Managed SLA.
We operate your AI agent system after launch. 24/7 on-call, drift monitoring, model upgrades, quarterly evaluation, audit support. Recurring monthly partnership for organizations without an in-house AI ops team.
AI systems are 30% engineering and 70% ongoing operation.
The build is where attention concentrates — but the lifetime of an AI agent system is in operation. Models get retired and replaced. Customer data drifts. Eval pipelines need re-running against new edge cases. Audit logs need monitoring. Compliance evidence needs to be ready when the auditor asks.
Most of our customers don't want to build an in-house AI ops function in year one. The Managed SLA gives them a senior AI ops team without the headcount. It also gives us continuous insight into what works in production, which compounds back into every new engagement.
What we actually do month-to-month.
24/7 on-call
Paged for production incidents. Standard 30-minute response SLO; tighter SLOs available for healthcare and customer-facing deployments.
Drift monitoring
Continuous evaluation against the eval harness from the build engagement. We flag accuracy drift, latency drift, and cost drift weekly.
Model-upgrade testing
When the upstream model is upgraded, we test the new version against your eval suite before promoting it. Roll-forward and roll-back paths.
Dependency upgrades
RAG pipeline, vector DB, embedding model, agent framework — all tested and upgraded with the same care as the LLM itself.
Audit support
When your SOC 2 / HIPAA / ISO 27001 auditor asks for evidence, we deliver. Pre-staged evidence packages save weeks of audit prep.
Quarterly re-evaluation
Every quarter we re-run the full eval suite, including new edge cases from production. Customer-facing readout and recommendations.
Three tiers. Pick the one that matches the workload.
Standard
Internal-use systems, business-hours response, quarterly evaluation. Single agent system, single deployment.
Production
Customer-facing or compliance-critical systems. 24/7 on-call with 30-minute response. Monthly evaluation cadence.
Mission-critical
Clinical-grade or revenue-critical. 24/7 with 15-minute response, weekly evaluation, named engineer on dedicated rotation, audit evidence pre-staged continuously.
All tiers per agent system. Multi-system customers get tiered pricing. Annual contracts standard; monthly available with a small premium.
We'll hand off cleanly. Many customers prefer not to.
The Managed SLA is optional. Every Build Engagement we ship includes a complete runbook, a documented codebase, and team enablement — your team can take over operation on day one if that's the preference.
Customers who keep us on do so because the operational depth is hard to replicate internally for a single AI system. Once your AI ops practice is at scale (multiple systems, dedicated team, mature eval infrastructure), the equation shifts. We're happy to operate during the ramp-up years and hand off when your team is ready.