Raising something at the front door

Some of the best ideas about AI come from the people doing the work. You know which task eats your Friday afternoon, which report gets re-typed every month, which step everyone waits on. This lesson is about what to do with an idea like that: when to just get on with it, and when to raise it so it can help more people than you.

People are already solving their own problems with AI, whether or not anyone asked them to.

Fix your own work

If a fix only touches your own work, uses tools you're already allowed to use, and costs nothing new, you don't need anyone's permission. Just do it. The framework calls this Improve. It's the most common move, and it's the one your organisation most wants you to make.

That might be a better way to draft your weekly update, a prompt that turns your own notes into a summary, or setting up the approved assistant to tidy a spreadsheet you build every month. Two conditions keep it yours: don't push extra work onto someone else, and anything you send on still meets the output standard (labelled, owned, short and checked).

And you won't be punished for it. Improving your own work is exactly what the second promise protects.

Two questions tell you when to raise it instead

Before you go further, ask two questions.

1. Would it cost money or engineering time? A new tool, a paid add-on, extra licences, or someone connecting two systems. That's Build, however small the price.
2. Would other people rely on it, or is a lot at stake? A template your whole team now uses, anything that writes into a shared system, or anything that touches customers, personal data or decisions about people. That's Own: it needs a named owner who answers for it.

If the answer to both is no, it's yours to fix. If either is yes, it goes in through the front door.

Try it on six situations from different jobs. Some are yours to fix; some need raising.

What the front door is

The front door is the one place your organisation names for raising things about AI. It might be a form, a shared inbox, a channel in your team chat, or a standing slot in a meeting. Anyone can raise anything there: a problem that keeps slowing you down, a tool you've made, or a request for something new.

Whatever you raise, you get a first answer within 10 working days, from the person who runs your organisation's AI work. There are only three answers.

• Adopt it. This is the default for something you've already made. It gets used more widely, and you get the credit. If other people will rely on it, it also gets an owner.
• Park it. It goes on the friction list (the shared list of where work waits, piles up or gets redone), with a status everyone can see. Parked isn't lost. It may be next.
• Build it. Money or engineering time is spent, but only once cheaper fixes have been ruled out: changing how the work is done, or using the tools you already have.

Why it exists

The front door solves two problems that show up in almost every organisation. The first is the idea that dies in an inbox. Someone has a good idea, mentions it to the wrong person, hears nothing, and stops suggesting things. The 10-day answer is there so that never happens quietly.

The second is the orphan: a tool someone built on the side that other people now rely on, with nobody who owns it, checks it or can switch it off. Think of a spreadsheet the whole team depends on, written by someone who has since left. When it breaks, or starts getting things wrong, nobody knows. Raising it means it gets an owner before that happens, and you keep the credit for making it.

Raise one thing

A good request is short. Describe the work, never a person. Say where it waits, piles up or gets redone. Say what you think would help, and mention anything you've already made, because that's how you get the credit. First, say where the front door is where you work. If nobody has told you yet, choose “Not told yet”, and the question to ask your manager goes on your record.