One assistant against three: when a chain works better
One good chat for every task works up to a certain volume; beyond it the context gets tangled and the result drifts. Below is a real chain where each assistant does its own part, and five standard combinations covering most tasks.
One good chat for every task works up to a certain volume; beyond it the context gets tangled and the result drifts. Below is a real chain where each assistant does its own part, and five standard combinations covering most tasks.
---
One chat for everything is the bottleneck
Somebody learns to work with assistants and does the natural thing: finds one good one and does everything in it.
That works — up to a certain volume. Then the symptoms appear.
The context gets tangled. In one conversation you discussed a client, and in the next message you ask for copy — and the model drags the previous thing in.
The result drifts. The same type of task comes out differently depending on how you phrased it this time.
Quality is mediocre everywhere. No part of the task has any depth, because the assistant is holding style, structure, logic and grammar in mind at once.
The problem is not the model. You are using one universal tool where a set of specialised ones is needed.
What a chain of assistants is
The task gets split into parts, each with its own performer, and each result feeds the next.
Not "do everything for me" but a sequence where each one works inside its own frame.
A real example from practice. A series of covers in a single style is needed.
The first — an art director. The task: how a series gets built so the frames read as one line. It produced a split into constants and variables: a modular grid and an anchor zone, one typographic voice, a running colour anchor at a 60-30-10 proportion, a single crop — those do not change. The subject figure, the secondary colour, the text hook — those do.
The second — a prompt generator. It took that brief and turned it into a prompt for generation: a modular grid in two columns at 9:16, exact colour codes — a graphite background, muted olive, a mustard-gold accent — and the bottom right zone left 15% empty for a duration badge.
What happened there. The art director does not know the syntax of generative models and does not need to. The generator does not know your brand system and does not need to. Each did its own, and not one requirement got lost in the handover.
Ask one assistant to "make a series of covers" and you get pictures with no system.
Why handing over works better
Three reasons, all practical.
Each works inside its own frame. An assistant does not have to think about meaning and form simultaneously.
The requirements get fixed as text. When you write the prompt yourself, half the decisions stay in your head — "well, it should be contrasty". In a handover between roles, everything is written down.
The intermediate result can be checked. If you do not like the brief, you correct that rather than redoing the final result.
Standard chains
Five combinations covering most tasks.
Meaning → form. A specialist assistant supplies the content, a copywriter turns it into text, an editor cleans it up.
Analysis → proposal. One assistant works through a client conversation and pulls out the pains, a second builds a deck brief from them.
Data → conclusion → material. An analyst works through the figures, a marketing one turns the conclusions into content.
Draft → critique → revision. One writes, a second reviews against criteria, the first reworks against the notes.
A role → checked by another role. The legal one looks at the wording, the sales one at how it sounds to a client.
The last combination is the most underrated: a check by a different role finds what the author cannot see by definition.
Modes that replace a second role
Separately, because almost nobody knows about this.
Most assistants have built-in refinement modes. After an answer they offer to:
Assess the work — a review of their own result against criteria with scores and reasoning.
Run a harsher assessment.
Give feedback as a focus group — the model goes through the result through the eyes of the recipient: a buyer, a client, a reader.
Give an expert review.
Refine against the notes — corrections against criteria rather than a general "make it better".
Propose a new approach or change the format, style or length.
The practical point. You do not rewrite the request ten times. You get a first answer, run the harsh assessment, then the focus group — and out comes a document that has been through three rounds of review. With no involvement from you beyond choosing the mode.
For many tasks that replaces the second role in a chain.
How to build your own chain
An order that stops this becoming extra work.
Step 1. Split the task into different kinds of work. Not into stages of time but into genuinely different types: invent, present, check.
Step 2. Find a role for each. The catalogue has 14 categories: marketing, sales, creative, media, legal, HR, analytics, e-commerce, engineering, entrepreneurs, personal development, PR, client retention, freelancing.
Step 3. Hand the result over whole. Not a summary but the text of the previous step. Summarising loses exactly what you started the chain for.
Step 4. Check at the joins. The most common place to lose something is the transition between roles.
Step 5. Do not lengthen it needlessly. Two or three roles cover most tasks. A chain of six assistants usually means the task was badly split.
When a chain is not needed
Honestly, because complicating things for its own sake is a common mistake.
A simple task. One question, one answer. No need to build a pipeline to write an email.
One-off work. Setting a chain up pays off through repetition.
When there is no intermediate result you can judge. If you cannot tell whether the first one produced a good brief, handing it to the second will improve nothing.
What this gives you in practice
Predictability. One type of task comes out the same, because the order is set.
Transferability. A chain can be handed to a member of staff or a client along with a description.
Accumulation. A worked-out combination stays with you and gets reused. A second client costs less than the first.
And scale. If the task repeats, a chain of assistants becomes a chain in AI Workflow — there each step's result gets passed automatically and an iterator runs it down a list.
What is available on the platform
More than 140 AI assistants across 14 categories. Each is a role with its own frame and its own clarifying questions.
Refinement modes on most of them: assessment, harsh assessment, focus group, expert review, criteria-based refinement.
AI Workflow — if a combination repeats: a chain where the handover is automatic. A "chat model" block with twenty models to choose from, an "AI assistant" block for conversing to the result you want, an "approval" block for stopping at your decision, and a "template as a step" block for nesting ready chains.
Building your own assistants — the Creator+ plan, when the ready roles are not enough.
Chat access comes with the Basic plan. Registration is free and opens three days of full Basic.
Where to start today
One action. Take a task you usually do in one chat and split it into two different kinds of work.
For example: invent and present. Or analyse and write.
Run it through two assistants, handing the first one's result to the second whole, with no summarising.
The difference is usually visible immediately — and it explains why one universal assistant was producing the average.
---
How this is arranged in multi-agent systems
A chain of two or three assistants is the manual version. Once there are many roles, a hierarchy appears.
The automations archive holds a system prompt generator for multi-agent teams. It distinguishes three levels of agent:
Director — the top level, distributing tasks and pulling the result together.
Supervisor — the middle, running its own area and managing the performers.
Basic — the performer of a specific operation.
The assistant determines an agent's level, extracts information about its tools and subordinates and produces a system prompt to that level's template. It works even from a screenshot of the settings.
Why know this if you work in chats. The same logic applies with no automation at all: one assistant sets the frame, a second works inside it, a third checks. The only difference is that in a system the handover is automatic.
The archive — Make, N8N and Full.
---
What to read next
[Six brief fields](/en/blog/six-brief-fields) — what to give the first one in the chain.
[How to build your own AI assistant](/en/blog/build-your-own-assistant) — when the ready roles are not enough.
[A virtual board of directors](/en/blog/virtual-board-of-directors) — a chain of six roles on one decision.
[Skills for Claude Code](/en/blog/claude-code-skills-guide) — the same logic inside one tool.