Your CEO has been to a conference, or read something on a train, and now the same question turns up in every meeting: “Why aren’t we using more AI?” It is a reasonable question asked in an unhelpful way, and “we are, a bit” is not an answer that survives a follow-up.
What follows is the set of questions that tend to come up, what is true in each case, and wording you can lift. Change the details to match your company. Do not change the honesty.
What the question usually means
“Use more AI” is rarely a request for a specific tool. It is usually one of three worries wearing a costume:
- Cost. Could we be doing this for less?
- Speed. Could we be doing more with the same people?
- Fear. Are competitors ahead of us?
Work out which one you are hearing before you answer. A CEO worried about cost does not want a tour of your prompt library. A CEO worried about falling behind does not want a spreadsheet of hours.
The eight questions
1. “Can’t AI just write all our content?”
It can produce all of it. Whether anyone reads it, believes it or finds it in search is a separate matter. Unedited AI copy is fluent, average and frequently wrong on details only your business would know. See how to spot AI slop for what that looks like in practice.
“It can draft all of it, and I use it for that. What it can’t do is know our customers, our products or what we actually think. Every piece still gets checked and edited by a person before it goes out, because our name is on it.”
2. “Can we cut the agency budget, then?”
Possibly some of it. AI makes production cheaper. It does not make judgement, strategy or accountability cheaper, and those are what you should be paying an agency for. If your agency is only selling you production hours, that is a fair conversation to have with them.
“Some production work should cost less now, and I’m going to ask our suppliers how AI has changed their pricing and their process. I wouldn’t cut the parts where someone experienced is checking the work and standing behind it.”
3. “What are our competitors doing with it?”
You mostly cannot know, and neither can your CEO. What you can see is their output: whether their content volume has jumped, whether it has gone bland, whether they are appearing in AI search answers. Report what is observable and refuse to guess the rest.
“I can’t see inside their process, but I can see what they publish. I’ll review the three competitors we care about and tell you what’s changed in the last year. If their output has got bigger and blander, that’s an opening for us.”
4. “Is it safe to put our data into these tools?”
It depends on the tool, the account type and the data. Free consumer accounts and paid business accounts often have different terms on how your inputs are stored and used. Customer personal data, contracts and unreleased financials should stay out unless someone has read the terms and decided otherwise. This is why you need a one-page AI policy.
“Some things are fine, some are not. I’ve written a one-page policy listing the tools we’ve approved and the data that never goes in. Customer personal data is on the never list. I’d like you to sign it off.”
5. “What’s our AI strategy?”
For a company your size, a separate AI strategy is usually the wrong document. You have a marketing plan. AI changes how some of the work in it gets done. What you need is a short statement of where you use it, where you do not, and who checks the output.
“We don’t need an AI strategy so much as a marketing plan that uses AI sensibly. Here’s where it’s saving time now, here’s where I’m testing it, and here’s what stays with a person.”
6. “What’s the ROI?”
Measure it on your own work, not on a vendor’s claims. Pick two or three recurring tasks, time them before and after, and be honest about checking time. Say a monthly report used to take you three hours and now takes one hour plus twenty minutes of checking. That is a real number you can defend. Then say what you did with the time you got back.
“On the tasks I’ve measured, here’s the time saved after checking. I’ve put that time into the things we kept postponing. I’ll report on those results rather than on the tool.”
7. “Do we need to hire someone for this?”
Almost certainly not yet. A dedicated AI hire makes sense when there is enough repeatable work to automate. A small team gets further by training the people it has and bringing in outside help for specific projects.
“Not at our size. I’d rather spend a fraction of that salary on training and on specialist help for the specific jobs where it pays back.”
8. “Will customers be able to tell?”
If nobody edits it, yes. Readers have become quick at recognising generic machine copy, and once they spot it they trust the rest of the page less. If a person who knows the business has reworked it, it reads like your company because it is your company’s thinking.
“They can tell when it’s unedited. They can’t tell when someone who knows our business has rewritten it properly, and that’s the standard I’m holding us to.”
The update to send before you are asked
The strongest position is to report first. Adapt this, fill in the brackets and send it. One paragraph is enough.
AI in our marketing: where we are. We currently use AI for [first drafts of blog posts, summarising research, meeting notes and monthly reporting]. On the tasks I’ve timed, it saves roughly [X hours a month] after checking. Everything customer-facing is reviewed, fact-checked and edited by [me / named person] before it is published, and nothing goes out without that sign-off. We do not put [customer personal data, contracts or unreleased financial information] into any AI tool, and we only use the approved tools listed in our one-page policy. Over the next quarter I’m testing AI for [one or two specific tasks] and will report back on whether it is worth keeping. The areas I’m deliberately keeping human are [positioning, customer conversations and final copy]. If you want more or less of any of this, tell me and I’ll adjust.
Every sentence in that paragraph answers one of the eight questions above. That is why it works.
What not to say
A few answers that feel safe in the moment and cause trouble later:
- “We’ll save X per cent.” You do not know yet. Promise a measurement, not a result.
- “AI can’t do that.” It probably can, badly. Say what it costs to do it well.
- “Everyone’s using it.” This invites the question of why you were slow.
- “I’ll look into it.” With no date attached, this is how the same question comes back next month.
If you want backup
If it would help to have someone outside the business explain where the line sits, that is a large part of what we do: AI-assisted work with a named person checking and signing off every piece. There is more on how that works here, or you can get in touch.