Advanced jobs
Harder bot jobs, and what it takes to do them well
Answering FAQs is easy. These 14 real-world jobs are not: each one needs the right tools, guardrails and a human in the loop. For every job you get why it is hard, a setup checklist, difficulty, matching bots and official docs.
All advanced jobs
Research-and-report agent
Searches many sources, cross-checks them and writes a cited report you can actually rely on.
Lead-gen and outreach workflow
Finds prospects, enriches them, writes personal messages and handles replies without burning your domain.
Codebase migration agent
Upgrades a framework, language version or API across a large repo without breaking production.
Automated QA and end-to-end testing
Explores your app, writes tests, runs them on every release and tells you what really broke.
Customer support triage with escalation
Answers what it can, routes the rest, and hands off to a human with full context.
Data-pipeline and automation monitoring
Watches your workflows and data jobs, diagnoses failures and tells the right person what to fix.
Compliance and contract document review
Reads long documents, flags risks against your rules and shows exactly where it found them.
Voice agent for appointment booking
Answers the phone, understands the caller, books into a real calendar and knows when to transfer.
Multi-agent workflow orchestration
Several specialized agents hand work to each other, with state, retries and a person in the loop.
AI code review gate
Reviews every pull request, flags real problems and enforces your standards before merge.
Browser agent for back-office tasks
Logs into web portals, fills forms and moves data between systems that have no API.
Internal knowledge assistant with permissions
Answers staff questions from company documents and only shows what each person is allowed to see.
Invoice and finance back-office workflow
Reads invoices and requests, matches them to records and routes anything unusual for approval.
Evaluating and regression-testing your agents
Measures whether your agent got better or worse every time you change a prompt, model or tool.
How we rate difficulty
Each job is scored from 1 to 5. Hard (3 to 4) means a capable team can do it with off-the-shelf tools plus careful review. Very hard (5) means mistakes are costly or the work spans many systems, so you need strong testing, narrow permissions and human approval before anything irreversible. These are our own judgements, not vendor claims.
Next steps
Paid bots with pricingDeep divesEveryday jobsFree templatesTake the quiz