In part two of Performance Delivered: Insider Secrets for Digital Marketing Success, host Steffen Horst welcomes back Navah Hopkins (Microsoft) and Shawn Walker (Symphonic Digital) to explore how AI is transforming paid search strategies.
This episode dives deeper into when to leverage AI, when to stick with manual strategies, and how to think critically about data, budgeting, and brand differentiation in an AI-first world.
When to Use AI — and When Manual May Win
Shawn explains that while AI can improve productivity in areas like creative development, case study editing, and email generation, it’s not always the best fit for every campaign.
For smaller advertisers with limited conversion volume, AI-based bidding may not have enough data to be effective. In such cases, manual campaign management or alternative platforms can produce better results.
Standing Out in an AI-Driven Marketplace
If all advertisers are using similar AI tools, differentiation becomes critical. Shawn emphasizes that success comes from a strong brand identity, unique messaging, and exploring underutilized channels — like Microsoft Ads, Reddit, or LinkedIn influencer ads — to find audiences your competitors haven’t reached.
Navah agrees and adds that advertisers should be intentional about campaign targeting. For example, rather than sticking to oversaturated locations, brands can target by DMA, county, or niche segments to maximize efficiency.
Strategy Shifts: Reporting, Funnel Growth & Brand Safety
Navah notes the biggest strategic shift isn’t just about bidding automation — it’s how advertisers report results. Top-of-funnel and brand-awareness campaigns, while essential for long-term growth, are often undervalued because they aren’t directly tied to return-on-ad-spend metrics.
She stresses the importance of educating stakeholders on why investing in the full funnel matters. Additionally, brand safety is more important than ever. Advertisers need to use content suitability controls to avoid placements alongside inappropriate or sensitive content, even if it means sacrificing short-term profitability.
The Importance of Asking “Why” in Data Analysis
Shawn warns against blindly trusting AI-generated insights or predictive metrics. He advocates for the “ask why three times” approach to uncover the true reasons behind performance changes.
AI can be a great starting point for analysis, but human oversight is crucial to validate findings, catch tracking issues, and interpret nuances that algorithms might miss.
Predictive Modeling: Promise vs. Reality
While predictive modeling offers exciting possibilities, both Shawn and Navah caution against over-reliance. Factors like politics, weather, economic shifts, and interest rates all influence results, and these external variables aren’t always captured in models.
Their advice: Use predictive insights as a guide, not a guarantee — and combine them with human judgment and on-the-ground context.
Budgeting in an AI-First World
Budget allocation strategies are evolving.
Shawn recommends simplifying campaign structures for small budgets to give algorithms more data to work with. For larger spenders, diversifying channels and focusing on less competitive spaces can improve ROI.
Navah suggests planning quarterly budgets rather than monthly, and building in at least 20% extra budget for the first 2–4 weeks of a campaign to clear the learning phase faster. She also encourages targeting no more than one major time zone per campaign to concentrate budget where it’s most effective.
Final Takeaways
AI can be a powerful tool in paid search — but it’s not a magic solution. Marketers need to:
- Know when AI will add value versus when manual oversight is better.
- Differentiate their brand to stand out in competitive auctions.
- Protect brand safety with proactive placement controls.
- Use predictive insights cautiously and always validate with human analysis.
- Budget strategically to give campaigns the best chance of success.
Episode Transcript
[00:00:02.960] - IntroThis is Performance Delivered Insider Secrets for Digital Marketing Success with Steffen Horst.
[00:00:11.120] - Steffen Horst: Welcome back to Performance Delivered Insider Secrets to Marketing Success, the podcast where we explore the strategies, trends and technologies driving today’s marketing results. I’m your host Stefan Horst and today we’re back with part two of our first episode on how AI is redefining paid search strategy. Here with me again are my two incredible guests. First, Nava Hopkins, a marketing strategist who has led digital growth initiatives for both startups and global brands. We had her on the podcast before where we talked about how to grow small businesses through digital marketing. Make sure to check those episodes out as well. She currently works at Microsoft where she collects and disseminates feedback from the advertising community to the product team, as well as shares accessible insights on new Microsoft ad features. Joining her is Shawn Walker, a performance marketing expert with deep experience in navigating platform changes, evolving bid strategies, and integrating AI into campaign planning. Shawn is the Chief Strategy and Media Officer at Symphonic Digital. Together we’ll explore how AI is redefining paid search strategy, what’s truly disruptive and or it’s just adding efficiency.
