03 — Decision tools

AI agent vs workflow automation — which do you need?

Not every process needs AI, and not every process can be handled by fixed rules. This tool looks at your inputs, variety and risk to recommend the right kind of system.

6 questions
Free · no sign-up
Runs in your browser

How to use · 4 steps

Swap in your own numbers anytime

01Pick one workflowThink of a single process, e.g. handling inbound emails or processing invoices.
02Answer six questionsAbout its input, variety, volume, risk, judgement and how often rules change.
03Read the fitEach approach gets a fit score. The highest one is the recommendation.
04See the exampleEach approach shows a concrete example of what that system would do.

Describe the workflowEditable

0/6

1
What does the input look like?
2
How different is each case?
3
How often does it happen?
4
What does a mistake cost?
5
Does it need rules or judgement?
6
How often do the rules change?

Best fitLive

Start answeringto build your result.

Each answer shifts the fit between the three approaches.

Fit by approach

Rule-based automation—
AI agent—
Keep it with a person—

Each approach as a share of the points it could have earned from your answers.

Rule-based automationExample · New form submission → create CRM record → send confirmation → notify the owner.

AI agentExample · Read an inbound email, work out what's being asked, pull the order, draft a reply, and flag refunds over a limit.

Keep it with a personExample · An AI summary of the case file and a drafted response, with a person making the decision.

Finish to unlock your resultOnce it's complete you can send it to us or share it as a link.

How it's worked out

No black box.Every step, in plain words.

Results are estimates for planning, based on the figures you enter. They aren't financial advice or a quote.

01
Each answer adds points to one or more of three approaches: AI agent, rule-based automation, or keep manual.
02
Each approach is shown as a share of the points it could have earned. The highest share is the recommendation.
03
When agent and automation are close, the result suggests a combined design — automation for the fixed path, an agent for the exceptions.

Questions

Good to know.

Automation follows fixed rules you define in advance. An AI agent reads unstructured input, decides what to do within limits you set, and escalates when it's unsure.

When volume is low, every case is different and mistakes are expensive. In those cases, tools that support the person usually beat tools that replace them.

Read · 4 minAI agents vs workflow automation: which does your process actually need?Read the guide ServiceAI Agent DevelopmentHow we build it

More tools

Keep going.

All 9 tools

8 questions

Ship or skip

Score a product or feature idea on market signal, feasibility, time to revenue and execution risk. Get a clear verdict and the reasons behind it.

Open tool

~3 min

Build vs buy

Compare building custom software with buying an off-the-shelf tool — on fit, control and three-year cost — and get a Build, Buy or Hybrid recommendation.

Open tool

~2 min