Open any AI writing product and the first thing you see is a template gallery. Jasper has about 100. Copy.ai has over 90. Notion's template directory has thousands. The unstated assumption is that more templates is better, because more templates means you can market to more use cases.
Projelli ships with 15. That's the whole list. When people ask whether I'll add more, the honest answer is "only if three founders ask for the same one." Here's the reasoning, because it's the core of how the product is different.
When I was doing competitive research, I signed up for Jasper and Copy.ai and Writesonic and clicked through every template they had. Most fell into one of four buckets.
SEO bait: "Write a blog intro," "Write a meta description," "Write a YouTube title." One-shot prompts a founder could write in 30 seconds by just typing the request into ChatGPT. The template adds nothing except a form field.
Corporate marketing for big teams: "Write a press release," "Write a product launch announcement." Useful if you're a marketing coordinator at a 500-person company. Irrelevant to a solo founder.
Social media: "Write a LinkedIn post," "Write 5 tweets." What the tools sell on, but the output is usually generic enough that it hurts your voice rather than helping it.
Then there's the good stuff. Maybe 10 or 15 templates, buried in the pile, that map to real founder work. That's the bucket I wanted Projelli's whole gallery to be.
A template belongs in Projelli if, and only if, both of these are true:
1. It's something a real founder I've met actually does on a real week.
2. Existing tools (ChatGPT, Notion, Google Docs, Jasper) don't do it well, because they're missing either the structure, the persistence, or the domain knowledge.
If both are true, it's in. If only one is true, it's out. No bucket for "things that would be cool," no bucket for "things a VC might ask about," no bucket for "things that look good in the template gallery screenshot."
This cut a lot. It cut anything a one-shot ChatGPT prompt handles fine. It cut anything where the founder's thinking is mostly numbers the AI can't see. It cut anything where another tool (TermsFeed for privacy policies, Ahrefs for keyword research) does it better. What was left was 15 documents that sit in the overlap between "founder writes it" and "existing tools are bad at it."
Let me walk you through six of them.
What the founder actually does: You're five weeks in. You've been telling yourself "I'll figure out positioning later." Now you're writing landing page copy and you realize you can't finish the sentence "unlike [competitor], we...". You open five browser tabs, get interrupted after the third, and when you come back you've forgotten what you noticed about the first one.
Why ChatGPT fails: ChatGPT will summarize one competitor fine. But there's no structure holding five comparisons side by side, no persistence between sessions, and no way to say "update the Tana row now that they shipped their new feature" without starting over. The comparison is the value, and ChatGPT gives you five separate summaries instead.
How Projelli handles it: The template walks you through a structured matrix. You name each competitor, answer the same 8 questions about each, and the output is a side-by-side table plus a written analysis of where you can win, where you can't, and what your positioning needs to sound like. Six weeks later when Tana ships something new, you open the file, update the one row, and ask the AI to re-do just the positioning section. You don't lose the other four analyses.
What the founder actually does: You're trying to figure out what to charge. You want to pressure-test a few scenarios (what if $29, what if $79, what if you bundle, what if you split into tiers) before you commit. The output you want isn't a number. It's a defensible story: why I'm charging this, what I tested, what would have to change for me to reconsider.
Why ChatGPT fails: Pricing conversations in ChatGPT go great for 45 minutes and then disappear. I've had this experience three or four times. You end the session with a pricing model you believe in, close the tab, and two weeks later you've lost the reasoning. You remember you decided $49 one-time, but not why you ruled out $79.
How Projelli handles it: The template produces two files: a pricing model (tiers, logic, math) and a decision log (what you considered, what you ruled out, under what conditions you'd revisit). Both stay in your workspace. When you open the pricing file six months later, the decision log is right there as a linked file.
What the founder actually does: You got off a 30-minute call with a potential customer. You took 12 bullets during the call. It's 4pm on a Thursday, and if you don't process the notes within 24 hours they'll rot. The specific phrases blur. You'll remember what you think they said rather than what they actually said. You need a structured debrief: the problem in their words, which of your hypotheses they confirmed or challenged, what new questions surfaced, and whether they're a fit for the beta.
Why ChatGPT fails: Paste the notes into ChatGPT, ask for a summary, and you'll get a summary. The problem is that a debrief isn't a summary, it's a lens. You need the output structured around YOUR hypothesis map, not a generic "key points" list. ChatGPT doesn't know what you're testing, so it can't tell you which notes are the interesting ones. Notion AI has the same problem, and the output gets stuck inside one page.
