Ep. 143: AI-Powered Content Automation, Creative Personalization with Andrew Swinand

Full analysis

Your draft

Andrew Swinand, CEO of Inspired Thinking Group (ITG), argues on the AdTechGod podcast that the real money in AI isn't the flashy stuff. Not generating new images and copy. The real opportunity is the boring stuff: using AI to tag, organize, and reuse the creative assets brands already own. The implicit decision for operators: where do you place your AI bets? On generative novelty, or on the plumbing that makes creative findable and personalizable at scale?

Reversibility: Type 2 for most operators. This is a where-to-invest-attention call, easily adjusted. For agencies restructuring around it, closer to Type 1. What's actually being decided: Whether creative infrastructure (asset tagging, dynamic creative optimization) becomes a competitive layer worth owning, or a commodity that gets absorbed into platforms and in-house teams. Forcing function: None worth naming. This is a directional signal, not an event. Agency restructuring is already underway at the holding companies.

The Market Analyst

This is a vendor telling a self-serving story that happens to point at a real gap. Strip the ITG pitch and the durable insight is the mismatch Swinand names: Google offers 55,000 targetable attributes, Meta 22,000, and the average advertiser runs 14 creative executions against them. That gap is where dynamic creative optimization lives. DCO automatically assembles ad variations from asset libraries, and it's underbuilt. For public ad-tech, this cuts two ways. The Trade Desk, Magnite, and the SSPs monetize targeting depth that advertisers can't fully use. Whoever closes the creative-side gap, whether Adobe, Salesforce, or an acquisitive DSP, captures spend that's currently wasted. In plain terms: the industry has built a Ferrari engine bolted to bicycle wheels.

The Skeptic

The whole thesis pivots on "60 to 70% of efficacy is the creative, every MMM ever." That's a rhetorical cannon, not a finding, and the episode's own fact-check flags it: the range swings wildly by category, and lower-funnel campaigns lean on media and targeting. The 3× response lift, the 80%-of-assets-never-touched, the 60%-never-used stats: all self-reported by the company that sells the fix, with no controls, no client names, no methodology. In plain English: the guy selling umbrellas is telling you it's about to rain. The underlying direction is real. Creative matters, DAMs are a mess. The magnitudes are marketing.

The Operator

Try to act on this Tuesday and the first thing that breaks is your data layer. Meta-tagging 33 million assets into 160 million descriptive phrases sounds clean on a podcast; in practice it means reconciling inconsistent taxonomies, rights and usage restrictions per asset, and brand-safety rules across 53 languages. The 90-day surprise: your AI tags are only as good as your worst-labeled input, and legal will freeze half the library over licensing and talent-usage terms. Swinand's "20 meaningful breaks" idea is the genuinely useful, executable nugget here. The pet-food client where dog size and life stage mattered but dog ownership didn't is a good illustration. Most personalization fails because teams chase one-to-one targeting instead of finding the handful of segments that actually move engagement.

The Customer / End User (the brand / CMO)

From the brand seat, the appeal is real and boring in the good way: cut production costs roughly in half, campaign lead times from eight weeks to two. That's a P&L line a CMO can defend. But there's a strategic tell. Swinand predicts brands consolidate production in-house on platforms like his, and specialist shopper/social/digital agencies collapse. A brand hears: fewer vendors, more control, lower cost. An agency hears: your production revenue is being platformed away. The uncomfortable question for the buyer is whether "in-house on a vendor platform" is really independence, or just a new dependency with a different logo.

The CFO

The economics only work if the tagging investment amortizes across enough reuse. Spending to classify 33 million assets when 80% were never touched is only smart if the newly-findable 20% drives incremental output. Otherwise you've paid to organize a junk drawer. The right frame isn't cost-per-asset; it's cost-per-used-asset. Swinand's real financial argument is sound: the ROI is in workflow and reuse, not generative showmanship. But "50% cost reduction" with no baseline is a number I can't bank. Before committing budget, I want the reuse rate before and after, and a payback period measured against the platform's annual license. Not against a one-time hero campaign.

The tensions worth watching

1. Moat vs. commodity. The Market Analyst sees asset infrastructure as an acquisition-worthy competitive layer. The Customer sees it as something brands will pull in-house. Both can't fully win. If it's easy enough for brands to own, it's not much of a moat for vendors.
2. Creative primacy vs. targeting primacy. Swinand says creative drives 60 to 70% of results. The Skeptic says that's category-dependent hand-waving. The whole "invest in creative infrastructure over media optimization" thesis rests on this contested number.
3. Efficiency vs. dependency. The CFO likes the cost story; the Customer worries that "in-house on someone's platform" trades an agency relationship for a platform lock-in.

What it hinges on

Two beliefs. First: does structured, object-level asset metadata actually become the prerequisite for agentic creative workflows? If yes, DAM and tagging infrastructure gets strategically valuable and acquisitive interest follows. Second: does the creative-to-targeting gap get closed by platforms (Adobe, Salesforce, the walled gardens) or by independents? The council leans toward the gap being real and the direction being right. Creative infrastructure is underbuilt. But there's heavy skepticism on ITG's specific magnitudes and on any single independent owning the layer. For an operator: the strategic signal (tag your assets, find your 20 meaningful segments, close the DCO gap) is worth acting on. The vendor claims are worth discounting hard.

What to verify before committing: your own reuse rate before and after tagging; whether your DSP/agency can execute DCO against a properly tagged library today; and the licensing/rights status of the assets you'd be reactivating.

Prediction: By the end of the holding companies' Q2 2026 earnings cycle (reported through early August 2026), at least one of WPP, Publicis, or Omnicom will publicly emphasize AI-driven creative production and asset automation as a named efficiency or growth pillar in its earnings commentary.

Confidence: Medium. The restructuring toward production platforms is already visibly underway at the holdcos.

Why: The episode's most grounded claim isn't ITG's. It's that specialist agency silos are collapsing into integrated brand-plus-production models with less headcount, a pattern the summary notes is already visible at WPP, Publicis, and Omnicom. These companies are under real margin pressure and have every incentive to tell investors an AI-efficiency story; production automation is the most concrete version of that story they can point to. The opposite outcome, none of them foregrounding creative AI production on an earnings call, would require them to ignore the single most investor-friendly narrative available to them right now, which is unlikely given how loudly each is already courting the AI-efficiency theme.

Revisit by 2026-08-31: We're right if any of WPP, Publicis, or Omnicom names AI-driven creative production or asset automation as an efficiency/growth driver in Q2 2026 earnings materials or the call. We're wrong if none of the three does.