[00:01:22.690] - Steffen Horst
Now in the first episode we talked about what trends in AI are truly disruptive or disrupting paid search versus just adding efficiency. Whether the shift to AI first campaign management is inevitable or just trendy. And what do the next 12 to 24 months look like for search advertisers as platforms evolve? Make sure to check the first episode out too. Now let’s dive into part two on how AI is redefining paid search strategy. So for new campaigns, all brands starting fresh AI offers an opportunity to build strategies in a completely different way. But should brands start planning with AI in mind from day one?
[00:02:04.600] - Shawn
Yeah, I think there’s some parameters you need to think about before deciding productivity wise. Yes, definitely. Pull in AI. It’ll help you send emails, edit your case studies, you know, anything that you need, it just speeds it up. But as far as launching AI based campaigns, things like that, if you’re a small client and you don’t get enough conversions, you really have to consider if something that requires data to be effective is going to work for you. And I will say we’ve had some challenges in retail. That’s a big one where let’s say the price of your goods is $20 or $30 and it’s five to $10 a click, you’re not going to get enough clicks in order to get a sale. So you really have to sit there and say how much budget do I need to get that statistical significance. And if we can’t get it, maybe we need to consider doing a manual campaign or thinking about other platforms or different tactics, things like that. And I also want to say one thing to go along with this is if everyone is doing the same thing and you have advertiser A and advertiser B and they have very similar offerings and they’re both using AI, dynamic, creative, scraping their website building keywords and everything is one for one who wins.
[00:03:27.960] - Shawn
Right. This is an auction based model to where it if you’re just bidding each other up all the time, what is the thing that is going to help you win at the end of the day? And I think more importantly than just thinking AI first is how do you differentiate your brand? Right. A lot of AI is, you know, going back to retail, a lot of things are commodities, you’re selling the same thing. But do you have a different brand of approach? Is it a different lifestyle, you know, do you have brand ambassadors, things like that? I think people need to think that way and think of a more robust marketing strategy and then start thinking how do we roll this out and can we use different things in AI to help us get the job done? But I think you need to start off with more of a brainstorm of who is our brand, what is our waterfall for getting leads or selling sales and do we have enough budget to achieve what we’re trying to do? And then once you answer these questions, then make some smart decisions based off of that.
[00:04:34.930] - Navah
I fully agree. In terms of account creation or campaign creation, there is a certain degree of privilege that older accounts are going to have because they’ll have that statistical significance, they’ll have those older conversions. It will be easier for them to lean into AI, but ironically they’ll be less likely to because they’ll have those beautiful account structures that they don’t want to necessarily move away from or it’s taking a risk to set up new AI. So it’s kind of like a chicken and an egg situation. For what it’s worth, I tend to push, if you’re brand new, test AI in your B level market and prove out your PPC knowledge and account building and kind of manual with your tier A. And then you can gradually move the AI components that are working well into that tier A mix. I’m not going to repeat all of the brilliant points Shawn made, fully agree with every single one of them, especially around points of differentiation. But there is something to also be said for being where your competitors are not. So I’m not going to shamelessly plug Microsoft as where people are not, although we do actually have over a billion people that interface with our screens globally.
[00:05:51.050] - Navah
But what I will say is that I think the average person defaults to Google search first or to Facebook’s meta or Instagram’s meta first. That tends to be the default. And I would take this opportunity in the AI first world to really consider where can I be where my competitors haven’t thought to be. So if you see that your impression share is 90% for an idea, you don’t have enough competition or there’s not enough volume for that idea. Like there needs to be some awareness put into the mix and if you see that your auction insights are pushing you all the way to the bottom and it’s super competitive, you probably want to look at some other places. So I would definitely encourage folks that are married to Google look for bidding on close variants, look at going after DMA location targeting if you’re in the US or kind of counties as opposed and then exclude the zip code rather than targeting the specific zip codes and getting really specific obviously yes, try Microsoft, also try Reddit. Think about LinkedIn specifically the influencer ad type because that’s a way to also bring a lot of humanity to your brand offering.
[00:07:07.750] - Navah
So everything Shawn said plus be where your competition is not great.
