Question 1

Is my data being used to train the AI, and can I turn it off?

Accepted Answer

Whether AI trains on your data depends on the plan, not the tool. On business and team plans, the major vendors don’t use your inputs, outputs or connected data for training, and that’s the default. On personal accounts it’s a setting each person can change. Check which plan your team actually uses before anything else.

Question 2

Is it safe to put client information into an AI tool?

Accepted Answer

Putting client information into an AI tool can be safe, on the right plan and within your contracts. 3 things decide it: whether the vendor trains on your content, how long it keeps it, and whether your client agreements allow it. On a personal account with default settings, the safe answer for client information is no.

Question 3

Is it changing anything in the original files?

Accepted Answer

AI reading your files doesn’t change them. It can only alter an original if it has been given permission to write, and a well-built system isn’t. In our builds the originals are never touched: the system works on a copy or writes a new output, and that output is checked against the source.

Question 4

We suspect a lot of shadow AI and have no policy. Should the policy come first?

Accepted Answer

Your company AI policy shouldn’t come first: find out what people already use, without judgment, and write the policy from that. A policy written from a template, or from fear, covers the wrong things and gets ignored. The use is already happening, so start by seeing it.

Question 5

Do we have to tell clients when AI helped produce their work?

Accepted Answer

Whether to disclose AI use to clients is a policy decision for now, because there’s no fixed rule yet. Check the client contract first: some require disclosure and some prohibit AI use altogether. A workable position is that AI is used as a reviewed tool, a person signs off the work, and client data is protected.

Question 6

Who owns what the AI produces?

Accepted Answer

Between you and the vendor, AI-generated content is normally yours: the major tools’ terms assign the output to the customer. Whether it can be protected by copyright is a separate question, and in the US that generally needs human authorship. Read the terms for the plan you’re on, and ask your lawyer about anything valuable.

Question 7

Do I have to upload the report every month, or does it remember?

Accepted Answer

An AI assistant remembers its standing instructions, and this month’s data is handed over each run. A project keeps its instructions and whatever files you put in it, and it only sees what’s there. It doesn’t reach into your other chats or go looking for the new report on its own.

Question 8

If I connect a tool, does the AI get free rein to read everything in it?

Accepted Answer

Connecting a tool doesn’t give the AI free rein: it reads what you point it at, when you ask, within the permissions you already have in that tool. It isn’t scanning or storing the whole mailbox. What a connector can do, as opposed to see, is a separate setting.

Question 9

What does connecting my email or spreadsheets actually unlock, compared with copy-paste?

Accepted Answer

Connecting your email or spreadsheets lets you refer to your things instead of pasting them in: “reply to that email”, “the totals from this month’s sheet”. It reads on demand, keeps the link to the source, and drafts. It doesn’t send. The gain is fewer steps and an answer you can trace back.

Question 10

I just have to trust it. I can trust it, right?

Accepted Answer

You can trust AI output only as far as you can check it. A person reviews exactly the outputs that would hurt if they were wrong, edge cases are flagged rather than quietly decided, and a human hand stays on the commit. Trust comes from being able to check its work.

Question 11

How do I know it didn’t miss something?

Accepted Answer

To check AI output for misses, look at what it read as well as what it found. You’re missing some things today, too, so perfection isn’t the bar. A run should state what it looked at, so a person can read it and see the gaps.

Question 12

How accurate is AI?

Accepted Answer

AI accuracy has no single number: it depends on the task and on what the AI was given to work from. The useful comparison is with today. It doesn’t have to be perfect, it has to beat your current miss rate, and its misses have to be catchable.

Question 13

Why does it solve my problem sometimes and completely miss other times?

Accepted Answer

When AI is inconsistent, the cause is almost always the context it was given. Cold, it guesses how your firm works. Given a map of it, it reasons. That map is usually a document nobody has written yet: how the process runs, what the terms mean, and what a good result looks like.

Question 14

Could AI draft our technical drawings?

Accepted Answer

AI can’t reliably draft technical drawings today. It handles text, analysis and summaries well, but it doesn’t produce CAD or survey drawings a professional could sign. The work around the drawing, such as specifications, reports and checks against the standard, is where it helps now.

Question 15

What does an AI consultant actually do?

Accepted Answer

A good AI consultant does 3 things: finds where AI would pay back in your firm, puts those options in order, and tells you what to leave alone. What you get should be a short, ranked plan grounded in how your work actually runs, with numbers behind each line.

Question 16

How do we choose an AI consultant?

Accepted Answer

To choose an AI consultant, ask 2 things: what they last recommended against, and who looks after the build a year later. The first tells you whether they’ll be honest about payback. The second tells you whether you’re buying something that lasts or something delivered and left.

Question 17

Should we build AI ourselves, buy a product, or bring someone in?

Accepted Answer

Build or buy AI based on fit, and on who will own it. Buy when a product already handles the work your way. Build when someone on your team can, and will be there to maintain it. Bring in a partner when the work crosses systems, needs judgment at a step, or nobody has time to own it.

Questions clients ask us about AI

Where your data goes after you type it in.

Is my data being used to train the AI, and can I turn it off?

Is it safe to put client information into an AI tool?

Is it changing anything in the original files?

Rules your team will actually follow.

We suspect a lot of shadow AI and have no policy. Should the policy come first?

Do we have to tell clients when AI helped produce their work?

Who owns what the AI produces?

What it keeps, and what you hand it each time.

Do I have to upload the report every month, or does it remember?

How far a connected tool can reach.

If I connect a tool, does the AI get free rein to read everything in it?

What does connecting my email or spreadsheets actually unlock, compared with copy-paste?

Checking it before you rely on it.

I just have to trust it. I can trust it, right?

How do I know it didn’t miss something?

How accurate is AI?

When the honest answer is “not yet”.

Why does it solve my problem sometimes and completely miss other times?

Could AI draft our technical drawings?

Bringing in help, or building it yourself.

What does an AI consultant actually do?

How do we choose an AI consultant?

Should we build AI ourselves, buy a product, or bring someone in?

Different firms,
the same questions.

Answered before you have to ask.

Questions about working with Milo Works

What brings you here?

A question that isn’t here?

Is my data being used to train the AI, and can I turn it off?

Is it safe to put client information into an AI tool?

Is it changing anything in the original files?

We suspect a lot of shadow AI and have no policy. Should the policy come first?

Do we have to tell clients when AI helped produce their work?

Who owns what the AI produces?

Do I have to upload the report every month, or does it remember?

If I connect a tool, does the AI get free rein to read everything in it?

What does connecting my email or spreadsheets actually unlock, compared with copy-paste?

I just have to trust it. I can trust it, right?

How do I know it didn’t miss something?

How accurate is AI?

Why does it solve my problem sometimes and completely miss other times?

Could AI draft our technical drawings?

What does an AI consultant actually do?

How do we choose an AI consultant?

Should we build AI ourselves, buy a product, or bring someone in?

Different firms,the same questions.

Answered before you have to ask.

What brings you here?

A question that isn’t here?

Different firms,
the same questions.