The Agent That Found a Million Mentions and Still Couldn't Choose
Yesterday a B2B startup raised €1.03 million to watch a million creator posts a minute. Discovery is cheap now. The founder still has to decide who gets the first miss.
Yesterday a Belgrade-born founder raised more than a million euros to solve a problem every operator already feels in their calendar. Astute, a two-person B2B media company, closed an oversubscribed pre-seed of €1.03 million to automate creator partnerships with a pair of AI agents. One watches more than a million posts a minute. The other handles the admin that used to live in a founder’s late-night spreadsheet. The Recursive reported the round on August 17. The product names the bottleneck correctly and still leaves the expensive part untouched.
B2B companies cannot find, reach, and run partnerships with newsletters, podcasts, and social creators without burning weeks of manual work. That is the problem the raise is built on. Founders hear that sentence and nod because their own pipeline looks the same. They have a list. They have a tracker. They have a person whose unofficial job is chasing people who sounded interested in March. What they do not have is a rule for who gets the first real miss. Discovery got cheaper overnight. Selection stayed a personality test.
Volume arrives before judgment
Vida Stanić, Astute’s CEO, said B2B marketing lived on ads, PR, and SEO while audiences moved to newsletters, podcasts, LinkedIn, and X. In consumer brands, that shift turned influencer spend into a line item. She said B2B is next. The investor quote in the same piece praised founder-market fit: one founder needed the product as a marketer and could not buy it, the other sat on the creator side of the table. That story sells a round. It also describes the trap operators walk into the moment a tool floods the room with names.
A million mentions a minute produces a feeling that looks like diligence. The dashboard lights up. Competitor names appear. A creator who mentioned your category last Tuesday becomes a candidate. The founder forwards three profiles to the team and calls it a pipeline. Nobody has asked what happens when the first deliverable slips, or who owns the rewrite, or whether the creator’s audience even buys the thing you sell. The agent did the hunting. The founder still has to do the choosing, and choosing under a flood feels like falling behind, so they approve the first three that look warm.
Agencies repeat the pattern with better branding. They sell access to a roster and treat the roster as the strategy. Consultants sell introductions and treat the calendar as proof of work. Creators collect brand deals the way founders collect partners, by volume, then discover in month four that none of the relationships can survive a hard conversation. The software did not invent this. It compressed the time between the first email and the first disappointment.
The cost sits in the first yes
The founder who treats a recommendation engine as a partner filter pays in two currencies. The first is calendar. Every new creator, agency, or channel partner who enters without a standard consumes a weekly check-in that used to belong to the one relationship that might have produced revenue. The second is credibility. After three soft exits, the market files you as someone who starts partnerships and cannot finish them. Future counterparties price that. They send junior people. They keep the real inventory for operators who look harder to get into the room.
There is a quieter cost inside the company. The team learns that a mention is a mandate. Sales starts quoting creators who have never seen a deck. Marketing starts writing co-branded copy for people who have not agreed to a date. Finance starts modeling a revenue share against a list that has never closed. The agent measured impressions, conversions, brand score, and visibility in AI search. Those are useful numbers after a relationship exists. They are decoration before anyone has named the work, the owner, and the consequence of a miss.
Selection systems that force the hard sentence early, including platforms like onSpark AI when they are used to surface how a counterpart behaves under friction rather than how often they appear in a feed, only work if the founder is willing to reject a warm mention. Warmth is abundant now. Judgment is the scarce asset, and it looks slow in a week when a competitor just announced twelve new partners.
Keep the filter human
The operators who will survive this wave write three sentences before they let an agent populate a list. They name the one behavior that disqualifies a partner in the first thirty days. They name the person inside their company who is allowed to say no after the dashboard says yes. They name the first miss in advance, so the miss arrives as a scheduled conversation instead of a surprise that everyone tries to soothe.
If those sentences feel too sharp for a tool that just made discovery feel infinite, the tool is doing its job and you are refusing yours. Infinite discovery without a standard is how founders stay busy and stay lonely at the same time. Astute raised money to watch the internet. The work on this side of the table is deciding, before the first outreach, which mention is allowed to become a relationship, and which one is only a mention.