Prepared by: Averity | Source: Internal recruiter roundtable (weekly team call) | Date: July 30, 2026
This brief reflects direct, front-line observations from Averity’s recruiting team based on active searches and candidate/client conversations, and is intended to give a grounded, practitioner-level view of the market beyond general press coverage.
Executive Summary
Our recruiters compared notes across their current open roles and candidate conversations this week. The clearest signal: “AI” as a hiring buzzword has splintered into two very different realities on the ground — companies building AI products (agents, copilots, embedded/on-device models) and companies using AI coding tools (Claude Code, Cursor, Codex) to accelerate delivery with existing teams. Almost every open req now references AI in some form, but what that means varies enormously by company stage and size.
A second, equally important signal: hiring behavior is diverging sharply between early-stage/seed startups and established mid-to-large companies. Startups are increasingly unrealistic — rejecting strong candidates over pedigree mismatches, culture-fit obsessions, and a preference for younger, “cooler” hires — while more mature companies (300–1,000+ employees) are hiring in a fairly business-as-usual way, just more selectively.
Titles & Technologies in Demand
Across the team’s active reqs, a few patterns repeated regardless of client:
• Full-stack / product engineering: TypeScript, Node, React, Vue, JavaScript — by far the most common stack mentioned. Go and Python appear consistently as secondary languages, especially on backend-heavy or AI/data teams.
• AI/ML-specific roles: titles are blurring — “ML Engineer” and “AI Engineer” are frequently interchangeable and depend entirely on what the person actually does day to day, not the label. Traditional ML engineering skews toward modeling; AI engineering increasingly means agentic/agent-building work.
• Seniority mix: senior/staff/principal engineer, lead AI engineer, and deployed/lead data scientist roles were all mentioned. Several clients are also hiring product managers and heads of product for AI/SaaS lines.
• Data/cloud fundamentals still matter: AWS and Databricks came up as near-universal baseline knowledge, particularly for ML tooling built on top of cloud platforms.
• Location model: results were mixed — some roles are fully remote with geo-banded (tiered) compensation that excludes candidates from NYC, the Bay Area, and DC; others (especially in regulated or defense-adjacent spaces) require 5 days on-site, U.S. citizenship, and security clearance eligibility.
AI Coding Tools: What Companies Are Doing
This was the most requested clarification on the call, since “AI” is used so loosely that it’s become nearly meaningless without more context. The team broke it into two distinct categories:
1. Building AI products: companies developing agents, copilots, or embedded/on-device models as their actual product. Examples raised: a defense-acquisition platform with an agent layer that answers sourcing questions; a browser company distilling large models into small, on-device models for local (non-cloud) inference; consumer/service companies building conversational AI to automate the front end of customer interactions (e.g., senior-care referral intake) before routing to a human deeper in the process.
2. Using AI to build faster (internal tooling): companies of nearly every size are now using Claude Code, Cursor, and Codex (OpenAI) to accelerate their own engineering. Anecdote shared: a manager at American Express described a codebase refactor that would normally take 10–12 months completed in about six weeks using AI-assisted coding.
Timing: multiple recruiters independently placed the shift around January–February 2026, describing it as a “light switch” moment where agentic AI tools (as opposed to earlier generative AI) suddenly became a baseline expectation in interviews and job requirements, after a full prior year of hesitancy about trusting AI-assisted code in production.
Effect on Interviewing and the Talent Bar
• LeetCode-style interviews are being phased out at multiple companies (Google was cited as having dropped them) in favor of evaluating how well a candidate can prompt/steer AI tools to fix or extend real code.
• There is real concern on the team that this overcorrection is risky: some strong, experienced engineers are being passed over solely for lacking hands-on production experience with Claude Code, when the underlying skill (engineering judgment) is what actually matters.
• Consensus view: AI coding tools are a force multiplier for engineers who already know what they’re doing (compared to “steroids” — they help someone who can already perform, but don’t create skill that isn’t there). Junior or inexperienced “vibe coders” being placed into sensitive systems (e.g., healthcare-adjacent) was flagged as a genuine risk companies aren’t fully weighing.
• A widely shared industry view relayed on the call: a senior engineer fluent in AI-assisted development can now credibly perform the output of two to three people, which is compounding concerns about a smaller number of available seats for junior and mid-level engineers going forward.
Startup vs. Established Company Hiring Behavior
| Category | Startups vs. established companies |
| Talent prefs | Startups avoid big-company candidates, often hiring in founders’ own image. Established firms target 10–15-yr candidates who know what they want. |
| Culture asks | Startups want younger, “cooler” hires (one client said no “nerds”). Established firms lead with mission and remote flexibility. |
| Responsiveness | Startups are frequently unresponsive despite active hiring, especially for 3–7 yr candidates. Established firms show markedly better, more “palpable” engagement. |
| Expectations | Startups “unicorn hunt” — unrealistic pedigree/hours vs. pay. Established firms hire business-as-usual with grounded offers (e.g., $265K base for an experienced hire). |
Enterprise & Infrastructure Signal
Less visibility into Fortune 500/finance hiring, but signal suggests steady, not urgent, AI integration — mirroring the decade-long cloud adoption curve in financial services. Real demand also sits in infrastructure/hardware — chips, power, cooling, drone/UAS and embedded systems — not just headline AI roles. This aligns with Averity’s historical strength in large, process-heavy clients (media, finance, insurance, healthcare) and is a deliberate re-engagement target.
RIF patterns behind current layoffs split two ways: genuine restructuring, or RTO mandates driving attrition while remaining staff absorb departed colleagues’ workload.
Create an Account or Sign In