Choosing / Reference
The AI You Already Pay For
Before subscribing to anything new, check what your existing tools now include. A significant amount of it arrived without announcement and is unused.
Most productivity software has added AI features over the past two years. Much of it is included in plans people already hold, and a lot of it is unused because it arrived quietly.
Auditing what you already have is the cheapest possible step and it frequently makes a new subscription unnecessary.
Where to look
Your office suite. Document drafting and editing, spreadsheet formula generation and data explanation, presentation generation from an outline, email drafting and summarising. Availability varies by plan and region.
Your code editor. Completion, chat, and increasingly agentic editing. Several offer meaningful free tiers.
Your browser. Summarising, translation, writing assistance, and increasingly a built-in assistant.
Your operating system. Both desktop platforms and both mobile platforms now include writing tools, summarisation, image editing and transcription, much of it running on the device.
Your note-taking app. Summarising, question answering over your notes, formatting.
Your design tool. Generative fill, background removal, layout assistance, image extension.
Your video and audio tools. Transcription, subtitle generation, filler-word removal, noise reduction, voice isolation.
Your customer or project tools. Summarising threads, drafting replies, categorising.
Your phone's camera and gallery. Object removal, subject extraction, text recognition, search over image content.
Why this matters before you subscribe
A great many AI subscriptions duplicate a feature already included somewhere. Transcription, summarising, background removal, image upscaling and writing assistance are all now standard in mainstream software.
Do the audit first. An hour spent looking through the tools you already pay for saves more than any comparison of new ones.
The quality question
Bundled features are frequently weaker than dedicated tools, and frequently sufficient.
Where bundled is usually enough: transcription, summarising a document you can check, formula generation, background removal, simple writing assistance.
Where a dedicated tool still wins: long-form drafting, complex analysis, anything needing the strongest model, specialised domains.
The deciding factor is usually whether you can verify the output. For a task where you would check the result anyway, the bundled version is fine.
The privacy dimension
Bundled features vary enormously in where processing happens.
On-device processing is the strongest position. Both mobile platforms now do a meaningful amount locally, and it is usually indicated.
Cloud processing under your existing agreement is generally acceptable for work data, because the same provider already holds the documents.
Cloud processing under a new agreement, where a feature routes to a third party, is worth checking. Some suites do this and disclose it in settings rather than prominently.
The relevant question: does using this feature send my document somewhere it was not already, and under what terms?
Turning them off
Some bundled AI is intrusive — suggestions that interrupt, summaries you did not ask for, features that upload content by default.
All of it can be disabled, and the settings are usually under a privacy or an AI heading. For regulated or confidential work, an explicit decision about which features are enabled is worth making deliberately rather than inheriting the default.
The practical sequence
Audit what you have. List your paid software and check each one's AI features.
Try them on real work for a week.
Identify what is genuinely missing rather than what a product marketed to you.
Then, and only then, consider a subscription for the specific gap.
Most people who do this in order find the gap is smaller than expected, and the subscription they were about to buy overlaps something they already own.
Running the audit
A structured hour that most people have never spent.
List every piece of software you pay for, personally and through work.
For each, search its settings and release notes for AI features. Many arrived in updates that nobody read.
Note what is on by default. Some features are enabled without announcement and process content immediately.
Note where processing happens — on the device or in the cloud, and under which agreement.
Try each one on real work, once.
Then write down what is genuinely missing. Not what a product advertised to you. What you tried to do and could not.
The gap between "features I have" and "features I use" is usually large, and the gap between "features I need" and "features I was about to buy" is usually larger. That second gap is what the hour buys you.