August 4, 2026 · By Ava Idris
ITT: Instructions, Tools, Triggers
ITT stands for Instructions, Tools, Triggers. An agent needs all three: standing orders it follows every run, tools that let it act, and a trigger that starts it without you. Miss any one and you have a prompt, a script, or a chatbot.
This is the test I teach first in every engagement, because it settles most arguments before they start. Almost every "is this an agent?" debate is really a disagreement about which of the three is missing.
The three parts
Each has its own entry, because each is where a different rollout goes wrong.
Instructions are the description of the job. What to do, to what standard, and what to bring to a human instead of deciding alone. The agent reads them on every run. Skip them and you get an agent with the logins and the schedule and no reason to touch either, which is the most common way a rollout produces nothing.
Tools are what give it hands. An agent with no tools can only talk. Access to a calendar, files, an inbox, whatever the job needs, is the difference between advice and action. The technical name for it is tool use. It is also where the real risk lives, which is why access gets earned rather than handed out.
Triggers are what start it without you. A time, an event, an email landing, a file appearing. This is the piece most people skip, and it is the actual line between using a tool and having something work for you.
How to use it on a demo
Sit through the pitch, then ask three questions in order. Where do its instructions live. What can it actually touch. What starts it when nobody is watching.
A real agent has an answer to all three, and the answers are specific. A very good prompt with a nice interface will answer the first vaguely and go quiet on the third.
That is not an insult and it is not a reason to walk away. It is a reason to pay prompt prices for prompt value, which is the thing most of the market is currently getting wrong.
If you want the short version aimed at what to actually learn rather than the framework itself, that's here.
Where it came from
I built this teaching people who were being sold agents and had no way to check the claim. It is deliberately small. A framework you can hold in your head during a sales call is worth more than a better one you have to look up.
If the term itself is still fuzzy, start with what an AI agent actually is and come back.
What does ITT stand for?
Instructions, Tools, Triggers. It is the three-part test for whether something is genuinely an AI agent rather than a prompt, a script or a chatbot.
What are the three things an AI agent needs?
Standing instructions it follows on every run, tools that let it act on something real, and a trigger that starts it without a person asking.
How do I tell if a product is really an AI agent?
Ask where its instructions live, what it can actually touch, and what starts it when nobody is watching. A real agent answers all three specifically.
What happens if an agent is missing one of the three?
It is still useful, it is just something else. Missing instructions makes it a prompt, missing tools makes it advice, missing a trigger makes it a tool you have to remember to open.