[00:07:12.070] - Steffen Horst
In part one of this episode we already talked a lot about manual bidding versus algorithm driven bidding and how has kind of how it has reshaped how advertisers approach campaigns and management. Now what I would love your perspective on is how has this evolution changed both strategy and or maybe we focus more on strategy because we already talked about execution Navah.
[00:07:36.640] - Navah
So it’s really fascinating. I think the biggest strategic shift is not actually in whether you use auto bidding or not. It’s in how we report on our outcomes. It tends to be really challenging for top of funnel or brand oriented buys or however you want to position it awareness oriented buys to win budget because they tend not to be grounded in return on ad spend metrics but they are undeniably part of the value and the outcomes that come from search or shopping branded, so on and so forth. So the biggest strategic shift I would say is have conversations with your stakeholders, have conversations with your clients and explain to them that if we only go after the very very bottom of the funnel there will come a point where that cohort runs out and you are not going to be able to forever serve these people and make money at a profitable make profitable sales off of these people. We Want to think about growing the cohort so the more that we can lean into tools like demand gen, lean into tools like audience ads, lean into maybe even some exploration with negatives in place with.
[00:08:53.260] - Navah
And this is actually, this is another thing that’s important. So not enough people use brand control or the content suitability tools and that is really critical right now. News placements used to be kind of a no brainer and mobile games were I reject you on site. It’s crazy right now in terms of news like wherever you fall, on whatever side of whatever spectrum, it is crazy right now. And if you are not excluding content that is sensitive or traumatic or bad, your ad will serve next to it. I saw an ad talking about the Texas floods and I’m not going to name who it was, but it was a B2B brand selling B2B services right on top of little girls dying to a flood. Like that’s just. You don’t want your ad serving there. So yes, we need to think about metrics but we also need to have conversations with our stakeholders of I can get you everything but are you okay with your ad serving here? And if the answer is no, I’m going to have to hurt our profitability to protect our brand sentiment, to protect brand equity.
[00:10:07.250] - Shawn
Yeah. I’m going to tack on to two things there and I want to start with the concept of dunning Kruger effect I like to bring in. It’s basically the idea if you have a little bit of knowledge, your confidence is super high and if you have a lot, your confidence goes down. This happens on a daily basis. And the reason this is important is we need to focus on where the bones are buried. Like for example, we had one scenario where we’re running PMAX and the client said, you know, PMAX is doing so much better, then our non brand campaign shut off the non brand campaign. Problem is when you go and look under the hood, Google doesn’t give you all the search terms. You have to put a script in to see what’s coming through. I think 80% of their terms were brand. And here’s the thing, when you carve that out, brand was doing as well as the brand regular brand search campaign. But the non brand that was part of PMAX actually did worse than their non brand campaign which I found super interesting. So when you just see a line item that says pmax you actually have to understand what’s in there.
[00:11:19.990] - Shawn
You know how much of our brand terms are in there. And I’m not saying don’t bid on brand, bid on brand, if competitors, you know, are stealing your market share and things like that. But reporting is so important to understand what am I actually looking at? And a big part of that too is data interpretation. So when you look in the ui, they have these predictive models, they even have like AI based insights and recommendations. To me, they fall a little bit flat on their face because they don’t answer the why. And I gotta say, we need to understand that more than anything. It’s not clicks have gone up 20% or click through rate is better for this thing. But why is that? You figure that out now, all of a sudden you have better insights that aren’t AI based. It’s actually based on reality, period.
[00:12:12.950] - Steffen Horst
Full stop.
[00:12:13.830] - Navah
Well, I actually have an interesting question for you. If Google or Microsoft or Meta or any ad platform told you here’s why your clicks are going down or here’s why X, Y, Z is happening, would you take that at face value?
[00:12:32.720] - Shawn
Absolutely not. We have this process that we always give the team and I’m a big fan of ask why three times eventually get to what’s actually happening. You know, so clicks went up 20%. Why? Oh, we changed the bidding and it’s cheaper now. Okay, keep pushing. Okay, you change the bids. What’s going on? You might end up realizing that either something is broken or you’re targeting different things, or the website changed, a URL was broken before you could get into such a big list of things you have to keep digging through. So I never trust something that spits out. I will say AI is good at starting the journey, but you need to finish it. It helps you understand where to look, but not to understand the interpretation of that data.
