This lesson is being recorded
Every video of this course is being uploaded to my YouTube channel. Subscribe so you do not miss it.
Bruno Ashford @brunoashfordSubscribeModule 7 · Lesson 7.4 · 11 min
Parallel Work
Run several specialists on the same problem at the same time to get multiple perspectives quickly.
Use ← → to move. The marker highlights text and the pencil draws on the slide; navigation locks while you annotate. Undo removes the last mark, Esc exits.
1Ask for parallel work explicitly
Tell Claude which specialists to use and that they should work in parallel and independently. Specify the same inputs for all of them so the comparison is fair.
I am considering raising our prices by 12 percent. In parallel and independently, ask three subagents to analyse this: the data-analyst (using data/ for the last 12 months of sales), the sdr-researcher (competitor prices in our market, with links), and the reviewer (risks and customer reactions to watch). Then combine their findings into one page: where they agree, where they disagree, and your recommendation.2Split big research into slices
Parallel work is also great for volume. Twenty leads to research? Ask for five researcher runs with four leads each. Each run stays small and focused, and the whole batch finishes much faster than one long sequence.
3Force a structured return
Ask every specialist to return the same format (for example: finding, evidence, confidence high/medium/low). Structured answers are easy to merge. Free-form essays are not.
4Look for disagreement, not just consensus
The value is in the tension. If the analyst says prices can rise and the reviewer warns about churn in one segment, that is the real conversation to have with your team. Ask Claude to highlight disagreements first.
5Watch cost and limits
Each subagent uses its own context, so parallel work uses more of your plan's usage than one prompt. Save it for decisions and batches that matter, not for everyday questions.
When you face a big decision, like raising prices or opening a second location, you would ideally ask your finance person, your sales lead and your most sceptical advisor, all independently, before they influence each other. In real life that takes a week of meetings.
With subagents, Claude Code can send the same question to several specialists at once. Each works in its own clean context, and the main session brings the answers together. You get independent perspectives in minutes, and you can see where they agree and where they fight.
Key points
- Parallel subagents give independent views before they influence each other.
- Use it for big decisions and high-volume batches.
- Same inputs and the same output format for every specialist.
- Disagreements are the most valuable part of the output.
Avoid
- Running parallel agents for simple questions and burning usage.
- Giving each specialist different inputs, then comparing apples with oranges.
- Reading only the combined summary and never the individual reports.
A roofing contractor debating whether to add gutter installation as a service asked for three parallel views: the analyst on his job history and margins, the researcher on local competitors and pricing, and the reviewer on operational risks. The analyst found gutters were already requested in 18 percent of quotes. The researcher found two competitors bundling them. The reviewer flagged that his crew had no certified installer. The decision became obvious: partner first, hire later.