Explore the roles emerging as work changes.
The new workforce brings people and AI agents together in shared workflows. Use market evidence to understand new responsibilities, then prepare your people for the work your organization needs.
This snapshot groups AI-related postings in engineering and business functions. The title patterns differ; use them to investigate responsibilities your enterprise needs, rather than treat the groups as a complete map of the labor market.
Job-market snapshot 2026-07-17 · 264,613 source job postings · 3,989 AI-Builder postings · 2,581 distinct titles
In the May 2026 snapshot, 47 of 422 Anthropic openings are Applied AI Architect or Forward Deployed Engineer variants. These roles connect model capabilities with customer work.
“Akin to a forward deployed engineer for internal functions. This is why, at Box, we're starting to hire for AI automation engineering roles.”
A framework for human responsibilities
In The Great Reorg, Foundation Capital proposes four human responsibilities in organizations using agents. They are a framework for discussion, not a forecast that these jobs are protected.
Owns the outcome. Signs the filing, stands behind what the agent shipped.
Designs how people and agents share the work and how exceptions are handled.
Builds and maintains relationships with the people involved.
Reviews and signs off on what AI produces.
A useful planning question is how people will gain the experience needed to check an agent’s work. Define what a reviewer must recognize, then give people practice with representative decisions and exceptions. Assessment can show performance on those cases; it cannot guarantee a future supply of expert validators.
Work beyond salaried job postings
Title-based posting counts do not capture every way AI work is organized. Domain-expert review may also be contracted per task rather than advertised as a salaried role. The counts above should not be read as a census of that work.
Prepare your people for these roles.
Connect the role to your workflows, then practise the responsibilities and assess readiness through Project Readiness. The AI Bootcamp runs for 6-8 weeks with simulations and assessment against agreed criteria. Production rollout and ongoing support are separately scoped.
Methodology
Counts are computed from 264,613 job postings in Nuvepro's real-world JD corpus (job-market snapshot 2026-07-17). Engineering buckets match by canonical title pattern. Business-function buckets match by function keyword AND any AI-flavor term (AI, GenAI, agentic, LLM, ML, prompt, applied AI). Operator buckets match canonical executive titles ("VP AI", "Head of AI", etc.). Strict to avoid over-counting.
"AI Builder" as a literal job title returns zero matches in this snapshot. We use it as a grouping term, not a claim that no employer uses that title. Cross-row title overlap is removed by DISTINCT job ID at the row level.