Let's get this out there: Braintrust's promise of a "decentralized talent network" with lower fees is compelling on a spreadsheet. But their operational model has a critical, hidden cost that doesn't show up in their 10% platform fee: the cost of your own time waiting for support.
I've been tracking our team's engagement for six months. The pattern is unmistakable. A technical issue with a contract or a payment discrepancy? You'll get a template email in 24-48 hours, followed by a 72-hour radio silence before the next non-answer. We once had a timezone mismatch on a milestone deadline that needed an immediate admin override. Took four days and three increasingly frustrated messages to resolve. For a platform built on speed and efficiency, their support latency is ironically high.
Now, contrast this with the experience of a colleague at a Fortune 500 that also uses Braintrust. Their "account manager" (a concept they never mentioned to us) resolves similar issues in under two hours. The difference in treatment is stark. It appears their support tiering is based purely on client spend volume, not on the principle of a decentralized, equitable network they market.
So, the real cost equation isn't just the 10% fee. It's the 10% fee *plus* the operational drag of slow resolution, which directly impacts project velocity and, ultimately, your own burn rate. I'd love to see someone from Braintrust actually refute this with their average response time metrics segmented by client tier. Until then, my data shows you're paying for "enterprise-grade" support but receiving "ticket-in-a-black-hole" service unless you're a whale.
- cost_observer_42
cost_observer_42
That's a really interesting point about the hidden time cost. It reminds me of setting up monitoring - if your alerting pipeline has high latency, the system might look efficient on paper, but you're losing hours diagnosing issues that could've been solved in minutes.
> the difference in treatment is stark
I wonder if this is just a scaling problem they haven't solved yet. In DevOps, we try to automate responses for common issues to avoid tiered support. But maybe their "decentralized" model makes that harder? Have you found any workarounds, like a specific contact method that gets better results?
Learning by breaking