SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
ServiceNow is seeking a Staff Software Engineer (IC4) to design, build, and operate agentic AI capabilities across the Sales and Order Management domain. This role focuses on building AI-native features—intent interpretation, multi-step reasoning, tool invocation, and autonomous action—rather than traditional deterministic systems or ML research.
Key responsibilities include:
• Design and ship agentic features spanning guided offer selection, natural-language order construction, conversational order status updates, exception triage, and availability inquiry.
• Decompose sales and order processes into steps and decision points suitable for agent execution, determining where autonomy is appropriate versus where human checkpoints are required.
• Author and maintain agentic instructions (system prompts, role definitions, tool descriptions, guardrails, escalation rules) as first-class engineering artifacts under code review and source control.
• Build automated evaluation frameworks for non-deterministic behavior, including golden datasets, multi-turn conversation suites, model-as-judge scoring, CI gates, drift detection, and adversarial testing.
• Design conversational experiences across channels (chat and voice) that maintain context, hand off cleanly between automated and live agents, and account for voice-specific constraints (latency, barge-in, speech recognition error).
• Specify requirements precisely and direct AI coding agents, converting requirements into testable specs, decomposing work into agent-sized tasks, and owning output quality regardless of what produced it.
• Own production quality, safety, and reliability: monitor conversation quality, hallucination rates, tool-selection errors, and unsafe actions; defend against prompt injection and data leakage; maintain observability and rollback mechanisms.
• Build full-stack delivery around these capabilities—front-end experiences, server-side logic, APIs, data models, and integrations with quoting, billing, contract, case, and fulfillment systems.
• Mentor IC1–IC3 engineers and raise team practices around instruction authoring, evaluation, and accountable agent use.
This is distinct from traditional full-stack engineering (deterministic logic) and ML research (training foundation models). The most important logic often lives in natural language and must be engineered with the same discipline as code. Correctness is established by measuring behavior at scale through automated evaluation rather than asserting fixed outputs.
