Let me tell you about the moment I stopped feeling guilty about my AI subscriptions.
I was staring at a PRD at 11 PM, dreading the rewrite my manager had requested. Three hours of work minimum, probably more. I opened Claude, dropped in my draft, and asked it to restructure the thing.
Twenty minutes later, I had a version that needed maybe 30 minutes of polishing. I sent it to my manager at 11:25.
That $20 subscription paid for itself in a single evening.
I’ve since become a convert. Not a cheerleader — I still roll my eyes at half the AI hype — but a genuine believer that the right paid tools change how you work. Here’s what I actually pay for, what I use it for, and whether you should too.
The Honest Reason I Started Paying
For the first year of the AI boom, I bounced between free tiers. ChatGPT’s free version. Gemini. Whatever browser extension caught my eye. I told myself this was sufficient.
It wasn’t.
The friction accumulated invisibly. Rate limits hitting mid-task. Context windows too small for my actual documents. Starting every conversation from zero because nothing remembered my product, my team, my current priorities. Uploading the same context repeatedly because nothing stuck.
I noticed I was spending more time managing the tools than benefiting from them.
The breaking point came when I calculated how much time I was wasting on AI-related overhead versus how much time I was saving. The math was embarrassing. Free tiers weren’t saving me time — they were just redistributing my inefficiency into new shapes.
I started paying for Claude Pro first. Within a month, I’d added ChatGPT Plus for specific use cases. Then Perplexity when competitive research became a bigger part of my role. Eventually Notion AI when my team moved our wiki. Granola when my calendar became wall-to-wall meetings.
My employer covers some of this. I cover the rest out of pocket. Here’s why each one stays.
The Five Tools and What They’re Actually Worth
Claude Pro ($20/month)
I use this for approximately 70% of my AI-assisted work. The writing quality is genuinely superior for product documents.
My PRD workflow now looks like this: I write the first draft myself because the thinking matters. Then I upload everything — competitor analysis, user interviews, data exports — and ask Claude to stress-test the spec. It catches gaps I missed. It identifies assumptions I didn’t state. It asks questions I should have asked before the design review.
The 200,000 token context window enables this. I drop in a quarter’s worth of user feedback and ask “what are customers most frustrated about?” The answer synthesizes across dozens of sources in seconds.
Projects feature means Claude remembers my product context across sessions. I don’t re-explain that I’m the PM for a B2B SaaS platform with particular technical constraints and stakeholder dynamics. It just knows.
Is it worth $20? In the time it saves me on PRDs alone, yes. But it also makes my specs genuinely better, which has career value beyond time savings.
ChatGPT Plus ($20/month)
I resisted this for months. Why pay for something when I had Claude?
Here’s the answer: different tools do different things better.
When I need to iterate rapidly on messaging or positioning, ChatGPT is faster. The voice mode changed how I work — I take walks and think out loud, and it responds conversationally. I’ve had more creative breakthroughs talking through problems hands-free than I ever did staring at a screen.
Code Interpreter (Advanced Data Analysis) handles data work that Claude doesn’t do as smoothly. I upload CSVs, ask questions in plain language, and get visualizations without touching Python myself.
Image generation for presentations and mockups. Quick answers to questions that don’t need depth. Voice brainstorming sessions.
Is it worth $20? Yes, specifically for data analysis and rapid iteration. Less essential if your work is primarily document-heavy.
Perplexity Pro ($20/month)
This replaced Google for most of my research work.
The citations matter in ways free search doesn’t. When I’m preparing competitive analysis for leadership, I need information I can stand behind. Perplexity gives me sources I can point to, not a list of links to manually verify.
Copilot mode handles complex research queries that require multi-step reasoning. “What features did our top three competitors launch in the last six months, and what user problems were they solving?” returns a synthesized answer with citations.
Is it worth $20? Essential for anyone who presents research to stakeholders. Skip if your research needs are minimal.
Notion AI (~$10/month as part of Business plan)
My team lives in Notion. Our specs, meeting notes, roadmaps, and decisions all live there.
Notion AI makes that documentation actually useful. Meeting notes auto-tag and connect to relevant specs. Action items extract into task databases. I can search across all our documentation by asking questions rather than guessing at keywords.
I write PRDs in Notion anyway. With AI, I can ask “what’s the current status of our mobile initiative across all docs?” and get a synthesized answer pulling from multiple pages.
Is it worth it? Only if your team is all-in on Notion. If your docs live in Google Drive and Confluence, this won’t help you.
Granola ($12/month)
This one surprised me with how much I use it.
I run five to eight meetings per day. The administrative overhead of note-taking, organizing, and distributing was eating an hour minimum every evening.
Granola transcribes meetings and structures the notes intelligently — themes, decisions, action items extracted automatically. I stopped manually taking notes during meetings, which means I actually pay attention. The follow-up reminders actually trigger.
Is it worth $12? If your calendar looks like mine, yes — immediately. If you run two meetings a day, probably not.
The Multi-Tool Reality
My complete stack runs roughly $80/month. That’s less than a single engineering hour, and these tools work 24/7.
The key insight: each tool has a clear lane. I don’t try to make one tool do everything. I route tasks to the right option.
- Claude for writing and reasoning
- ChatGPT for speed and data
- Perplexity for research
- Notion for team documentation
- Granola for meetings
This sounds complex but becomes muscle memory. The mental overhead of maintaining five subscriptions is minimal compared to the mental overhead of fighting a single tool that doesn’t quite fit your workflow.
The Case Against Paying (An Honest One)
I want to be fair: five subscriptions isn’t right for everyone.
Early-career PMs learning fundamentals probably shouldn’t pay for tools that reduce the friction of writing specs, analyzing data, or synthesizing research. That friction is where you develop skills.
If your team has strict tool policies, personal subscriptions create complications around data handling and compliance.
If your actual work involves minimal documentation, research, or meetings, you won’t get value from these tools.
The honest advice: start with one paid subscription. Track the time savings for a month. If the ROI is clear, add the next one based on where you feel the next friction point.
Don’t pay for five tools because a stranger on the internet told you to. Pay for tools that measurably change how you work.
The Real ROI
Here’s what changed when I started paying:
My PRDs got sharper because I had tools to stress-test them. My competitive analysis became more credible because I had better research infrastructure. My meetings became more productive because I wasn’t drowning in administrative overhead.
The tools didn’t replace my thinking. They gave me better material to think with.
That’s worth paying for.
Rating: 8/10 — Honest, practical breakdown of which AI tools actually earn their subscriptions for PMs. The focus on specific use cases and clear value propositions makes this immediately actionable. Deducted points for the inevitable overlap with other “which AI tools should PMs pay for” content already on the internet.
Meta Title (57 characters):
5 AI Tools I Actually Pay For as a PM (Honest Review)
Meta Description (153 characters):
A product manager breaks down which AI subscriptions are actually worth paying for: Claude, ChatGPT, Perplexity, Notion AI, and Granola with real use cases.
