When you put someone in a job, you make sure they can do it. They have the background, the tools and the training. ISO calls that competence.
Bots need it too, but it comes from different places.
Competence in ISO
ISO asks an organization to:
- decide what competence each role needs
- make sure people have it, through education, training or experience
- take action to close gaps
- keep evidence of competence
Where a bot's competence comes from
| For a person | For a bot |
|---|---|
| Education and background | The model it runs on |
| Tools and equipment | Its tools and connectors |
| On-the-job training | Its skills and instructions |
| Experience | Its saved memory and records, plus testing |
| Training records | Test results and version history |
The model: background
The model is like the education someone brings. A stronger model handles harder reasoning, longer documents and trickier judgment calls. A lighter one is fine for simple, repeated tasks, and costs less. See Cost and model choice.
Models also come with settings. xAI's docs, for example, list a reasoning setting for Grok 4.7 that goes from low to xhigh (docs.x.ai). More reasoning suits harder problems.
The tools: equipment
A bot without the right tools can't do its job, however smart its model is. A bot that's supposed to check the calendar needs calendar access. A bot that researches needs search.
And the reverse: a bot shouldn't have tools its job doesn't need. See Tools and connectors.
The skills: training
Skills are the on-the-job training. They tell the bot how your business does things. A strong model with no skills is like a talented hire on day one: capable, but not yet useful for your work. See Work instructions are skills.
Evidence of competence
In ISO, you keep training records. For bots, the evidence is:
- Test results showing the bot handles its job. See Your tester bot is your internal auditor.
- The version of its skills and its model when it passed.
If you change the model or the skills, the old test results don't fully count anymore. Retest.
Onboarding
Many bot setups use a getting-started skill that runs right after setup: it learns the owner's preferences and walks through the basics. In our review of 90 marketplace templates, 26 had one (good-practice.md). It's like the first-day orientation.
Where bots differ
A person's competence grows with experience on its own. A bot's mostly doesn't. It grows when someone updates its skills, gives it better tools or moves it to a better model. Competence for a bot is something you build on purpose.
What this means for you
- For each bot, write down: model, tools, skills.
- Check each against the job.
- Test the bot, and keep the results with the version.
- Retest after changing any of the three.
A competent bot isn't just a smart model. It's the right model, with the right tools, trained for the job.