AI Guides › Getting Started
The Three Questions To Ask Before Any New AI Tool
By Nigel Guy · 2 min read
A new AI tool gets recommended to you and the instinct is to sign up and
poke around — see what it does, decide later whether it's worth keeping.
That's how you end up with a dozen half-used accounts, three of them holding
data you'd forgotten you'd given them, and no clear memory of why you picked
any of them.
The rule: ask what it costs, what it does with your data, and what it's
actually for, before you create the account — not after.
The three questions
- What does this actually cost, once you use it properly? The free tier
is rarely the story. Check what the next tier up costs, and check what
triggers the upgrade prompt — a message limit, a feature gate, a file-size
cap. If you can't find pricing in under two minutes, that's itself useful
information about how the product wants you to sign up first and find out
later.
- What happens to what you put into it? Specifically: is your input used
to train their models by default, and can you turn that off? Is there a
retention period, or does it keep things indefinitely? A vague privacy
page that talks about "improving our services" without specifics is
answering a different question than the one you asked.
- What is this actually for, and is that your actual task? Not "AI tool
for productivity" — the specific job. A tool built for summarising
meetings and a tool built for writing code are both "AI tools," and using
the wrong one for your task usually means you're fighting the interface
more than the model.
What to skip
Skip reading a full terms-of-service document end to end — the data-handling
question above is answerable from the privacy policy summary or a search
within the page, and most of the rest is generic legal boilerplate. Skip
comparing every plan tier in detail before you've decided you'll use the
tool at all; you only need the tier you'd actually land on. And skip taking
a sales page's own claims about accuracy or capability at face value — that's
a marketing document, not a spec sheet.
Guardrails
- Pricing and features change; a screenshot from a blog post six months old
may already be wrong. Check the current pricing page directly rather than
trusting a secondhand summary, including this one.
- "We don't train on your data" is a claim worth finding stated plainly, in
the tool's own policy, not inferred from silence.
- None of this substitutes for your organisation's own data-handling rules,
if you have them — check those first if you're using the tool for work,
not personal, tasks.
All 751 AI guides · JulieMango plans from £17/mo