The AI Text Tool: A Complete Guide to AI Writing

The text tool is the workhorse of WorkCrafter: articles, product copy, emails, summaries, rewrites and brainstorms all come from it. This guide covers what it does, which engine to pick, and the prompt habits that separate usable drafts from wasted credits.

What the text tool does

You describe what you want in plain language and pick a mode — write, rewrite, summarize or brainstorm. The tool returns a structured draft you can copy, refine or regenerate. It's built for real work, not novelty: marketing pages, blog sections, outreach emails, meeting notes and idea lists.

Choosing a model

Each generation lets you pick the model. WorkCrafter's text tool offers twelve engines across three families — Claude from Anthropic and GPT from OpenAI. The right choice depends on what the task is worth.

• Automatic (recommended): WorkCrafter picks a sensible model for the task. Start here for most work.
• Claude Sonnet: the everyday default — strong quality at a good speed for most writing.
• Claude Opus: reach for it on nuanced, high-stakes copy — a sales page, a delicate email, anything where tone and reasoning matter more than speed.
• Claude Haiku: quick drafts, bulk variations, or when you just need a starting point to react to.
• GPT-4o: OpenAI's flagship — strong all-around performance, especially for structured tasks, summaries, and content that benefits from a broader knowledge base.
• GPT-4o Mini: fast and cheap — good for short drafts, rewrites, and bulk generation where you need decent quality at low cost.
• GPT o3-mini: OpenAI's reasoning model — useful when the task needs step-by-step thinking, comparison, or a careful breakdown rather than a single draft.

“Pick the model by what the task is worth. Claude Opus for the page that has to convert; Claude Haiku for the ten subject lines you'll throw nine of away — but always verify specific claims before publishing.”— WorkCrafter

Write a prompt that gets results

The model answers what you ask, so the quality of the ask sets the ceiling. A strong prompt names four things: who it's for, what it's for, the tone, and the format.

Write a 150-word product description for a stainless-steel water bottle.
Audience: outdoorsy 25–40 year-olds.
Tone: confident, warm, no hype.
Format: one short paragraph + three bullet benefits.

Notice what that prompt does: it bounds the length, fixes the audience, sets the tone explicitly, and dictates the shape of the output. Vague prompts ("write about my bottle") land on the average of everything the model has seen — which reads like every other bottle description.

The four modes

• Write: produce new content from a brief.
• Rewrite / improve: paste your own text and have it clarified, tightened or re-toned without losing your meaning.
• Summarize: distil a long passage into its key points.
• Brainstorm: get a diverse, numbered list of ideas to react to.

Rewrite is the most underused. Writing a rough draft yourself and letting the tool polish it keeps your voice and your facts while removing the friction — often better than generating from scratch.

Tips that consistently help

• State the length. "A short paragraph" or "120 words" prevents rambling.
• Give one example of the tone you want if you can — it's worth more than three adjectives.
• Ask for structure (bullets, headings, steps) when you'll scan the output rather than read it.
• Regenerate once or twice and keep the best; the second attempt is often stronger because you'll refine the prompt.
• For facts and figures, verify — the tool writes fluently but doesn't check reality for you.

The four things a good brief always states

A weak prompt is rarely too short — it is unspecified. The model fills every gap you leave with the most average answer available, which is exactly why generic prompts produce writing that reads like everyone else's.

1. Audience — "for finance directors at mid-size firms" changes the vocabulary, the examples and the assumed knowledge.
2. Purpose — is this meant to inform, persuade, reassure or sell? They are different pieces of writing.
3. Tone — name it, and name what to avoid: "confident, plain, no marketing superlatives".
4. Format and length — "120 words, one paragraph then three bullets" produces something you can paste; "write about X" produces an essay you have to cut.

If you only add one of these, add the format. It is the constraint that most reliably turns a wall of text into something usable.

Rewrite beats generate, more often than people expect

The instinct is to ask for a finished piece from nothing. But a messy draft in your own words, handed over for tightening, keeps your voice, your facts and your argument — and removes only the friction. It is also faster: you spend two minutes writing badly instead of ten minutes prompting toward something that still sounds generic.

This is especially true for anything where being wrong matters. The model cannot invent your pricing, your customer's objection or last quarter's number, but it can make your clumsy sentence about them clear.

Editing the output like an editor

• Cut the opening sentence. AI drafts almost always warm up before they start.
• Delete hedges — "it's worth noting", "in today's landscape", "can help you to".
• Replace any claim you cannot verify with one you can, or cut it.
• Read it aloud. Anything you stumble over is a sentence to rewrite.
• Check every number, date and name. Fluency is not accuracy.

Where the output should never ship unchecked

The model writes confidently about things it does not know. Statistics, citations, quotes, legal or medical specifics, and anything about your own business are the categories where a plausible sentence can be entirely wrong. Treat those as a draft to verify, never as a source.

A realistic article workflow

1. Brainstorm angles, then pick one — do not accept the first.
2. Generate an outline and edit it yourself. This is where the piece is actually decided.
3. Write each section separately, so nothing drifts off-topic.
4. Rewrite the whole thing for one consistent voice.
5. Fact-check every specific claim before publishing.

What it costs

A text or code generation costs 2 credits — about $0.04 at the $5 starter rate, and less on larger packs. Failed generations are refunded automatically, so experimenting is cheap.

Frequently asked questions

Which model should I default to?

Automatic. It picks a good model for the task, and you can override it to Claude Opus when a specific piece really matters or Claude Haiku when you're generating in bulk.

Why did my draft drift off-topic?

Usually the prompt tried to do too much at once. Split it: generate the outline first, then each section separately. Shorter, single-purpose prompts stay on track.

Can I trust the facts it writes?

Treat them as a draft to verify, not as sources. The tool is excellent at language and structure and unreliable at specific numbers, dates and citations — check those yourself.

Start writing

Open the text tool, start on Automatic, and give it a brief with an audience, a tone and a format. Iterate one variable at a time and you'll have a repeatable process in an afternoon.