. 

But here’s the catch: AI agents are only as effective as the execution and data frameworks beneath them. In CTV, those foundations are still highly fragmented.

Real applications, real questions 

Many of these AI announcements were strikingly substantive. For example, the demo video alongside NBCUniversal’s seller agent launch showed exactly how independent agency RPA would build a buy, using a natural language brief to receive information about inventory availability and then optimize based on desired audiences and outcomes. The video even showed how the approved order would be inserted into NBCU’s order management system. 

Even more impressive, the demo highlighted the ability to buy ads in NFL’s Sunday Night Football – CES re-confirmed that live sports inventory is the hottest for brand advertisers. The first campaign transacted via NBCU’s seller agent is booked during American football playoff games in Q1 2026.

That’s not to say brands and agencies were rushing home from Las Vegas to dump their entire advertising budgets into buyer agents. While the releases and demos were impressive, Razorfish’s Chief Social and Innovation Officer Cristina Lawrence commented that AI agents “aren’t ready for prime time just yet, so we need to think of this as inevitable but incremental.” 

Fortunately, the adtech AI conversation seems to have moved past “It will replace all the humans!” to “OK, this is pretty different, but it’s going to make a lot of stuff easier and faster” – most notably in shortening the time and friction between planning and activation. Still, while the path forward may be clearer, progress will still take time – observers should keep their eyes on how agencies leverage their new high-level orchestrators.

Identity crisis in a fragmented world

Streamlining buying and selling via agents sounds particularly appealing in the heavily fragmented landscape of CTV, which seems to grow only more convoluted and difficult to manage by the year. CTV streamers and platforms have long cribbed from the walled garden playbook to demand advertisers to use proprietary identifiers – part of what TVREV’s Alan Wolk labels “Feudal Media” – which has created challenges in holistic targeting and reporting.

Identity resolution has been a central tool for targeting CTV campaigns across platforms and measuring outcomes, but conversations at CES conceded an overreliance on IP addresses – which have become increasingly unreliable. Recent research from the Coalition for Innovative Media Measurement (CIMM) discovered that CTV IP-address-to-email-address linkages are only accurate 16% of the time; IP-to-postal linkages sit at 13%. Identity resolution providers only agree 2.8% of the time on IP-to-email linkages and 6.4% on IP-to-postal linkages.


Further exacerbating IP address reliability, connected TV manufacturers are switching from IP4 to IP6 addresses. “The era of IPV6 has brought dynamic addressing, carrier-grade NAT, router resets, ISP practices, and IPv6 privacy features that rotate parts of the address,” Amol Waishampayan, Co-Founder at fullthrottle.ai, told Mike Shields of Next in Media. “All of that to mean IPs change too often to function as a durable household identifier.”

Context to the rescue?

Perhaps recognizing the challenges of targeting audiences in such a fragmented landscape, Havas Media is stressing “Context-First Planning” with its new AVA global AI portal for managing AI tools. 

“We are big believers in context-first planning, and that’s important because it allows us to identify growth moments that are more important for a given brand or business throughout a planning cycle,” said Jackie Lyons, Chief Planning Officer at Havas Media Network, told Beet.TV.

And even if agentic transactions can help buyers navigate a terribly fragmented landscape for planning and activation, unifying measurement is another massive hurdle. Context is also playing an interesting role here as SSP Index Exchange announced a partnership with Gracenote allowing CTV impressions to be reported at the individual show or program level. With Spectrum Reach as a launch partner, advertisers buying through Index Exchange can now get full details about show context without a GraceNote subscription.

A sharper picture

CES offered more clarity as to how agentic transactions will operate, particularly in regards to CTV, even if adoption may lag. However, the event also made clear that CTV fragmentation challenges have become more than the industry can handle. AI may be able to solve some of these fragmentation issues from the planning and execution front, but measurement woes are likely to continue to plague CTV.

At DanAds, that’s exactly what we’re focused on. As agentic buying evolves, incrementally, not overnight, solutions like ours will become increasingly important in helping the industry move from experimental automation to scalable, trustworthy CTV performance.

And based on what we saw and heard at CES, that work has never been more relevant.