intermediate 8 min read Feb 24, 2026

Thinking About Tools

A framework for tool selection

#thinking-about-tools #topic

How to Think About AI Tools

AI tools are multiplying. There are platforms for writing, platforms for coding, platforms for images, platforms for research, platforms for specific workflows, platforms built inside tools you already use. It feels overwhelming and it’s getting worse.

You don’t need more tools. You need a framework for deciding which tools actually matter for your situation, and when to ignore everything else.

New to AI? Start with the On-Ramp to find your entry point, then come back.

The Two-Category Framework {#two-category-framework}

Every AI tool fits into one of two buckets:

Generalists: AI chat platforms that can handle a wide range of tasks - writing, analysis, coding, research, brainstorming, explanation. ChatGPT, Claude, and Gemini are the big three. You talk to them in plain English. They do a lot of things reasonably well.

Specialists: Tools built for a specific job or workflow. An AI that only does slide decks. An AI that only handles legal contracts. An AI that only generates images. An AI built into your design software. An AI that connects your calendar to your email to your to-do list.

The mistake most people make is obsessing over specialists before they’ve mastered a generalist. Specialists can be powerful, but they’re optimization tools. Generalists are foundation tools. Get the foundation right first.

Start With One Generalist {#start-with-one-generalist}

Here’s the honest truth: most people should pick one major platform and actually learn it. Not try all three, not bounce between them based on whatever article they just read, not subscribe to five things and use none of them well.

Why one platform?

AI is a skill, not just a tool. The people who get enormous value from AI aren’t using magic features nobody else has access to. They’ve gotten good at:

Those skills transfer between platforms, but you learn them by going deep on one, not by going shallow on three. You’re better off becoming genuinely proficient with Claude or ChatGPT or Gemini than barely scratching the surface of all three.

How to choose:

If you’re already deep in Google’s ecosystem (Gmail, Docs, Drive), Gemini is the practical starting point. The integration is seamless and you’re probably already paying for it.

If you care about writing and want thoughtful, detailed responses, Claude tends to be the better choice. It excels at sustained work on complex documents.

If you want something widely used with strong all-around capabilities and lots of third-party integrations, ChatGPT is the safe bet.

Don’t overthink it. They’re closer in capability than the marketing suggests. The platform that fits your existing workflow is usually the right one.

When to Add a Specialist {#when-to-add-specialist}

Add a specialist tool when you hit a clear, specific limitation with your generalist that’s costing you time or quality. Not because you saw an ad, not because a newsletter recommended it, not because it looks cool.

Real examples of when a specialist makes sense:

  • You’re doing presentations weekly and your generalist’s slide output is consistently mediocre and takes forever to fix. A deck-focused AI like Beautiful.ai or Gamma becomes worth it.
  • You’re a designer and AI image generation is core to your work. See the Image Generation guide for tool comparisons.
  • You’re running a business and need to extract data from documents constantly. A tool built specifically for that workflow will pay for itself.
  • You’re learning to code and hitting walls with your generalist’s coding help. A specialized coding assistant like Cursor or GitHub Copilot becomes valuable.

Notice the pattern: Each of these is someone doing something frequently, hitting a real limitation, and able to point to exactly where the generalist is falling short. That’s the signal to look at specialists.

The Two-Subscription Rule {#two-subscription-rule}

Most people should cap AI subscriptions at two. Not ten, not five, not three. Two.

Your two slots:

  • Slot one: Your main generalist platform
  • Slot two: Either a second generalist (if you genuinely use both for different workflows) or a specialist tool that you use heavily enough to justify the cost

Everything else should be:

  • Something you already have access to through tools you’re paying for — see what you might already have
  • A free tier you use occasionally
  • A tool you pay for only when you need it (project-based pricing, one-off purchases)

If you’re paying for more than two AI subscriptions and you’re not a heavy user or a professional using these for work, you’re almost certainly overpaying for capabilities you’re not using enough to justify. See Cost Management & ROI for help calculating value.

Decision Tree: What Do I Actually Need? {#decision-tree}

Work through this honestly. Don’t think about what sounds cool. Think about what you actually do.

Question 1: Are you already using AI regularly?

No: Start with a free tier of one generalist. Spend two weeks using it for real tasks before you even think about paying. If you haven’t found a use case in two weeks, you don’t need to pay for anything yet.