[00:13:20.530] - Navah
I apologize, I’m totally hijacking this, but I find that really fascinating. So say you have that question with the conversational reporting in Google. We’ll click on Google just so that it’s not super loaded with me asking. And it tells you the same things that your direct report came back with because you keep asking, okay, so this happened, okay, why did that happen? And then it comes back and it keeps answering. Would you take that information or would you want a human to verify it?
[00:13:52.450] - Shawn
I’d want a human to verify it because I don’t want to get in trouble for something else that happened. You know, there’s been so many times that something else happened. Like I said, like a website was down or somebody changed a button or a form broke and you’re sitting there saying our cost per conversion is bad. Like and the ads we uploaded aren’t working, but in fact your tracking was off or you added a cookie banner. So all of a sudden you’re not tracking this. We’ve seen it all. So you’re never going to be a hundred percent sure, right? But at least if you get rid of some of the details that you know, you answer 90% of the questions, at least you can come into the room with a good answer rather than, I don’t know, yes, you know, or just accept it at face value of this is, this is why this is happening. And then five years from now you realize that, you know, a button was missing on the submit form, you know.
[00:14:48.790] - Steffen Horst
Is that a question of mistrust? And if you have a system that in hindsight gave you the right answers over and over again, would you come to a point where you actually start trusting it and not ask the team to do the analysis?
[00:15:04.550] - Shawn
It would take a lot. I think it comes down to the individual, right? I think everyone’s a little differently. Some people are order takers, some people are critical thinkers. I think you need both. You need to be able to not just accept what’s there. You have to ask, is this the end of the story? Well, why is this happening? If AI can get there? And I see enough cases to where it did the same thing I was going to do, I’ll trust it. But it’s going to have to take a lot more trials because right now I don’t think we’ve had enough trials with it to where I like what it’s spitting out, or I trust it. I trust that it got the metrics right, but I don’t trust the reason why it’s saying it happened.
[00:15:48.440] - Navah
This goes back to what we talked about in the first part of this episode of sharing your information. So in the agentic world, I could see AI 100% answering the question in the way that you, Shawn Walker, would go about solving it. Because you would have conveyed how you think, what you would look for, and it would be taught and trained on your specific processes. So it would be an agent in your account managing according to your style. But, and this is a big but, that requires you passing along all of who you are and how you think into the agent. And I don’t know that the average business is prepared to do that. Maybe we’ll get there in that 12–24, maybe 36 month time frame. Maybe we’ll all still be sitting here saying AI is okay, but that feels like the main question going back to the original question that we’re answering of net new campaign for AI versus leaving it alone or kind of the strategy — how much we’re willing to put our own brains into AI and our personal data in is going to determine how much value we can get out of it, but also determines how much value we bring to the equation.
[00:17:10.840] - Shawn
Yeah, and 10,000 foot view. Just remember we built AI, humans built it, right. So it’s not like a computer built AI. I understand AI comes up with its own methods, but somebody had to program it. So the most important part, and I think even in colleges it might be a good course to have if they don’t have it, is how to prompt, how to interpret data, what to look for, how to train your agent. Because if you train it wrong, you’re going to have a poor virtual employee. So you have to know what to look for. Because even in, let’s say you can continue on with AI, you can know what to look for, but you have to keep digging. So you can still use AI and say I have a theory, can you support this? You know, type that in. Okay, great, that sounds interesting. Let’s dig further. You have to think critically in order to ask the right questions. Because if you ask the wrong question, you’re going to get a wrong answer.
[00:18:10.080] - Steffen Horst
Now predictive modeling is one of AI’s biggest promises. Being able to anticipate results before they happen. But the big question is, and we just talked about trust, how much can we actually trust these forecasts? Are we seeing dependable results or do we still need a cautious eyes.
[00:32:28.630] - Steffen Horst
Well, this was the first episode, part two of I think four episodes that we’re planning on this topic. Make sure to listen to the next episodes as well. Thanks everyone for listening. If you like the podcast, please subscribe to us or leave us a review on iTunes or your favorite podcast application. If you want to find out more about Symphonic Digital, you can visit us at symphonicdigital.com or visit us on @symphonichq. Thanks again and see you next time.
[00:33:03.060] - Outro
Performance Delivered is sponsored by Symphonic Digital. Discover audience-focused and data-driven digital marketing solutions for small and medium businesses at symphonicdigital.com.
.png)
