Before You Automate
A 15-point AI readiness checklist for small companies. Find out whether your team has the process, knowledge, quality controls, safety rules and ownership needed to build its first useful AI agent.
Your team may already use AI - without having an AI system.
Five signs of scattered AI use
These signs show where the team needs shared context, operating rules and ownership before automation can scale.
Do not start with a tool. Start with one workflow.
Your first pilot should be narrow enough to inspect, useful enough to matter and frequent enough to evaluate within a month.
Which repeated task consumes predictable time every week?
Who understands the task well enough to explain its exceptions?
What does a good result look like - and how will you recognise it?
What must the agent never send, change, delete or decide alone?
The 15-point AI readiness checklist
Tick what is genuinely true today. Your score will update automatically and show the most sensible next step.
Process
A bounded piece of work with a visible outcome.
Company knowledge
The context and rules that make your way of working specific.
Quality
Evidence that the system works beyond an impressive demo.
Safety
Limited permissions and explicit human control.
Adoption
A person and a review cycle around the technology.
Start with the foundation
Choose one workflow, name its owner and gather the company knowledge needed to perform it well.
Use the score to identify the next concrete step: build the foundation, run a controlled pilot or start developing the first working agent.
A realistic first 30-day pilot
The goal is not autonomous work everywhere. It is one bounded workflow that the team can observe, test and improve.
Choose
Select one workflow, name the owner, record the baseline and collect real examples.
Structure
Prepare company context, sources of truth, definitions, rules, exceptions and approval boundaries.
Build & test
Create the first narrow agent and test it manually against known cases. Record failures, not just successes.
Run & decide
Use it with limited access. Review quality, time saved, errors and team adoption before expanding.
What not to do first
Complexity is easy to add and expensive to remove. Earn it only when the simple version has reached its limit.
Do not try to automate the whole company at once.
Do not begin with a polished interface before the workflow works.
Do not give an experimental agent access to everything.
Do not build multiple agents before one narrow agent is reliable.
Do not measure progress by the number of prompts or tools.
Do not remove human judgement from risky decisions.
Want to start this process in your company?
Book a call with Barcelona Code School. We will review your current situation, identify a sensible first workflow and suggest the right solution - whether that is a focused workshop, a controlled pilot or a deeper implementation path.
Bring one workflow, one recurring problem or the questions your team needs help structuring.