Free tools/AI Workflow Complexity Score

Score how complex an AI workflow is to build.

AI scores 7 build dimensions 1–5 · blends to a Low / Medium / High band

Describe one task the way you'd explain it out loud and let AI score each of the 7 things that drive build cost — the data, the cleanup, the sources, the decision, storage, the cost of being wrong, and checking — then blend them into a Low / Medium / High complexity band. Adjust any score yourself; the total updates live.

Describe the workflow — let AI score it, then adjust.

Replay the last time you actually did this task: what you started with, the systems you touched, the calls you made, what happens if it's wrong. The more detail, the sharper the score. (Or skip this and rate the seven stages yourself.)

What lands in front of you when the task begins?

Simple · 1–2
A clean web-form submission, or one tidy spreadsheet — all text, predictable fields.
Complex · 4–5
Call recordings + scanned PDFs + free-text emails, with missing or inconsistent fields.
SimplerMore complex

How much fixing/tidying before the data is usable?

Simple · 1–2
Use it as-is; maybe map a couple of fields.
Complex · 4–5
Match the same customer across CRM + billing, dedupe, fix names, and fill blanks first.
SimplerMore complex

How many places do you pull from in one run?

Simple · 1–2
One source — a single form, inbox, or system.
Complex · 4–5
CRM + email + a calendar + a phone system + a website you click through.
SimplerMore complex

How much judgment and how many steps to reach the answer?

Simple · 1–2
One obvious step — summarize the note, or copy a field across.
Complex · 4–5
Branching judgment — route by intent, weigh edge cases, escalate when unsure.
SimplerMore complex

What has to be saved — including half-finished work?

Simple · 1–2
Nothing, or one row in a sheet.
Complex · 4–5
Transcripts, versions, and a searchable history other people rely on.
SimplerMore complex

If it's late, wrong, or inaccurate — how bad is that?

Simple · 1–2
You notice and fix it in seconds; no real harm.
Complex · 4–5
A wrong or late output loses a sale, breaks trust, or exposes private data.
SimplerMore complex

Who can tell if the result is good, and how?

Simple · 1–2
You glance at it and instantly know if it's right.
Complex · 4–5
It takes an expert and several reviewers, plus stored good/bad examples to judge quality.
SimplerMore complex
Score by stage
The data you start with3/5
The cleanup before you can act3/5
Where the information comes from3/5
The decision you make3/5
Where things get stored3/5
The cost of getting it wrong3/5
Checking the output3/5
Blended complexity
60/100
Medium complexity

Medium → scope it properly and keep a human in the loop while it beds in.

Get this as a branded one-pager

A clean PDF you can save or send to your team. We'll email you occasional, useful AI notes — no spam.

How to use it.

1. Describe one real task

Not a department — one task you'd want off your plate, like "after a call, write up the notes and update the CRM." Replay the last time you actually did it: what you started with, the systems you touched, what happens if it's wrong. The more detail, the sharper the score.

2. Let AI score the 7 dimensions

AI reads your description and scores each stage 1 (simple) to 5 (complex) with a one-line rationale: how messy the data was, how much cleanup it needed, how many sources you pulled from, how much judgment the decision took, what had to be stored, how bad it is if it's wrong, and who can judge the output. Each stage shows a concrete simple-vs-complex example, and you can override any score — the blended total updates live.

3. Read the complexity band

Your answers roll up into Low, Medium, or High. Low is a fast prototype; High needs real engineering — integrations, guardrails, reliability, a deployment story. It's the honest version of "could we vibe-code this in a weekend" versus "this needs the tech team."

4. Feed it into your roadmap

The band maps directly to build effort. Drop it into the AI Roadmap Generator as the complexity input and it becomes the scheduling weight — so a backlog of scored tasks turns into a realistic, capacity-aware plan.

Frequently asked questions.

What actually makes an AI workflow complex to build?
Rarely the model — it's everything around it: messy data from several places, brittle hand-offs between systems, a decision that takes real judgment and has lots of exceptions, and a high cost if it gets things wrong. A task with one clean input, one obvious step, and no real downside is easy; one that pulls from five systems, weighs judgment, and must never be wrong is hard.
Which of the 7 questions matter most?
The decision you make and the cost of getting it wrong carry the most weight, because that's where AI projects actually stall — judgment-heavy work with exceptions, and workflows where a mistake is expensive or sensitive. A task can pull from many sources and still be simple if the decision is obvious and a wrong answer is harmless.
Does a high complexity score mean don't build it?
No — high complexity often means high value, since the hard workflows are the ones competitors can't quickly copy. It means budget for real engineering, not a weekend prototype, and don't hand a High-complexity build to someone without the support to take it to production.
What's the difference between complexity and whether it's worth automating?
Complexity is how hard it is to build; worth is whether the payoff justifies it. A workflow can be simple but low-value, or complex but transformative. Score complexity here, score the payoff with the ROI calculator, and only the workflows that clear both bars belong on your roadmap.

More free AI tools.

Numbers looking promising?

A free tool gives you a hypothesis. The 30-minute diagnostic is where we pressure-test it against your actual workflows — and decide whether the project is worth building, buying, or skipping.

Book a 30-min diagnostic →