For a long time, local plumbers and HVAC companies had the cash to buy local TV airtime but hit a brick wall at production. A business might easily budget $5,000 for media, only to be priced out by the five-figure cost of actually shooting the commercial. Because of that math, regional service providers mostly stuck to search and social platforms. Now, generative AI is acting as a direct path into connected TV, obliterating the old barrier to entry for SMB accessibility. Early startup platforms designed to bridge this gap proved that automating the creative process makes television advertising financially viable for a massive new class of local buyers.
Anant Joshi has spent over two decades in global adtech, building international operations for companies like Sizmek, Perion, and Meetrics. He currently serves as Director of Business Development International at Magnite, the largest independent sell-side platform. He recently helped commercialize Streamer.ai, an AI-powered CTV creative generator for SMEs that was acquired by Magnite to scale the technology globally. He views these tools as the great equalizer for local advertisers.
"Smaller businesses can now compete with the larger companies that used to spend millions on advertising and big creative agencies. It just makes it more accessible," says Joshi.
What began as a straightforward tool to generate 30-second spots for local service providers is pulling heavier weight. Today, some larger brands are beginning to test more mature video models alongside smaller advertisers, while platforms like Magnite build agentic offerings that generate thousands of audience-specific, modular ad variants at scale.
In practice, these AI tools automate strategy, creation, and deployment, compressing the timeline for end-to-end campaign management. Magnite's BioAgent represents the leading edge of that compression.
"It is basically an agentic offering to buy media using conversational AI," Joshi says. "We're the first agent that has the ability to do the whole end-to-end campaign management of doing the strategy, creating the campaign, adding the creative, deploying that on media and doing all the reporting."
The workflow impact is immediate. "Previously, something that used to take weeks to negotiate with individual CTV publishers is reduced down to minutes," he says.
The push toward automation also reflects how many in the industry are expanding their definition of CTV formats. For early adopters, these AI systems function as universal visual asset generators, capable of populating different screen layouts like vertical video, tile-based display units, and home screen placements.
Pause ads are gaining traction as an entry point. "What we're finding especially here in the UK with ITV, they're finding good success with pause ads," Joshi says. "When you pause content on CTV, an ad comes up, and that is effectively an image or a static ad, but that is also created using AI. Within that format, you can have the text, you can have QR codes."
Simultaneously, some tools built for the smallest SMEs bypass traditional production entirely by integrating social media feeds directly into the television environment. "To make the whole process of creating the ads in the first place a lot easier, you can actually just take a social feed, like an Instagram reel, and put that inside an ad and run that as a CTV ad," he says.
The speed of this technology is changing how some agencies pitch and resource CTV campaigns. Because campaign generation requires almost zero upfront capital, certain agencies are now pitching fully realized CTV campaigns on spec, resulting in high conversion metrics.
"We have a client in North America which is an agency client of ours. And they're typically doing search and social with their clients," Joshi says. "For every four previews of an ad that they're sending out, they're winning three campaigns on the back of it."
Meanwhile, several major publishers and hardware platforms, including Channel 4, TF1, and LG, are adopting self-serve tools designed to capture omnichannel spend. "Whether you're a broadcaster or a media agency, the idea is to capture that spend," Joshi says, noting the importance of direct agreements with publishers. "Each of those publishers are also coming to us directly because they want their own self-serve."
But letting AI generate ads at scale requires some basic quality control. Many platforms manage intellectual property risk through hardcoded LLM content restrictions and client-owned reference imagery. They also integrate human QA to catch the models' current limitations in rendering physics and object movement. Because of those visual hiccups, human judgment remains a mandatory step in many of these adtech workflows.
"We have to make sure that any reference images are their own reference images that they have rights to," Joshi says. "So we put the ownership back on to the client when it comes to image rights."
The technical limitations are real but narrowing. "One thing that all the LLM models can improve is more of physics dynamics. Objects moving," Joshi says. "When something is generated, it's always going to have a human look over it from a creative point of view, to make sure they're happy with it."
The underlying technology already supports a wide spectrum of creative ambition, from a local plumber's basic promotion to a larger brand's multi-variant campaign. Joshi expects the technical limitations around motion and realism will continue to narrow as computing power advances.
"Effectively, you could create your own Super Bowl ad. And it's really down to your own imagination," Joshi says. "I think soon we'll probably start to see GenAI winning Cannes awards."
