A 70% Faster Campaign Still Needs a Good Idea

A campaign can be 70% faster without being 70% better. This week's launches show AI removing setup, outreach, optimisation and measurement work, while the parts that still separate strong marketing from cheap output remain human: customer insight, brand judgement and choosing the idea worth putting into the world.

Two unrelated marketing systems landed on almost the same number this week. Magnite said an agentic connected-TV campaign cut setup time by about 70%, while Wondrlab said its seven-agent influencer system can get campaigns live up to 70% faster.¹² That symmetry is useful because neither result came from asking a model to write more captions.

The most consequential change in AI in marketing is happening around the creative work. Systems are beginning to research, coordinate, activate, measure, check and recommend, which means the production bottleneck is moving. The question for marketers is no longer how many assets a model can generate, but what happens when much of the machinery around a campaign starts running itself.

The 70% number matters

Magnite and Amnet France used buyer and seller agents to identify and activate connected-TV inventory through natural-language instructions. Magnite reported roughly 70% less campaign setup time and a 95% video view-through rate.¹ Wondrlab's WondrAgents attacks a completely different workflow, using seven specialised agents across briefing, creator discovery, outreach, negotiation, contracting, content evaluation and reporting.² In both cases, the claimed gain comes from removing hand-offs and repetitive coordination rather than inventing the campaign idea.

That distinction matters because marketing teams spend a surprising amount of time on work customers never see. The brief needs formatting, the creator shortlist needs researching, the inventory needs configuring, the contract needs routing, the assets need checking and the reporting needs assembling. Removing that coordination can create real capacity, especially for small teams that cannot afford a specialist for every step. It can also expose whether the team has anything valuable to do with the time it gets back.

Clearcast's new iClear product makes the same point from a less glamorous corner of advertising. Teams can upload scripts, images and video for AI-powered pre-screening against the UK CAP Code, with expert human review still available through allClear.³ Compliance is essential work, but few creative teams want senior people spending their day repeatedly checking obvious issues that software can flag first. The useful outcome is not fewer humans; it is fewer human hours spent on work that does not need senior judgement.

This is where many AI adoption stories become more interesting than the headline suggests. A company that saves 70% of setup time has not automatically created 70% more value. The gain only becomes valuable if the recovered time moves towards stronger research, sharper offers, better creative direction or more thoughtful testing. Otherwise, the organisation has simply become more efficient at feeding an average idea into the market.

The agency economics behind that are uncomfortable in a useful way. If a client used to pay for hours of manual campaign assembly, a 70% reduction in setup time removes work that was easy to count and invoice. Agencies then have to prove value higher up the stack through audience understanding, creative choices, experimentation and commercial judgement. Automation makes billable effort a weaker proxy for value, which is good for clients and demanding for service businesses built around labour-heavy delivery.

There is also a management question hiding inside the efficiency claim. When seven agents can move a campaign from brief to reporting, the cost of a poorly specified brief rises because the system can propagate it through every later step. Faster execution therefore increases the value of clear boundaries, approval points and source-of-truth brand information. A vague instruction used to waste one person's afternoon; a vague instruction inside an automated workflow can shape dozens of decisions before anyone notices.

Generation is losing centre stage

For the last few years, the easiest AI demo in marketing was generation. Type a prompt, get a headline, image, email or social post. That was useful, but it also encouraged an unhealthy assumption that the main marketing problem was producing more material. This week's product launches suggest vendors are starting to attack a different problem: the work surrounding the output.

Position² launched Arena around shared client context, campaign history, performance signals, human experts and AI agents, claiming early users save eight to ten hours per marketer each week on routine work.⁴ Earlier in the week Adobe acquired Rilo, a startup that had built agents for competitor research, prospecting, outreach, social content and other end-to-end marketing workflows.⁵ Agile & Co went further in its language, describing Core for local service businesses as an 80% AI and 20% human strategy model that remembers services, pricing, brand voice and previous campaign performance.⁶ The unit of value is expanding from a generated asset to a completed sequence of work.

That is a healthier direction because marketers rarely suffer from a shortage of possible words. They suffer from fragmented information, repeated manual steps, weak hand-offs and context that disappears between tools. A model that writes ten versions of an advert is useful for a moment. A system that remembers the brand rules, knows which offer underperformed last quarter, pulls the right product information and prepares a reviewable campaign draft can become useful every week.

The design choice that matters is where the system stops. Position² says Arena keeps expert guidance and approvals inside the process, while Agile & Co describes its Core platform for local service businesses as 80% AI and 20% human strategy.⁴⁶ That ratio will vary by company, but the principle is sensible: automate the repeatable work while keeping authority over identity and judgement with people. An owner should not need three hours to prepare a week's Instagram content, but they should still recognise their own business in every post.

