TaskdriverWorkflow ReviewKnow where you stand
How good is
your AI setup?
Ask ChatGPT to review your AI workflow. You get a score out of 100, what makes your setup special, what you would lose if you reset it to the defaults, and the three best things to improve next.
Use it in ChatGPT
Workflow Review is built to run inside ChatGPT, and there is no form on this page. It is waiting for review by OpenAI, so it is not in ChatGPT’s directory yet. There is nothing to install or sign in to today.
Once it is approved, search for “Taskdriver Workflow Review” in ChatGPT and ask: “Review my AI workflow setup.”
Only the options you pick are sent · Your instructions and memory are never sent to us · No sign-in
The scoring sheet
Exactly how your score is worked out
Each of the seven areas earns a level from 0 to 4. Add the seven levels (28 at most), divide by 28 and multiply by 100. That is the whole formula: the same answers always give the same score, and no AI model marks it.
Custom instructions
| What you have | Level |
|---|---|
| No custom instructions | 0 of 4 |
| A short note about you or your tone | 1 of 4 |
| Detailed rules for how it should work | 3 of 4 |
| Layered rules, with different ones for each project | 4 of 4 |
Memory and context
| What you have | Level |
|---|---|
| Memory is off | 0 of 4 |
| Memory is on and left to fill by itself | 1 of 4 |
| Memory you tidy and curate | 3 of 4 |
| A knowledge base kept outside the assistant | 4 of 4 |
Connected apps
| What you have | Level |
|---|---|
| No connected apps | 0 of 4 |
| 1 to 2 connected apps | 1 of 4 |
| 3 to 5 connected apps | 2 of 4 |
| 6 or more connected apps | 3 of 4 |
Add 1 if any connected app can create, send or change things (not only read), up to a maximum of 4.
Reusable workflows
| What you have | Level |
|---|---|
| None | 0 of 4 |
| 1 to 2 projects, custom assistants or saved prompts | 1 of 4 |
| 3 to 9 | 2 of 4 |
| 10 to 29 | 3 of 4 |
| 30 or more | 4 of 4 |
Automation
| What you have | Level |
|---|---|
| Nothing automated | 0 of 4 |
| Jobs you start by hand | 1 of 4 |
| Jobs that run on a schedule | 3 of 4 |
| Assistants that work unattended | 4 of 4 |
Checking the answers
| What you have | Level |
|---|---|
| You rarely check the answers | 0 of 4 |
| You sometimes check | 2 of 4 |
| You always check the sources or test the result | 4 of 4 |
Using more than one assistant
| What you have | Level |
|---|---|
| One assistant | 0 of 4 |
| 2 to 3 assistants | 2 of 4 |
| 4 or more assistants | 3 of 4 |
Add 1 if you have a rule for which assistant gets which job (and use 2 or more), up to a maximum of 4.
The four bands
- Out of the box0 to 24
- Customised25 to 49
- Advanced50 to 74
- Expert-built75 to 100
The ranking
How the rank works
You are ranked against other people who used this review, and nobody else. It is not a survey of everyone using AI, so treat it as a rough guide.
- The number of setups is always shown. Under 100 setups, the result is labelled an early sample, because it can swing a lot.
- One score per person per day. We tell people apart by the anonymous user id ChatGPT sends, or by the network connection when there is none. If you check again the same day, your new score replaces your old one, so nobody can pile up the ranking.
- Anonymous. We keep the score, the seven levels and the date. A scrambled version of that id (or address) is kept for about 2 days to apply the one-per-day rule, then erased.
- Ties are shared. People with exactly your score count as half below you.
Want the gaps closed for you?
Taskdriver sets up AI assistants, connects them to the tools a small business already uses, and keeps them running. If your review shows gaps, we can talk through which ones are worth fixing first.
See Taskdriver’s AI services