Compare
Probative vs ChatGPT
ChatGPT will write you an award submission in about four minutes. The question is what you had to do before those four minutes, and what you have to do afterwards.
What ChatGPT is doing
It writes what you tell it. You supply the case, the evidence and the emphasis, and it produces fluent prose from what you supplied. That is genuinely useful, and if your problem is the blank page it will solve it today for almost nothing.
The three things you still have to do yourself
- Decide the case. Which project, which angle, which of your four possible stories the evidence will actually carry. It will argue whichever one you hand it, competently. Choosing it is still your job, and it is the part that decides the result.
- Check every claim against a source. Paste your documents in and tell it to stay inside them and a careful operator gets a long way. What you do not get is a record that outlasts the session: a link from each claim to the document it came from, made the same way every time, on the day somebody questions a number.
- Notice what is missing. A gap gets written around smoothly unless you already knew it was there. Working out which absent fact would have changed the entry means reading everything against the criteria first, which is the work you were hoping to skip.
The difference is not that a general model is incapable. It is that with ChatGPT you assemble and supervise the process yourself every time, and the entry is only as good as how well you did that on the day. Here the process is the product.
You paste in some rough notes, it produces a confident paragraph, and somewhere in that paragraph is a specific-sounding claim that came from the model’s sense of what usually goes there rather than from anything your firm can evidence. It reads better than the truth, which is exactly why it gets left in.
Side by side
| ChatGPT | Probative | |
|---|---|---|
| Writes fluent prose | Yes | Yes |
| Cost | Near zero | $1,125 per entry |
| Works from your raw project documents | Only what you paste | Yes, all of it |
| Models the published award criteria | If you paste them | Yes |
| Decides which case to make | No | Yes |
| Links every factual claim to a source | No | Yes |
| Refuses claims it cannot support | No | Yes, and lists them |
| Tells you what evidence to go and get | No | Yes |
When ChatGPT is the right answer
If you have already decided the case, already have the evidence in front of you and you just want help with the prose, use ChatGPT. It is cheaper than us and it will do that job well.
Start with two answers.
Send an award you are actually entering and whatever project material you already have. You get two complete answers, the case we would make, the questions we would send your team and the claims that one more fact would unlock.
Credited in full if you go ahead. If you read it and walk away, you owe nothing further.