That is also the logic behind Asteris's approach to Instagram marketing AI. The useful role for software is to work from a business's own media, context and brand cues, prepare strong drafts and reduce the repetitive effort around planning. The owner or marketer still decides what gets published. For small businesses, that division of labour is more valuable than an infinite supply of generic content.

Discovery is becoming machine-mediated

The same shift is appearing on the customer side. John Lewis said AI-agent product searches now account for 2.5% of searches, up from 0.3% a year ago, an increase of more than eightfold.⁷ Semrush separately reported that 57.5% of AI users in a survey of 2,338 US consumers had decided not to buy something because of information from a chatbot, rising to 69.7% among people who use AI at least weekly.⁸ A growing share of discovery and evaluation is happening before a person reaches a brand's website.

That changes what content has to do. A page is no longer written only for the human visitor who lands on it after a click. It may also be read by an AI system comparing products, checking claims, assembling a recommendation or deciding which sources deserve to be cited. Clear facts, credible evidence and distinctive brand information become commercial inputs, not merely SEO hygiene.

JCB's new brief with Sleeping Giant Media makes this shift unusually concrete. The agency's remit includes traditional search, digital PR, thought leadership and AI search optimisation across Europe.⁹ This is a large industrial brand treating visibility inside AI answers as ordinary marketing work rather than an experimental side project. The practical response is not to flood the web with machine-written pages, but to make the existing information about the business specific, consistent and worth trusting.

For smaller brands, the same principle applies at a different scale. An ecommerce company using an Instagram AI content tool still needs the product facts, visual identity and customer evidence underneath the content to be accurate and recognisable. If an AI assistant reads the website, social posts, reviews and third-party mentions, generic volume will not make the brand more memorable. Distinctive source material gives both people and machines more reason to understand what is different.

That requirement changes the relationship between brand content and performance content. Product specifications, FAQs, customer evidence, category expertise and social proof used to sit in separate parts of the funnel, each optimised for a different channel. AI assistants can collapse those sources into one answer, which means contradictions and vague claims are easier to expose. The brand that has spent years being precise about what it sells and who it serves gives recommendation systems better material than the brand that relied on clever copy to patch over weak information.

It also makes consistency more valuable across surfaces. An AI system may encounter a product page, an Instagram caption, a press mention and a customer review before it ever sends a person to the site. If those sources describe the business differently, the machine has to resolve the ambiguity itself. Good Instagram content strategy therefore starts to overlap with good information architecture: the brand should sound recognisably itself while repeating the important facts accurately wherever they appear.

Advertising enters the conversation

OpenAI's advertising direction brings these threads together. CFO Sarah Friar described ChatGPT advertising as what might happen if Google and Meta "had a baby": search-like intent from the current conversation combined with the context a platform may know from previous interactions.¹⁰ Topsort has already announced an integration that lets marketplace sellers add OpenAI Ads as an offsite channel, with products, budgets, dates, countries and catalogue synchronisation managed through its existing workflow.¹¹ The ad is moving closer to the moment where the customer's problem is being articulated.

That raises the standard for relevance. Search advertising can be clumsy and still survive because a keyword provides limited context. Social advertising can be broad because the feed is built around interruption. In a conversational interface, an irrelevant recommendation is more obvious because the system has far more information about what the person is trying to solve.

For marketers, this increases the value of accurate product data and honest positioning. If the model understands the customer's request in detail, the brand cannot rely on a vague claim or a generic lifestyle image to carry the sale. The recommendation has to make sense against the actual need being discussed. More context should create less tolerance for weak relevance.

It also complicates measurement. Comscore has launched measurement for sponsored advertising inside AI chat, linking exposure to traffic, engagement and consumer behaviour.¹² That is an early attempt to connect a conversational exposure with what the person does afterwards, rather than treating the appearance inside the answer as valuable on its own. Marketing teams will increasingly need to measure what happened before the click, including cases where the click never arrives.

Speed only pays when taste survives

The broad pattern this week is not that AI can make more marketing. We already knew that. The more consequential development is that software can now remove large amounts of the coordination, setup, checking, search and interpretation that sit around a campaign. That can make a small team far more capable, but it also concentrates the remaining human responsibility.

Once routine execution gets cheaper, weak judgement becomes more expensive. A poor assumption can be repeated across channels faster, an off-brand message can travel further and a forgettable idea can be turned into hundreds of polished variations. The marketer who knows what should be automated, what needs review and what should never leave the draft stage becomes more important as the machinery gets faster.

This is why the human role should move closer to the decisions that shape everything downstream. Customer insight, brand voice, evidence, creative direction and final approval are not decorative layers added after automation. They are the constraints that stop automation from turning speed into noise. The best AI content marketing systems will make those constraints easier to apply consistently, not try to route around them.

A 70% faster campaign is a meaningful achievement when the time saved returns to the part of the job customers can actually feel. Better ideas. Better evidence. Better choices about what belongs to the brand and what does not. Speed is useful when it buys more room for judgement, not when it simply increases the volume of output.

Sources