Start ’em, sit ’em is back, with Craig Leshen, President of adops.com, once again in the interview chair. It’s time to drill down into which aspects of adtech deserve real focus in H2 2026, and which ones should stay on the bench until the plan, or the market, is ready.
For the latter half of 2026, it looks like the strongest plays will help teams use AI sensibly, package their audience signals more clearly, build credible video and CTV products, and reduce dependence on channels they don’t control.
Craig: Publishers are making decisions in a market that feels less predictable than it did a few years ago. Search traffic is under pressure, buyer expectations are sharper, and more of the audience journey is happening inside platforms. For example, Reuters reported that social media and video networks are now go-to news sources for 54% of respondents globally, ahead of news organizations’ own websites and apps at 51%.
At the same time, publishers are being pushed to experiment with AI, streaming, and contextual signals, often without enough operational space to do everything well.
Some ideas are interesting, but they aren’t ready to be treated as core priorities. Others sound exciting in a pitch deck, then fall apart when someone asks who owns the workflow after the kickoff call.
Craig: Yes. AI is a Start ’em when it’s practical, contained, and tied to work the team already understands. In ad ops, the best uses are usually the ones that help a campaign operator reach a better first judgment faster, whether they’re checking campaign quality, reviewing performance changes, or trying to understand why something has started to drift. AI should support that work, not take ownership of it.
The audience side is moving, too. Use of AI chatbots for news rose to 10% (from 7% last year), rising to 16% for under-35s. That doesn’t mean publishers should hand strategy over to AI, but it means teams need to understand where AI is already entering the news journey, then decide what role it should actually play.
Internally, AI use definitely requires human review. It can make good operators faster, but it doesn’t fully understand the commercial and technical context a publisher’s team carries around every day. A recommendation might look efficient on the surface and still create trouble later, so someone with real knowledge of the business still needs to catch that before it becomes a real problem.
The best use cases are the ones where success is easily measured. If an AI tool saves time on a report, or provides monetization improvement insights, the team can compare the result with what it would normally produce. If it helps summarize troubleshooting notes, an experienced operator can quickly see whether the summary is useful or misleading. That also makes adoption easier because the team isn’t being asked to trust a black box with something vague.
Craig: A publisher shouldn’t begin with a broad instruction like, “we need an AI strategy,” then go shopping for something impressive. It’s better to ask where the team is losing time, where quality checks are inconsistent, and where a tool could help without taking control away from people who know the business.
Training is part of that. Giving people an AI tool without guidance is a bit like giving them a complicated spreadsheet and assuming they’ll automatically build a useful model. Some people will figure it out, but a lot of the value will be lost because the organization hasn’t explained what good usage looks like.
AI adoption isn’t finished when the license is bought. Publishers need guidelines, review steps, escalation rules, and a clear owner for each workflow where AI is being used – and then make sure you actually use it. Otherwise, the tool becomes another thing everyone is expected to use, but nobody is really responsible for improving. We’re all busy ad ops people, but we should try adopting newer “quality” tech (not just any tech or AI tool), along with utilizing current (modern) best practices, and not just remain comfortable with what’s worked for us in years past.
Craig: Intent and contextual are still Start ’ems because they help publishers turn the content itself into a stronger commercial signal. The real opportunity is to understand what a reader or viewer is likely trying to do in that moment, then shape that context into something buyers can understand.
That’s where publishers have an advantage. They know why certain content performs, which environments are trusted, and where an ad can show up without feeling forced into the experience. That knowledge becomes more valuable when it’s translated into clean taxonomy, clear packaging, and reporting that connects the context to the buyer’s goal.
But the signal has to be parsable for buyers. They need to be able to understand the offer, compare it with alternatives, and see how it performed.
The ad ops work can affect whether that promise survives the handoff from sales to activation. If a seller promises a high-intent audience environment, the platform setup has to match that promise. If the deal is built around a specific content signal, the reporting has to show whether that inventory actually ran as planned.
Craig: It is for the right publishers. Audience behavior has shifted, and buyers are putting serious money into digital video and streaming environments. Nielsen’s report found that streaming represented 47% of US TV usage, while IAB projects that US digital video ad spend will surpass $80 billion in 2026, which would account for more than 60% of total TV and video ad spend.
