Sep 2, 2026 · 33 min · 11 segments
I agents are bringing a new level of automation to ad operations. Aaron Lu, AI Product Lead at Notion, shares how his team is using AI across growth marketing, building agents that automate workflows…
Aaron LiuGuest
Nick LaffertyHost
I wouldn't say how we've measured the success of the ad campaigns has changed, but the rigor in which we can go in one level deeper than what we may have done before has changed pretty dramatically.
And what I mean by that is like, if I've gone through half the day and I've gotten like 10 leads at like a cost per lead of, let's say, $50.
But it doesn't tell the deeper story of the last 12 hours of the day, I didn't actually get any leads and, in essence, have wasted the last half of the day's worth of money.
And so the ability for like an agent to always be on and always be monitoring systems while we are probably not able to look at things for 24 hours straight makes it so that like we can catch these kinds of things and kind of conserve budget where we want to and reallocate that to other areas where like experimental budget might be needed.
Yeah, as someone who's run ads also for a living for the last 10, 15 years, I think this really hits home for me is when you're not looking at the platform, you might miss things.
And so do you have agents that specifically check for, have you had a conversion in the last few hours and then make bid or budget adjustments there?
I've had a couple of agents that do like day to day monitoring and I think like self healing and learning loops are something I'm really focused on right now.
And so, for example, like if we have like an agent that does catch something like that, we should also have another agent on top of that that actually looks and sees like after you have made that change, how is that actually impact the campaigns over time? Um, and so what I'm currently trying to do is like create these self-learning loops across all of ads.
So like not only are we catching these things before they happen or before we waste a ton of budget, uh, but we're also able to monitor the efficacy of that over time.
Um, as well as like make additional suggestions on like other things that we might be missing when we're doing that.
I wouldn't say how we've measured the success of the ad campaigns has changed, but the rigor in which we can go in one level deeper than what we may have done before has changed pretty dramatically.
And what I mean by that is like, if I've gone through half the day and I've gotten like 10 leads at like a cost per lead of, let's say, $50.
But it doesn't tell the deeper story of the last 12 hours of the day, I didn't actually get any leads and, in essence, have wasted the last half of the day's worth of money.
And so the ability for like an agent to always be on and always be monitoring systems while we are probably not able to look at things for 24 hours straight makes it so that like we can catch these kinds of things and kind of conserve budget where we want to and reallocate that to other areas where like experimental budget might be needed.
Yeah, as someone who's run ads also for a living for the last 10, 15 years, I think this really hits home for me is when you're not looking at the platform, you might miss things.
And so do you have agents that specifically check for, have you had a conversion in the last few hours and then make bid or budget adjustments there?
I've had a couple of agents that do like day to day monitoring and I think like self healing and learning loops are something I'm really focused on right now.
And so, for example, like if we have like an agent that does catch something like that, we should also have another agent on top of that that actually looks and sees like after you have made that change, how is that actually impact the campaigns over time? Um, and so what I'm currently trying to do is like create these self-learning loops across all of ads.
So like not only are we catching these things before they happen or before we waste a ton of budget, uh, but we're also able to monitor the efficacy of that over time.
Um, as well as like make additional suggestions on like other things that we might be missing when we're doing that.
The rest of this transcript — segmented and speaker-labeled, so you land on the exact moment something was said
Search every transcript — by keyword, by phrase, or by meaning, across every show Radar indexes
Trends — what is surging across podcasts, measured against its own baseline
Alerts — when a name you follow appears in a newly indexed episode
No account is needed to search Radar.
