AI agent examples by role — 20 concrete use cases

A common frustration with AI agent content: it's abstract. 'Agents will transform work.' Sure. But what does one actually do on a Tuesday morning? Here are twenty concrete examples, one per role.

Recruiting

Candidate sourcing agent — given a job description, the agent pulls from LinkedIn, GitHub, and the firm's ATS; scores each candidate against the JD; drafts personalized outreach; schedules intro calls when candidates reply. Reduces 500 applications to a ranked 15, every day. See full recruiting agent.

Financial analysis

Deal screening agent — for every inbound memo, the agent scores fit against the firm's thesis, extracts the key financials, builds the comps table, and drafts a one-page screening memo. Reduces a week of partner-level screening to a same-day decision. See analyst agent.

Legal

Contract review agent — the agent reads every inbound NDA, MSA, or SOW; flags deviations from your playbook; proposes redlines in your firm's style; and drafts the client response. A 40-minute review becomes a 4-minute review. See legal agent.

Consulting

Case research agent — from a client question, the agent runs market sizing, competitor scans, and customer research; produces a sourced briefing document in your firm's template. Three hours of desk research becomes thirty minutes. See consulting agent.

Underwriting

Loan application triage — the agent parses every application, verifies docs, runs credit + income analysis, computes DTI/LTV, and drafts the UW memo in your bank's format. Automates the first 45 minutes of every file. See underwriter agent.

Sales

Call briefing agent — 15 minutes before every discovery call, a pre-read hits your Slack: priorities of the contact, LinkedIn highlights, likely objections, 5 tailored questions. No more cold calls. See sales agent.

Accounting

Transaction classification — every transaction across every client is classified against the chart of accounts. Ambiguous ones surface for one-click approval. 2,000 transactions classified in the time it used to take to classify 50. See accounting agent.

Founder / CEO

Inbox triage — reads every inbound, drafts replies, schedules the meetings worth having. The founder sees a distilled 10 instead of 200. See founder agent.

Operations

Daily ops briefing — every morning at 7am, one Slack message with key metrics, anomalies, overdue work, and today's priorities across the team. Replaces an hour of tracker review with a 2-minute read. See ops agent.

Product

User research synthesis — reads every interview transcript, clusters themes, pulls verbatim quotes, writes the research readout. 30 interviews synthesized in 20 minutes. See product manager agent.

Marketing

Content brief agent — from a topic, the agent produces a brief: angle, outline, keywords, examples, CTA. The writer fills in the voice. Shaves the blank-page tax off every piece. See marketer agent.

Real estate

Comparable sales agent — for any address, the agent pulls recent comps, adjusts for condition and features, produces a CMA in your template. Three hours of Zillow scraping becomes three minutes. See realtor agent.

Bookkeeping

Bank reconciliation — the agent reconciles bank, credit card, and A/R against the ledger daily. Only variances that need human judgment surface. Turns month-end close into a non-event. See bookkeeper agent.

Legal operations (paralegal)

Discovery review — reads discovery docs, clusters by issue, flags privileged and responsive docs, outputs the privilege log draft. What took a team of contract reviewers now takes an associate reviewing the flags. See paralegal agent.

Customer success

Account health monitor — watches every account for usage, tickets, NPS, sentiment. Flags churn risk 60 days before it surfaces in the data. A 200-account book becomes manageable. See CSM agent.

Design

Design system auditor — watches shipped work for drift: wrong tokens, off-brand type, inconsistent spacing. Weekly audit delivered to the design lead. System stays clean without policing. See designer agent.

Research

Literature scanner — watches arXiv, PubMed, SSRN, and key conferences. Summarizes new papers in your field weekly with relevance scores to your research agenda. Eliminates the 'what came out this week' reading task. See researcher agent.

Strategy

Theme watcher — tracks the themes you care about across 20+ sources. Weekly digest with signal-to-noise ratio. What strategists used to get from three paid feeds, now personalized. See strategist agent.

Coaching

Session prep agent — before each coaching session, the agent briefs the coach: last session's themes, homework status, what to probe today. Preserves deep presence for 20 clients a week. See coach agent.

Journalism

Background research agent — for any source or company, the agent drafts a background brief: public record, prior statements, litigation, funding, known relationships. Thirty minutes of background research becomes three. See journalism agent.

Patterns across all twenty

A few things to notice:

• Every example is a process, not an answer. Agents do work; chatbots answer questions.
• Every example has measurable output. An agent you can't measure isn't an agent — it's a chatbot on a cron.
• Every example leaves the judgment layer to the human. The agent does the 80%; the human does the 20% that earns the salary.
• Every example lives in the tools the role already uses. Agents don't force a workflow change — they live inside one.