I Used 14 AI Tools Daily for 90 Days. Only 4 Stayed.
The experiment
For 90 days I committed to using 14 AI tools every single day, across every category I work in — writing, research, coding, design, video, audio, email, content, knowledge management, productivity. The premise was that I wanted to know which tools actually earn ongoing use versus which ones survive only because of switching costs. The result was uncomfortable in a way I want to share, because most “best AI tools” content is written by people who have never actually committed to using all the tools they recommend.
Out of 14, four stayed in my workflow after 90 days. Ten got cut. Here’s what happened.
Day 0 baseline
The 14 tools I committed to, picked because each was the leader (or co-leader) in its category as of early 2026:
ChatGPT, Claude, Cursor, Notion AI, Perplexity, Suno, Midjourney, Runway, Descript, Otter, Beehiiv, Loom AI, Lavender, Jasper.
Total monthly cost: $312/month all in. Target: use each daily, observe friction, write notes weekly.
The full 14 tools
I’m not going to itemize all 14 here — it’d take 3,000 words and you’d skim. Instead I’m going to skip to the part that matters: which ones I dropped and why, and which four survived.
Week-by-week churn
Week 2: Jasper dropped. I was already paying for Claude and ChatGPT, and Jasper’s pitch (team brand controls, content templates) was solving a problem I didn’t have as a solo. Same content output came from Claude with a tighter prompt. -$59/month.
Week 3: Lavender dropped. Real-time email coaching is great if you’re writing 60 cold emails a day. I was writing 10. The tool was solving a different scale of problem than mine. -$29/month.
Week 4: Loom AI dropped. I record async videos maybe once a month. The AI features were fine; the underlying behavior wasn’t supporting them. The tool wasn’t wrong; my use case was. -$15/month.
Week 5: Otter dropped. I’d been using it for call transcripts but Fireflies’ tighter CRM integration won me over for the calls that mattered. -$20/month.
Week 6: Beehiiv kept. I run a small newsletter and the platform itself is the workflow. Stayed.
Week 8: Runway dropped. I generated maybe four videos in 90 days. The use case wasn’t daily for me. If I’d been making a film, Runway would’ve stayed. -$15/month for the moment.
Week 9: Midjourney dropped from daily to monthly. Beautiful tool. Hero-image work isn’t daily for me. Kept the cheapest tier as an “as needed” account. -$30/month most months.
Week 10: Notion AI kept. Daily Q&A across the workspace + smart writing features actually integrated into how I think.
Week 11: Suno dropped. Made about a dozen songs for fun. Wasn’t a workflow tool — it was a toy. Toys are great but they don’t earn $24/month. -$24/month.
Week 12: Descript dropped from primary to backup. I edit fewer videos than I expected; the few I do edit get done in DaVinci faster. Kept the free tier just for transcription. -$30/month.
Week 13: Perplexity moved to “daily” status. Was using it occasionally; promoted to default search after Comet browser made it constant. Stayed.
The remaining tools at week 13 were: Claude, ChatGPT, Cursor, Notion AI, Perplexity, Beehiiv.
Week 13.5: dropped ChatGPT in favor of Claude. Both were good; running both was redundant. Claude won my workflow for writing nuance; saved $20/month.
The 4 that stayed
By end of 90 days:
- **Claude.** The foundation. Every day, multiple times. Writing, thinking, debugging conversations, drafting copy, decision support. $20/month.
- **Cursor.** Every day I write code. The compounding return on the tab autocomplete + Composer agent makes this irreplaceable. $20/month.
- **Notion AI.** The Q&A on my workspace + the writing assist inside docs makes Notion sticky in a way the base product wasn’t. $10/user/month add-on.
- **Perplexity** (with Comet browser). Replaced default search. The agent features in Comet promoted Perplexity from “occasionally” to “daily.” $20/month for Perplexity Pro.
Beehiiv stayed too but as platform infrastructure rather than an AI tool I use; it’s the rails the newsletter runs on.
Total spend at day 90: $70/month for the four. Down from $312. Savings: $242/month, or $2,900/year.
The data
| Metric | Day 0 | Day 90 | |—|—|—| | Tools in daily use | 14 | 4 | | Monthly spend | $312 | $70 | | Tools opened daily without thinking | 4 | 4 | | Tools requiring forced use | 10 | 0 | | Subjective productivity (1-10) | 7 | 8 |
The productivity went *up* when I cut the 10. The mental overhead of which tool to open for which job was the hidden tax. Picking four and going deep beat trying to maintain expertise across 14.
Would I do it again?
Yes, with a tighter starting list. The 14-tool baseline was too wide. If I were redoing this experiment, I’d start with 6 tools — the leader in each of the categories that matter most for my work — and commit to those for 90 days, with a clear “stay or cut” decision at day 30 and day 60.
The bigger lesson: tool sprawl is the real bottleneck, not tool selection. If you’re trying 11 tools and feeling AI-overwhelmed, you don’t need a better tool. You need to cut to four, go deep, and reclaim the cognitive overhead.
Mine were Claude, Cursor, Notion AI, Perplexity. Yours will be different. Figure out which four and stop reading lists like this one.