A recent Digiday piece on influencer agencies and the shift toward automation platforms argues the industry is undergoing real structural change. The pattern it identified (platforms raising capital, sophistication moving from service to software) tracks with what we're seeing.
"The influencer marketer's role is dead," Jonathan Chanti declared in the piece. "All the jobs they used to do—find people, curate them, reach out to them, nurture them, report on them—you can build an AI that does a better job of that right off the bat."
Ten years managing Delta Rae through Warner Bros., three at Religion of Sports, two and a half building Basa. The directional thesis could be right. The question is what automation actually looks like once you're in production, and whether we're building toward the foundation that makes it possible or just automating the easy parts while the harder infrastructure problems go unaddressed.
What "Human in the Loop" Actually Means
The phrase "human in the loop" has become ubiquitous. The question is: doing what?
A brand wants to activate two hundred creators for a trending moment. Discovery platforms surface candidates instantly. AI drafts outreach and suggests rates. Then the human part starts. A creator's manager says the rate is too low and they only do sixty-day exclusivity, not thirty. Legal flags usage rights concerns. The creator delivers early with extra content someone needs to review. Their manager mentions expanding into a new vertical. Worth exploring?
Email threads branch. WhatsApp conversations happen outside the system. Spreadsheets track responses. Slack holds approval status. Contracts go through DocuSign and get saved as PDFs somewhere. The intelligence generated (what rates closed, which terms caused delays, what delivery patterns emerged) scatters across tools. Most of it never becomes structured data.
The Data That Exists Versus The Data We Can Access
The automation confidence assumes the data problem is solved. It's not. The bottleneck is what data exists in accessible form, and most of the data that matters doesn't.
Discovery platforms pull from social media APIs. Everyone has the same follower counts, engagement rates, demographics. This intelligence is commoditized.
The differentiated intelligence lives in scattered communications. What you paid, what terms you negotiated, which creators over-deliver, which managers respond quickly, what contract structures work. That data sits in email archives, WhatsApp messages, PDFs on desktops, departing employees' institutional knowledge.
AI can't analyze data that doesn't exist in structured form.
Jamie Gutfreund saw this: "Access isn't the advantage anymore. Data is." She's right. The question is which data. The performance metrics everyone can access, or the operational intelligence nobody's capturing?
Why Consolidation is Harder Than It Looks
The natural response: build one platform that handles everything. Discovery, outreach, negotiation, contracting, performance tracking.
This runs into a problem that's behavioral, not technical.
People use different tools for different contexts: email for formal negotiation, WhatsApp for urgent coordination, Slack for internal approvals, DocuSign because legal requires it, spreadsheets for tracking what platforms don't handle, discovery platforms for sourcing intelligence.
Each tool serves a purpose. The fragmentation reflects how workflow actually functions. "One platform to rule them all" fails because human behavior naturally fragments across contexts. The better question is how to consolidate the intelligence those tools generate. The tools themselves are going to stay fragmented.
You need to know what performed well, what you paid, what terms you negotiated, what delivery patterns emerged, whether those patterns scale. Right now those things sit in different places, captured in different formats, owned by different people.
The Behavioral Complexity Nobody's Accounting For
There's an assumption embedded in the automation narrative. Micro and nano creators can be automated because they have less leverage. No managers, no agents, more need for the income. So the deals are supposed to be simpler and more standardized.
Are we sure?
When you're activating two hundred micro-creators instead of twenty mid-tier influencers, coordination complexity explodes. Each has their own communication style, schedule, comfort level with legal language. And risk compounds: with two hundred creators, you're exponentially more likely to encounter someone who doesn't understand exclusivity, misses deadlines, delivers off-brand content, or has a controversy emerge.
We saw this building Basa. A teenage creator's parent joins calls about image rights worried about college applications. A micro-creator reviews terms on their phone between work shifts with questions about unfamiliar language. Someone delivers technically compliant content that's tonally wrong in ways automation wouldn't flag.
The Spectrum of Complexity
Complexity changes by tier. Mid-tier creators often have managers but not full representation, which leaves a real question of who approves contracts when those people disagree. Macro and mega influencers have agents, managers, lawyers, business managers, publicists. A single deal might require approval from four people with different priorities.
Each tier has its own choreography. Micro-creators need hand-holding but decide quickly. Mid-tier moves faster on communications and slower on approvals. Mega influencers have sophisticated teams that negotiate every clause. Creative production varies the same way. Nano creators deliver on phones. Macro influencers have content teams. Celebrities involve photographers and stylists.
Then there's the brand side. Straightforward posts need legal and brand approval. Complex partnerships (video series, product claims) require legal, brand, product, PR, sometimes executive sign-off. Each layer adds friction. When something goes wrong, the response can't be automated. Do we extend deadlines? Request reshoots? Terminate? Those calls require understanding contract terms, relationship history, timing, and risk tolerance.
The automation narrative assumes that once you solve discovery and outreach, the rest is straightforward execution. Execution is where behavioral complexity lives. It's different at every tier, shaped by relationship history. That's not a problem AI solves without infrastructure designed around behavioral reality.
How Close Are We Really?
So how close are we to total automation? It depends on what we mean. Can AI handle discovery and initial outreach? Increasingly. Generate reports and surface performance patterns? Absolutely.
Can it synthesize intelligence from scattered communications? Navigate behavioral complexity that compounds with volume? Make judgment calls about risk, relationship history, strategic trade-offs? Not without infrastructure that doesn't exist yet.
Agencies are struggling because they never built infrastructure to capture their own intelligence. Platforms are raising money because everyone recognizes something needs building. Nobody's actually solved it yet. Brands are going in-house because they're tired of operating blind on contract economics.
The real question is whether we'll build the infrastructure that makes automation meaningful. The kind that captures operational intelligence, connects it to performance data, makes both accessible for decisions. The material human judgment works from is what's missing.
