We were using AI every day and still getting average results

The uncomfortable realisation

Like most builders, my co-founder and I use AI constantly. It’s part of our daily workflow. We use it for writing, brainstorming, structuring ideas, debugging, and research. On paper, it felt like we had access to incredible leverage.

And yet, something felt off.

Everyone online kept saying, “AI is insane.” Timelines were full of mind-blowing outputs. But when we observed how most people were actually using these tools, the pattern was different. They would open ChatGPT, type one vague sentence, hit enter, and expect magic.

When the output was generic, the conclusion was predictable:
“The model isn’t that good.”

The more we examined this, the clearer it became that the model usually wasn’t the bottleneck.

The prompt was.

The real problem: articulation

Most people don’t struggle with ideas. They struggle with articulation.

There’s often a clear outcome in your head, but translating that into structured, detailed instructions is hard, especially when you’re moving fast and just want a result.

We noticed a consistent loop:

• vague input

• generic output

• frustration

• blame the tool

But AI systems are extremely sensitive to input quality. Small improvements in clarity can create disproportionately better results.

The gap wasn’t intelligence.
It was the structure.

Our thesis

Instead of building another AI wrapper or a new model, we asked a simpler question:

What if there were a thinking layer before the model?

a system that helps users clarify intent, add missing context, and structure their request properly before sending it anywhere.

not more AI.

better inputs.

Building Promptly

promptly starts with a rough idea. a half-baked thought. something you might otherwise throw directly into ChatGPT.

Instead of sending it immediately, promptly guides you through a short process that surfaces the details that matter. It asks clarifying questions, extracts context, and turns your messy input into a structured, optimised prompt.

You can then use that prompt in any major model.

We’ve seen firsthand how much difference this makes. The same model, given better instructions, produces dramatically better output. fewer generic responses. less back-and-forth, more usable results on the first attempt.

We’re not trying to replace AI tools. We’re trying to make them more effective.

Who this is for

promptly isn’t for everyone.

It’s for people who:

• Use AI daily for real work

• care about output quality,

• know what they want, but struggle to phrase it

• are tired of generic responses

builders. marketers. writers. students. operators.

If AI is part of your workflow, the quality of your prompts directly impacts your results.

Why this matters long term

models will keep improving. That’s inevitable.

But as models become more powerful and more commoditised, the advantage shifts toward those who can think clearly and instruct precisely.

Prompting isn’t about tricks.

It’s about clarity.

and clarity compounds.

What’s next

We just launched v1, and we’re building in public.

right now, we’re closely watching how people use promptly. where they hesitate. What confuses them. What genuinely improves their workflow.

If this resonates with you, try it. stress test it. break it. Tell us what feels unnecessary and what’s missing.

And if it meaningfully improves your AI workflow, consider upgrading and supporting us.

We’re excited to see how far better inputs can take us.