Kovva 联合创始人谈用 AI 智能体自动化广告买手的跨平台工作
From Puppies To Programmatic To AI Agents
- 1Kovva 联合创始人 Mimica 称,各广告平台 AI 只优化自家,买手跨平台的工作仍无人自动化
- 2Kovva 智能体接入约 50 个买手常用工具,从 DSP 到 Slack,无需学习新面板
- 3买手通常从每日或每周简报起步,交叉核对后再排期任务,交接复杂流程至少需三到四周
- 4Kovva 把护栏设在智能体之外,所有写入操作需人工批准,Mimica 称打算长期保留该机制
广告自动化正从平台内优化转向跨平台工作流,但落地节奏取决于买手对智能体的信任积累。
- 影响:营销团队可先把对账、QA 等跨平台杂务交给只读简报类智能体;写入操作仍需人工批准,当前定位是辅助买手而非接管投放;交接复杂流程需数周,提效是渐进过程。
- 参考:可考虑先在邮件或 Slack 每日简报这类只读场景试点,验证数据口径后再排期周期性任务;值得盘点团队跨平台的手工操作清单,评估哪些适合先行交接。
- 依据:内容为 AdExchanger 对 Kovva 高管的访谈,接入工具数量、人审机制等均属官方口径,暂无独立验证。
Tuesday, October 27th, 2026 – 1:30 am
SHARE:
Tanja Mimica will be honored at AdExchanger’s 2026 Top Women in Media & Ad Tech Awards. The gala takes place on November 4 in New York City. Join us to celebrate the leaders shaping the future of our industry. Use promo code TWTALKS to save on your ticket.
Every major ad platform is building AI to optimize itself.
But media buyers don’t spend their entire day hunkered down inside of any one platform. They hop around pulling reports from a DSP, a social platform and an ad server, reconciling the numbers when they don’t match, QA-ing campaigns and troubleshooting why a line item isn’t delivering.
And that’s the stuff that nobody’s automating, says Tanja Mimica, COO and co-founder of agentic AI startup Kovva, on this week’s episode of AdExchanger Talks.
“Meta is building AI and getting really good at optimizing Meta, and The Trade Desk is really good at optimizing The Trade Desk,” says Mimica, but the “majority of the work of a media buyer lives between the platforms.”
Mimica, who studied animal physiology in Australia and went into veterinary practice management before landing in ad tech, co-founded Kovva to tackle exactly that.
Kovva has agents that plug into around 50 tools that media buyers already use, including DSPs, search and social platforms, verification partners, Slack and Teams, so there’s no new dashboard to learn.
Buyers typically start with a daily or weekly briefing sent to email, Slack or Teams that recaps the past 24 hours in their accounts. They can cross-check that with what they’re seeing in the platforms, ask the agent follow-up questions and, once it proves itself, start scheduling recurring tasks and handing off more complex workflows, a shift that takes at least three to four weeks.
But buyers are still pretty cautious about delegating anything high-stakes or hard to undo, like setting up a campaign with a large budget.
“Which I think is perfectly understandable,” Mimica says. “As someone who’s trafficked campaigns before, I wouldn’t either.”
Kovva builds its guardrails outside its agents, meaning that the rules and limits are set by people, she says, and an agent doesn’t have the authority “to raise its own budget limit or approve its own change.”
And, for now at least, every “write” action – the term for when an agent changes something inside of a platform, as opposed to just reading data – needs human approval.
“We’re planning on keeping that in place for a long time,” Mimica says.
Also in this episode: Mimica’s journey from animal hospitals to ad tech (including how she used a baby name book to christen her first ad tech startup), the AI holdouts and internal champions who can make or break adoption and pitching the business case for agentic AI beyond just the amount of time saved.
For more articles featuring Tanja Mimica, click here.
来源:AdExchanger · adexchanger.com