That gives publishers a real reason to look at CTV, especially if they have credible video, live programming, sports-adjacent coverage, or a strong audience niche. But CTV shouldn’t be an experiment. It needs to be treated like a product, with clear inventory rules, buyer expectations, brand safety controls, and measurement plans. A publisher that simply says, “We have streaming inventory now,” won’t get the same value as one that can explain the proof points behind the offer.
Measurement is where many teams get caught out. Buyers want outcome accountability, but CTV still has a lot of fragmentation, technical gaps, and inconsistent identifiers across the market. Live and sports-related content can raise the stakes further because execution becomes more sensitive in real time.
Craig: It does, and the case is getting stronger. We should actually shine a spotlight on this, so bright that you should feel like you’re staring at the sun. Publishers can’t afford to depend too heavily on one traffic source or platform relationship, especially as organic search becomes more uncertain. But diversification doesn’t mean every publisher needs to do everything. For one publisher, the right move might be stronger direct-sold packages. For another, it might be registration, newsletters, or a paid product that fits the audience. The point is to build revenue around the audience relationship, rather than waiting for traffic to arrive through channels the publisher doesn’t control.
Audio deserves a place in that conversation, especially for publishers with communities or specialist coverage that people want to spend time with. It can be a practical extension of the audience relationship when the format fits the brand and the commercial model has a clear owner.
Subscriptions need the same careful treatment. Paid products can work very well when the brand, content, and audience habits are strong enough. But treating subscriptions as a universal fix ignores how hard it is to convince most people to pay for news or information online. Keep in mind that the share of people willing to do this has stayed mostly flat since last year, at around 17%.
Craig: Sit the idea that leadership can buy a tool, roll it out across the company, and suddenly replace staff expertise. That kind of thinking creates unrealistic expectations inside the business and makes people more resistant to the genuinely useful parts of AI.
The mistake is treating AI like a finished employee instead of a tool that still needs direction. If you ask it to build a workflow or recommend decisions, someone still has to understand the work well enough to check what comes back. Without that ownership, the publisher may save a few hours upfront but create a larger mess later. So don’t sit out training, governance, maintenance, and judgment - all the things that should be done by a human.
Craig: A persistent placement identifier could help buyers understand the same ad placement across different paths, which would support transparency and make buying decisions cleaner. The problem is that real-world implementation has been uneven. If IDs are missing, unclear, duplicated, or changed dynamically, the signal becomes much less useful.
So the advice is to treat GPID as basic setup work, not a star player. If a publisher uses it, the signal should be persistent, descriptive, and easy for buyers to interpret. But publishers shouldn’t present GPID as a major growth lever.
Craig: Sit anything that depends on vague promises and weak ownership. If a vendor can’t explain what changes in the ad stack, the publisher should slow down. If nobody can explain who owns the workflow after launch, the publisher should slow down again.
The same test applies to measurement. A new initiative should come with a clear view of what success looks like, how the result will be checked, and what the team will do if performance doesn’t show up, and it needs to be backed up by data. Show me the numbers! Otherwise, the project risks becoming another interesting idea that takes up time without changing the business.
Ad quality belongs in this conversation too. Poor ad experiences can damage trust, weaken engagement, and create brand safety headaches for publishers and buyers. If a new monetization idea improves short-term yield but makes the site or app feel worse, the publisher needs to ask whether that trade-off is really worth it. Experienced ad ops people are vital here, since they know where a nice commercial story can break during implementation.
Craig: The first starter is practical AI that improves work the team already understands. The second is better intent and contextual packaging, because publishers need to make their own content signals easier to buy. CTV also deserves attention where the publisher has a credible product and the measurement plan is strong enough to support the sales story.
And again, shining a spotlight on revenue diversification, as it should be kept front of mind, too. Revenue diversification leads to revenue growth. Publishers need a broader base of revenue and a closer relationship with their audiences.
The bench is for AI hype, GPID overconfidence, subscription absolutism, and any strategy that can’t survive basic questions about ownership, measurement, and execution.
To learn more about how Craig and the team can help you turn the right Start ’ems into real operational progress, while keeping the wrong Sit ’ems from draining time and budget, contact us today.