How Projelli handles it: The template loads your hypothesis doc and uses it as context. You paste raw notes, the AI runs them against the hypotheses, and tells you what's new, what's confirmed, what's contradicted. The output is filed under interviews/ with a filename like 2026-04-14-sarah-design-lead.md. Ten interviews later, you can ask the AI to synthesize patterns across all of them, because they're all real files in the same folder.
What the founder actually does: If you've taken money, you owe a monthly or quarterly update. Structure doesn't change from month to month: metrics, wins, losses, asks, a short narrative. What takes the time isn't writing. It's remembering what you did. By Friday afternoon you can't recall what you shipped on Tuesday, so you scroll through your calendar, your commits, your Notion, your Slack, reconstructing your own month. Writing is 20 minutes. Reconstruction is two hours.
Why ChatGPT fails: ChatGPT has no memory of your weeks. You'd have to feed it everything from scratch every time, which is the thing you're trying to avoid. Notion AI is slightly better if you live in Notion, but the update pulls from five tools anyway.
How Projelli handles it: The template reads from your weekly review files. If you've been running the weekly review, the investor update pre-fills with the last four weeks of wins, losses, and metrics you already captured. You're no longer reconstructing. You're editing. That only works if the weekly reviews are real persistent files the template can read from, which is what a chat window can't do.
What the founder actually does: You're ready to hire your first person. You need a job description, an interview plan, a rubric so you score candidates the same way, a scorecard, and a 30/60/90 onboarding plan. All five should be internally consistent. The rubric should match the scorecard. The onboarding plan should align with what you said you needed in the JD.
Why ChatGPT fails: Ask ChatGPT for a job description and you'll get a decent one. Start a new conversation and ask for an interview rubric, and it won't reference the JD. You patch them together, realize they contradict each other, and end up rewriting the whole thing. The tool doesn't know these five documents are one system.
How Projelli handles it: The first hire playbook runs as one workflow. You answer the questions once, and the workflow produces all five documents with wiki-links between them. Edit the JD to add a requirement, and the rubric file picks it up as a backlink. You're running one workflow that treats the hire as a single object.
What the founder actually does: You have an investor conversation next month. You know what slides go in a pitch deck. The problem isn't the layout, it's the story. The opening line, the problem frame, the customer quote that makes it real. A pitch deck is a narrative, not a design. The design is a day of work once you have the story. The story is where founders burn a week.
Why ChatGPT fails: ChatGPT will generate a 10-slide outline that reads like 1,000 other generic deck outlines. It doesn't know your customer language. It doesn't know the numbers from your beta. It can't pull from your competitor analysis or your interviews, because they don't exist in its context. You'd have to feed it everything every time, and you won't.
How Projelli handles it: The pitch deck template reads from your customer persona, your competitor analysis, and your pricing file if they exist. The customer quote in the problem slide is something you actually captured in a real interview, not a plausible-sounding invention. The output links back to sources for every major claim. Update your persona file, re-run the workflow, and the deck narrative updates too.
Read those walkthroughs again and the failure mode for ChatGPT is the same every time. It's not that ChatGPT is bad at writing. It's excellent at writing. It's that ChatGPT treats every request as independent, with no memory of your other requests, no structure spanning documents, and no way for today's output to be tomorrow's input.
The value of these templates isn't the prompts. You could reverse-engineer the prompts from any one of them. The value is that the outputs land in a workspace where they persist, cross-reference each other, and become context for the next workflow. A pricing decision that references your competitor analysis. A pitch deck that pulls from your customer persona. An investor update that reads your weekly reviews.
That only works if the documents are real files sitting in a real folder on your hard drive. Which is why Projelli is local-first, and why the template count is 15, and why those 15 aren't going to become 150 just to look better in a feature comparison chart.
For the record, the other nine are New Business Kickoff (the flagship, turns a fuzzy idea into Vision + PRD + Lean Canvas), Customer Persona, MVP Scope, Go-to-Market Plan, Landing Page Copy, Content Strategy, Financial Model, Weekly Review, and Board Meeting Prep. Each passed both rules. I wrote about the selection process in more detail over here.
If you have a workflow you wish was in the list, tell me. I keep a spreadsheet of user requests, and a template graduates into the product the third time somebody asks for it. That's the only way the list grows.
See the 15 templates in ProjelliJameson Daines builds Projelli on weekends and evenings around a Senior Product Designer day job. If you're a founder with a workflow that isn't on the list, email [email protected]. Read about why local-first matters for founders or get Projelli at projelli.com.