FAQs
Common questions from sessions — answered honestly.
Getting started
Why Claude over ChatGPT?
Both are excellent. For most business use, Claude edges ahead on the things that matter most day-to-day:
Writing quality. Claude produces less obviously-AI output. It's more nuanced, more natural, and better at matching a specific tone — which matters when you're writing anything that represents you or your brand.
Long documents. Claude handles large amounts of text better. Drop in a full contract, a long email chain, a detailed report — it works with all of it without losing the thread.
Cowork. ChatGPT doesn't have an equivalent to Cowork — an agent that works inside your actual files and applications with persistent context about your business. This is the biggest practical differentiator.
Honest responses. Claude tends to tell you when it's uncertain or when your brief has a problem. ChatGPT is more likely to produce confident-sounding output regardless.
ChatGPT has its own strengths — better voice mode, image generation, a larger ecosystem of Custom GPTs. Keep it around. But Claude is the right starting point.
Which tool should I start with?
Claude. Set it up properly with your business context and it'll outperform a generic ChatGPT session on almost everything — writing, analysis, document work.
Then add Granola for meetings and Wispr Flow for dictation. Those three cover most of what you'll want to do.
See Which tool when for the full breakdown.
Do I need to pay?
For serious daily use, yes — the free tiers are too limited.
One important distinction: Claude Cowork requires Claude Max (£90/month) — it's not available on the standard Pro plan (£18/month). If Cowork is central to how you want to work, that's the plan you need.
For everything else — Claude in the browser, Granola, Wispr Flow — you're looking at £15–18/month each.
If you save an hour a week, the maths works. Cowork's higher price reflects what it can actually do.
How long until I see results?
You'll save time on the first day if you use it for something real. The shift from "occasionally useful" to "genuinely indispensable" usually takes 2–3 weeks — once you've got Cowork set up properly and prompting becomes second nature.
Privacy & security
Is it safe to use for work?
On Claude Pro and Max: your conversations are not used to train models, and Anthropic holds SOC 2 Type II certification — the same security standard used by major enterprise software. The data handling is genuinely serious.
What you share is ultimately a personal and organisational decision. Some people include financial reports and client information without concern; others prefer to keep sensitive material out entirely. Both are reasonable positions depending on your obligations and your company's policies.
What helps with Cowork specifically: Claude only sees files inside the folder you point it to. Nothing else on your machine is accessible. So you control exactly what's in scope — it's not connected to your whole system by default.
Passwords and credentials are the one clear exception — never include those anywhere.
See Cowork Security for more detail.
Does Claude train on my conversations?
Not on paid plans. Claude Pro and Team accounts don't use your conversations to train the model. Free accounts may.
Always use a paid account for work.
Can my company see what I'm doing?
On a personal paid account: no, unless you share it. On a company-provided account: assume yes.
Getting good results
Why isn't my output any good?
Almost always one of these:
- Prompt too vague — add more context, specifics, and format instructions
- Not iterating — the first response is a draft. Tell it what to change.
- Missing business context — once Cowork is set up with your context files, results improve significantly without extra effort
See the Prompt Writing Guide.
Should I feel uncomfortable using AI for my content?
No. There's no shame in it — and with the right systems in place, there's no quality compromise either.
If Claude has your tone-of-voice file, your brand guidelines, examples of writing you're proud of, and proper context about your business, the output can be indistinguishable from content written entirely by hand. The quality ceiling is high. It's the setup that determines it, not whether AI was involved.
The practical approach: let AI do the heavy lifting, then add the human touch for the final 10–20%. That last pass is where you catch what doesn't sound right, add the specific detail only you know, and make it yours. That's not a workaround — it's just a good editorial process.
Can people tell I've used AI?
With the right context in place and a human edit at the end — usually not. The tell is when people copy-paste without reading: generic phrasing, wrong tone, missing specifics that only you would know.
The output reflects the quality of what you put in. Strong context files, a good brief, and a final read-through produces content that stands on its own.
Is it accurate?
AI is excellent at structure, drafting, and synthesis. It is not reliable for facts, figures, dates, or anything that needs to be precisely correct.
Always verify anything consequential. Treat outputs like a first draft from a smart intern — capable, but needs checking.
Claude Cowork
What's the difference between Claude.ai and Cowork?
Claude.ai is the browser-based chat interface. Cowork is the desktop app — it works inside your actual files and applications (Excel, Word, Chrome, Google Docs) rather than requiring you to copy and paste in and out of a chat window.
Cowork is what makes Claude genuinely change how you work, rather than just being a useful add-on.
What can Cowork actually connect to?
Via Connectors: Gmail, Outlook, Calendar, Slack, Notion, Google Drive, Salesforce. It can read, draft, and take action inside all of them.
It can also run scheduled tasks (e.g. a weekly competitor check every Monday morning) and record workflows it can repeat automatically.
Is Cowork safe for sensitive files?
It runs locally on your machine. Files aren't uploaded to a cloud server for processing — Claude reads them directly from your file system.
That said, follow your organisation's data policies before connecting it to sensitive systems.
Bigger picture
Will this replace jobs?
The honest answer: it will change what jobs look like. The people who adapt fastest will be able to do more — better output, faster, with less grinding work.
The risk isn't AI replacing you. It's someone who uses AI well doing what currently takes a team.
How do I get my team using it?
Don't mandate it. Find one person who's enthusiastic, let them demonstrate something concrete, and let word spread from there. Mandate tends to produce checkbox adoption; enthusiasm produces real change.
A session like this one is a good starting point — showing it rather than explaining it lands differently.
Where do I go after this session?
- Get the core tools installed and set up
- Use Cowork for something real this week — ideally a task you do regularly
- Come back to these pages when you're at your desk and want to go deeper
Questions after the session: ollie@aytonlabs.io
The answers here reflect Practical AI's current view and experience. AI tools move fast — capabilities, pricing, and best practice all change. Treat this as an informed starting point, not a definitive guide. What works best for you will depend on your context, your business, and your own judgement.