Create Stuff: From Blank Page to Finished Output

A year ago, a decent first draft took hours whether it was a presentation, blog post, sales email, set of social assets or a video script. You sat down, stared at the blank page and in my case, stared at it a bit longer. Then you typed, deleted and started again, that was the job.

That job has completely changed. “Create Stuff” is the first category in the Task Framework because it’s where most people start with AI. It’s also where the biggest misunderstanding sits.

What “Create Stuff” actually means

Create Stuff is any task where you go from nothing to something. A blank page to a finished output. Writing, images, video, audio, presentations, design assets. If it didn’t exist before the prompt and it does after, that’s it, simple.

Well, clearly it was never going to be that simple, because “create” doesn’t mean “press a button and accept the result.” If you’re going to use AI like a vending machine you’ll enjoy the convenience but will create something that is pretty standard. It’s always best to think about it as a collaborator in a process that still depends on your judgement at every stage.

Most knowledge work has three layers. There's the thinking (what should we make and why), the making (actually producing the thing) and the judging (is it good enough). AI has essentially collapsed the cost of the middle layer. First drafts of copy, decks, frameworks, workshop plans all of it can now be generated in seconds.

But that doesn't make everything easier, it just makes the top and bottom layers harder to ignore. If anyone can produce a draft, the value shifts to the people who know what's worth drafting and the people who can tell whether the draft is any good. Strategy and editorial judgement. The bookends.

This is what makes this classification of generative AI, Create Stuff, so interesting for teams. They handle the production layer, the bit that used to consume hours of skilled people’s time but was never really where those people added their distinctive value.

When you free up the middle, you don’t lose expertise you, liberate it and concentrate it where it matters.

The tools to use right now

This moves fast. What follows is a snapshot of what’s good in April 2026, not a fixed recommendation. The categories matter more than the tools.

Writing:Claude and ChatGPT are the two serious options for long-form and strategic work. Claude tends to produce more natural prose and holds context better across longer documents. ChatGPT is quicker for short-form and integrates more easily with other tools. Jasper and Copy.ai serve teams producing on-brand content at scale. For most people reading this, Claude or ChatGPT will cover 90% of use cases. Here at Jointly we’re deeply into Claude but as much for Claude Code and agentic AI as for writing, but more of that another time.

Images:Midjourney v7 produces the best-looking results. ChatGPT's image tool (GPT Image 1) isn't as polished but it's the easiest to use as you just describe what you want and refine in conversation. Adobe Firefly is worth knowing about if you need commercially safe images or already work in Adobe. Canva is the fastest route to mockups and social content.

Video:This is where 2026 feels different. Runway, Sora 2, Veo 3.1 and Kling can all generate usable video from a text prompt. Runway is the most production ready. Sora looks the best but is less reliable. Kling is the budget option, especially good for social content. For talking-head videos, Synthesia and HeyGen use AI avatars to produce training and internal comms material that would previously have needed a camera crew. None of these replace proper production for high stakes work, but they've opened up a whole tier of video content that simply didn't exist before

Audio: Voice cloning and text-to-speech have quietly got very good. ElevenLabs leads on quality, you can generate natural sounding voiceovers in minutes, which changes the economics of podcasts, product videos and training content. Descript takes a different angle, it lets you edit audio by editing a text transcript. Delete a sentence from the transcript and it disappears from the recording. It still feels like it shouldn't work until you try it. NotebookLM from Google is also worth a look it can turn documents into surprisingly natural podcast-style conversations. Try it instead of sharing that 80 page deck. 

Presentations:Still the weakest category. AI can generate slides but rarely something you'd actually stand up and present without serious reworking. The layouts tend to be bland, the visual hierarchy is off and the storytelling is flat. The best all-round option right now is Gamma, which gets closer than most to something usable. But honestly, the smarter approach is to split the job: use Claude or ChatGPT to build your narrative and structure, then design the slides yourself or hand it to Canva. AI is good at thinking through what your deck should say but it’s not good at making it look like you care. Yet.

How to prompt for creation

Most prompts fail because they’re under-briefed. “Write me a blog post about leadership” produces something written by nobody, for nobody. That’s not an AI problem, it’s a briefing problem. If you participate in one of our AI Immersion sessions something you’ll be given up front is the BRIEF framework for writing prompts that actually work, because AI only as good as the brief you give it and the better the brief the less time you spend fixing what comes back.

Background -What’s the situation? Set the scene. Who’s the audience? What’s already happened? What constraints exist? Give the AI the context it would get if it sat next to you for a week.

Role - Who should the AI be? Tell it what expertise to draw on. This shapes vocabulary, depth and perspective. You’re casting the right person for the job.

Intent - What exactly do you want? Be specific. Not “help me with a presentation” but “write five slide titles that tell the story of why we’re launching this product now.” The clearer the ask, the less you’ll need to fix.

Examples - What does good look like? Show the AI a reference point. Paste in something you liked, or describe what to avoid. Examples do more work than adjectives. “Make it punchy” is vague. Showing a punchy paragraph is specific.

Format -How should the output land? Specify structure, length, tone and where this will end up. A Slack message and a board paper need different formats even if the content is the same.

Then, to really create stuff that’s usable accept that the first response is a draft, not the answer. Get into the habit of challenging or redirecting your AI to sharpen or extend what’s been created.

Iterate, don’t accept.The first output is a draft. Push it. “Too formal.” “Weak opening.” “Cut this in half.” Version two and three are where it gets good.
Feed it your thinking first.The best outputs start with your ideas, even if they’re messy. Bullet points, voice notes, rough paragraphs. Give it something real to work with and it sounds like you, not everyone else.
Use it where you’re weakest.Everyone has formats they avoid. If you’re strong at long-form but bad at social, that’s where AI earns its keep. Let it carry the work you’d otherwise put off.

Becoming fluent in Create Stuff

Fluency isn't a technical skill. It's not about knowing which buttons to press or how to write the perfect prompt, it’s judgement, a feel for when AI accelerates your work and when it gets in the way.

A useful rule of thumb is use AI for the things you're weakest at. If something needs your voice? Write it yourself first then let AI clean it up. If it's fifty social posts and they all need to exist by Thursday? Let AI draft them. Need images for a deck and you can't draw. Go for it. Tricky email to a difficult stakeholder? Do that one yourself. The skill isn't using AI more. It's knowing when to put it down.

The people getting the most from Create Stuff aren’t the ones with the smartest prompts, they’re the ones who reach for it at the right moment and apply judgement and taste to what comes back. They’ve stopped treating it like magic and started treating it like a tool, which is when it creating stuff feels natural.

Next time:Find Stuff— how AI changes the way you search, filter and surface what matters. This is the category where most people are still using AI like a slightly better Google.