How to spot a good first task
You do not need a strategy. You need one task that passes four quick checks. The more boxes it ticks, the easier the first win.
- It repeats. You do it every week or every day. A one-off is not worth training, a routine is.
- It has rules. You could write the steps down for a new hire. If it lives only in your head, write it first.
- The data is reachable. The work can be done from the inboxes, files and tools you already use.
- It has a low blast radius. If a draft needs a fix, nothing breaks. You approve before anything goes out.
All four true? It is ready to hand over. The five jobs below almost always pass.
The five to start with
- 01
The weekly recap
BEST FOR · CEOs, GMs, sales and ops leadersSales numbers, ops status, a team update, written in your format and sent on schedule. It is the same shape every week, which is exactly what an AI worker is good at. You approve the first few, then it just lands in your inbox.
- 02
Inbox triage
BEST FOR · Founders, anyone drowning in emailSort, label and draft replies to the routine messages, so you open your inbox to a short list that actually needs you. Nothing is sent without your nod, so the risk is near zero.
- 03
Follow-ups
BEST FOR · Sales leaders, coaches, foundersAfter every call or meeting, a draft follow-up ready before you forget. Quotes that went quiet, proposals waiting on a reply, the chase you keep meaning to do. It drafts, you approve, it sends.
- 04
Meeting and lead prep
BEST FOR · Sales leaders, consultants, GMsWho you are meeting, what they do, what to say, gathered and on one page before the call. Tedious to do by hand, easy to hand over, and a quick way to look sharp every time.
- 05
The recurring report
BEST FOR · GMs, finance, marketing leadsThe monthly deck, the competitor scan, the variance write-up, built on your template from the same sources each time. Slow and repetitive for a person, a natural fit for an AI worker.
What to keep for later
An AI worker is not meant to do anything and everything from day one. Some work earns a human first, and some should stay with a person for good. Hold these back until the agent has proven itself on the easy jobs:
- Judgement calls. Pricing a tricky deal, a sensitive client reply, anything that turns on reading the room.
- Undocumented work. If nobody can write down how it is done, an agent cannot run it. Map it first, automate later.
- One-offs. A task you do once a year is not worth training. Routines pay back, exceptions do not.
- High-stakes actions. Payments, contracts, legal commitments. These stay behind a hard approval gate, or with you.
This is the point a lot of AI tools get wrong. The goal is not an agent that does everything. It is a system you can control, that takes the routine off your plate and leaves the judgement with you.
Pick yours in ten minutes
List the work that repeats every week. Cross off anything that lives only in your head or carries real risk. From what is left, pick the one that eats the most of your time. That is your first job. If you want a second opinion before you commit, our free readiness check scores you in 60 seconds, and the free audit picks the highest-value task with you.
FAQ
- What should I automate with AI first?
- One recurring task that follows clear rules, runs on data your tools already hold, and won't cause harm if a draft needs a fix. The weekly recap, inbox triage, follow-ups, meeting prep and a recurring report are the safest first jobs.
- How do I know a task is a good fit?
- It repeats every week, can be written down as steps, runs on data the agent can reach, and has a low blast radius if it gets something wrong. Four for four means it is ready.
- What should I not give it first?
- Judgement-heavy calls, one-off projects, undocumented work, and anything high-stakes like payments or legal commitments. Those come later, once the agent has earned trust on smaller jobs.