What Changed
Clay shipped auto-mode as the new default for Claygent — a post-trained model router that selects the appropriate LLM for each task without any manual configuration. Simple lookups go to faster, cheaper models. Complex research tasks get routed to frontier models. Each routing decision resolves in under a second.
This isn't a beta or an opt-in experiment. Auto-mode is now the default for all new Claygent columns, workflow nodes, and the Claygent Builder test environment. If you're creating new enrichment columns starting today, you're already running on it.
Why the Number Matters
Clay published a hard figure from production traffic: auto-mode costs 47% less than routing every task to the most capable model, while holding frontier-level output quality.
For teams running enrichment at scale — prospecting lists in the tens of thousands, continuous data refreshes, multi-step research workflows — that's not a marginal optimization. If you're spending $2,000/month on Clay enrichment credits, you're looking at a potential $940/month reduction with zero workflow changes. At $10,000/month, the math speaks for itself.
The key detail is that quality is maintained, not traded away. The router doesn't just pick the cheapest model — it picks the right model. That distinction matters for teams that have already calibrated their enrichment outputs to a quality standard and can't afford regressions in data fidelity.
What Operators Should Do Right Now
For existing Claygents, your current model selection stays in place unless you manually switch. That means there's a one-time audit worth doing: review your highest-volume Claygent columns and switch them to auto-mode. The expected outcome is cost reduction with equivalent output quality — and if you see quality drift on any specific task, you can pin that node back to a specific model.
For teams building net-new enrichment workflows, auto-mode is already the default. There's nothing to configure.
The broader takeaway for operators: this is the kind of AI infrastructure improvement that compounds quietly. Model routing optimization isn't visible in your pipeline the way a new data source is — but it directly affects the unit economics of any AI-augmented GTM motion. If your RevOps team is running Clay at scale and hasn't reviewed enrichment costs against the new baseline, that's a 30-minute audit that has a defensible ROI before you leave the building.