What You Need to Know
Claude gives its best output on complex work when you split the work into pieces that Claude can do properly, one at a time. That splitting is task decomposition. Connected parts pass results to each other: a later part needs an earlier part's output or decisions. Independent parts work on their own.
A single well-formed prompt handles a lot of work, but it has a limit. When you put a whole project's requests into one message, the result is shallow and uneven. Claude does some parts well and gives other parts little attention.
This lesson uses one made-up example throughout: a supermarket. Its manager writes a quarterly customer-feedback report for the company's main office. The report needs an outline, sections in one voice and an action list. She also has three side jobs: a competitor summary, a press release and a budget table.
The Help Center's list of remedies for an unhelpful answer includes breaking complex requests into substeps. The Help Center also says that a prompt should contain as much context as possible. Context is the background information that Claude has for a task.
Four Techniques, Lightest to Heaviest
Numbered steps inside one prompt. List connected parts explicitly, in order. The manager writes: first summarise the feedback themes, then rank them by frequency, then draft one improvement action per theme. Claude follows an explicit sequence far more reliably than one long paragraph of requests. Anthropic's guidance is to use numbered sequential steps when order or completeness matters, and to be specific about output format and constraints.
Staged prompting in one conversation. Ask for one part at a time, and review each part before the next one starts. For a report, ask for the outline, review and adjust it, then draft section by section. The approved output becomes the starting point for the next stage. So you correct the work while correcting still takes little effort.
Plan-first prompting. When the right structure is not obvious, ask Claude to propose an approach first. For example: before writing anything, list the sections you would include and what each covers. Reviewing a plan takes seconds. Reworking a wrong 2,000-word draft does not take seconds.
Separate conversations per sub-task. Put genuinely independent pieces, such as a competitor summary, a press release or a budget table, in separate chats. Separate chats keep each context clean and focused. The manager gives each of her three side jobs its own chat.
Choosing the Technique
Ask two questions about each part of the work.
1. Does a later part need this part's output or decisions? If not, the part is independent. Give it its own chat. The manager's competitor summary, press release and budget table each get a separate chat, because nothing in one of them changes the others.
2. If the parts are connected, must you check a result or make a decision before the next part can be right?
- No check needed: write numbered steps in one prompt. Numbered steps need an order that you already know, and a result that you can check once at the end. The feedback themes work this way: summarise them, rank them, then draft one action for each theme.
- Check needed: stage the work in one conversation. Put a stage break at each point where you must review the output or make a decision before the next step can be right. Anthropic's advice for a long chain of steps is to ask for an intermediate result that you can check before Claude continues. A document with several sections in one voice needs a check after each section, because each later section builds on the approved ones. The manager corrects her outline before any section exists, because every section follows the outline.
When you do not know the right structure, ask for the plan first. After you approve the plan, ask question 2 for each part.
A mismatch between technique and task fails in one of two ways. The first is one very large prompt that asks for a finished 20-page product in one pass. The second is connected work that you split across chats. Each new chat has then lost the outline and tone decisions that the previous chat made. Report sections that the manager splits across chats would lose her outline and tone.
Carrying Context Forward
A new conversation does not automatically contain another conversation's working material. Claude does now remember across chats. This memory is a set of individual topics that cover your role, projects, preferences and ongoing work context. Memory is context about you, not a transcript. Memory does not carry your approved outline, tone guide or half-finished draft into a new chat. On paid plans, you can ask Claude to search past chats and bring relevant details into the current chat. That search is retrieval: Claude fetches the earlier details only when you ask. It is not automatic carry-over.
The dependable pattern is explicit persistence: you carry outputs forward yourself. Paste the output into the new conversation, re-attach the file, or keep the output in a Project's knowledge base. Projects are self-contained workspaces with their own chat histories and knowledge. The knowledge base holds the files that you upload to the Project. When the manager returns to the report in a new chat, she first pastes her outline into the chat. Carrying outputs forward is your job, not the product's.
In the exam
The exam tests matching a technique to a situation, not reciting the four. The two questions in "Choosing the Technique", plus plan-first when the structure is unclear, find the right technique. Both failure modes appear as wrong options. Where persistent material should be stored is a Domain 3 and Domain 5 question.
Current state (November 2026)
Memory became available on claude.ai on 11 September 2025 for Team and Enterprise plans, then for Pro and Max from 23 October 2025. As of 14 November 2026 memory is on by default for Free, Pro, Max and Ultra. As of that date, on Team and Enterprise an owner decides whether memory is available. In organisations created before 31 October 2026, memory stays off for each member until that member enables it. In most newer ones, memory starts on. Chat search, which lets you ask Claude to reference past conversations, is available on paid plans. Neither feature changes the rule: memory saves context about you as individual topics, and you must ask for chat search.
Exam Traps
One very large prompt that asks for everything, chosen to keep the whole deliverable in one place
Attention spreads unevenly across a project of requests in one message. As a result, some sections are strong and others are skimmed. Corrections also arrive only after the whole thing is wrong. Breaking complex requests into substeps is the documented remedy.
Fragmenting connected work across separate conversations
Each new chat lacks the outline, tone decisions and approved material that the previous chat built. So the pieces arrive in different voices. They need reconciling by hand. Separate chats are for independent deliverables, not connected ones.
Plan-first prompting offered where the structure is already known and agreed
Asking Claude to propose sections you have already settled is a small unnecessary extra step: it adds a review round without adding information. Plan-first is useful only when the right structure is genuinely unclear.
Practice Scenario
A programme director at a charity must produce a 20-page annual impact report: an overview, five sections drafted from the notes of five different programme teams, and one consistent voice throughout. The director wants to approve each part as it is produced, and never to explain settled decisions again to a new chat. How should the director structure the work with Claude?
Build Exercise
Stage a Multi-Part Deliverable
What you'll learn
- How to sort the parts of a deliverable into connected and independent
- How to assign the simplest technique that fits each part
- How to correct an outline before any drafting starts
- How to carry an approved output forward into later stages
- Choose a genuinely multi-part deliverable from your work: an annual report, a project plan, a course pack
Why: Decomposition only shows its value on work too big for one prompt; a small task will not need the techniques.
You should see: One deliverable named, with its parts listed as bullet points.
- Sort the parts into connected, meaning a later part needs an earlier part’s output, and independent, meaning they stand alone
Why: This sorting answers the first question in Choosing the Technique. Independent parts need separate chats. Connected parts stay together in one prompt or one conversation.
You should see: Each part labelled connected or independent, with one phrase saying why.
- Assign a technique to each part: numbered steps, staged conversation, plan-first, or a separate chat, with one sentence of justification each
Why: Matching technique to task is the objective itself; the justification makes you check the choice.
You should see: A short list pairing every part with a technique and a reason.
- Run the first stage, the outline or the plan, and correct it before any drafting starts
Why: Corrections at the outline stage take seconds; the same corrections after 2,000 words take an afternoon.
You should see: An approved outline you have actively edited, not just accepted.
- Each time a later stage needs an earlier output, note how you carried it forward: pasted in, re-attached, or stored in a Project
Why: A new conversation does not automatically contain the working material of an earlier one; explicit carry-forward is the reliable pattern.
You should see: A note per stage recording the carry-forward method you used.
Sources
- My prompt isn't giving me a helpful answer — Anthropic
- Prompting best practices — Anthropic
- How do usage and length limits work? — Anthropic
- Use Claude's chat search and memory to build on previous context — Anthropic
- Bringing memory to Claude — Anthropic
- What are projects? — Anthropic
- AI Capabilities and Limitations: Steerability — Anthropic