Yes: Move to question 2.

Question 2: Do you have a specific problem that your current tool isn’t solving well?

No: Stick with what you have. Focus on getting better at prompting and workflow design. The limitation is probably skill, not tools.

Yes: Move to question 3.

Question 3: Is this problem something you do frequently enough that a dedicated tool would pay for itself?

Rarely or occasionally: Don’t add a tool. Learn to work around the limitation or accept that it’s not a perfect solution.

Weekly or more: Consider a specialist tool. Calculate whether the time saved is worth the subscription cost. If it saves you an hour a week and your time is worth more than the subscription, it’s probably worth it.

Daily or constantly: Absolutely look at specialists. At this level of use, optimization tools pay real dividends.

How to Evaluate Whether a Tool Is Worth Paying For

Most AI tools are not worth paying for. Here’s how to tell the difference.

The free tier test: If you haven’t outgrown the free tier, you don’t need to pay. Period. Free tiers today are genuinely capable. Rate limits exist but they’re generous for casual use.

The workflow test: Does this tool fit into an existing workflow or create a new one? Good AI tools fit where you’re already working - your email, your documents, your design software, your code editor. Bad AI tools want you to come to them and change how you work.

The specific limitation test: Can you name exactly what this tool does better than what you have now? “Better AI” is not an answer. “Handles long documents without losing the thread,” “generates slides that don’t need complete redesign,” “connects to my actual calendar and email” are real answers. If you can’t be specific, you don’t need it.

The frequency test: How often will you actually use this? Be realistic. Once a month is not worth a subscription. Once a week might be, depending on what it saves you. Daily use is easy to justify.

The cancellation test: Is it easy to cancel? If a tool makes cancellation hard, that’s a red flag. Good tools don’t need to trap you.

Building Your AI Stack From Scratch {#building-your-stack}

Here’s a practical starting workflow, regardless of who you are or what you do.

Step 1: Pick one generalist platform

Use the criteria above. Don’t spend more than 15 minutes deciding. Just pick one and go.

Step 2: Spend two weeks with the free tier

Use it for real tasks. Things you actually need to do, not toy problems. Draft emails you’d normally write yourself. Ask questions about things you’re trying to learn. Summarize long documents. Brainstorm ideas for projects.

Step 3: Notice what works and what doesn’t

Keep a running note in your phone or wherever: places where AI genuinely helped, places where it fell down, tasks you wish it could do but can’t seem to handle.

Step 4: Upgrade if you’re hitting rate limits or need better models

If you’re using it constantly and the free tier is getting in the way, upgrade to the paid tier. If you’re barely using it, don’t.

Step 5: Add specialists only when you hit a clear wall

Once you’ve been using a generalist regularly for a month or two, you’ll know what you actually need. Then you can make an informed decision about whether a specialist tool makes sense. Not before.

When to Switch Tools

People switch tools too often. The cost of switching isn’t the money, it’s the lost momentum and the need to rebuild your workflow and habits.

Stay with your current tool unless:

  • You’ve hit a genuine limitation it cannot solve, and another tool can
  • Your workflow has changed and a different platform fits better
  • The tool has gotten meaningfully worse, not just different
  • You’ve given it an honest try and it simply doesn’t click for how you think

“I saw a review saying X is better” is not a reason to switch. “I need feature Y that my current tool doesn’t have and it’s blocking my workflow” is.

What About AI That’s Already in Your Tools? {#ai-in-your-tools}

There’s a good chance you’re already paying for AI and not using it.

Google Workspace subscribers have access to Gemini. Microsoft 365 subscribers have access to Copilot. Notion users can use AI within Notion. Canva users have AI design tools. GitHub users have Copilot.

Check what you’re already paying for before you add another subscription. You might have a capable AI tool already available and simply not turned it on.

Use what you have first. Add only when you’ve outgrown it.

The Bottom Line

Most people need one generalist AI platform, used well, plus whatever AI capabilities are already built into tools they’re paying for. That’s it.

Specialists have their place, but they’re optimization tools, not foundation tools. Get the foundation right first. Add specialists when you hit a real, specific limitation that’s costing you time or quality.

The people getting the most out of AI aren’t using magic tools. They’re using basic tools well. Start there.