I get this question every other week, on stage and in DMs. "Krista, what's actually in your stack? Like, what do you really use, not the sponsored stuff, the real one." Fair. Here it is. The actual tools running my business as of this month, what each one does, what it costs, and where each one is still rough.
This isn't a list of every shiny tool I've tried. This is the surviving stack after I cut everything that didn't earn its keep.
The principle behind the stack
One rule: every tool has to either save me hours or earn me money. Not "be cool." Not "have AI features." Save time or make money. If it does neither after 60 days, it goes.
The second rule. I'd rather have 6 tools I deeply use than 30 I half-use. Half-used tools are worse than no tool, because they put data in places I forget to look.
The 6 core tools
1. Claude (writing and agents) , ~$20/month for Pro
Claude is the brain. I use it for almost all of my long-form writing, content repurposing, sales page drafts, email replies that need actual thought, and increasingly as the model powering my AI agents (more on those below).
Why Claude and not ChatGPT or Gemini. The writing voice. It's the only model I've used that doesn't make me rewrite every other sentence to sound like me. ChatGPT writes like a startup employee giving a TED talk. Gemini writes like a smart intern who reads a lot of HBR. Claude writes like a person who actually finished a draft.
Cost: Claude Pro subscription is in the $20/month range for individual use last I checked. API usage for agents is separate and pay-as-you-go. We probably spend $80-150/month combined across team usage at this point, though that fluctuates.
Where it's still rough. Long context windows are great, but when I dump 40,000 words of source material, it sometimes prioritizes the wrong parts. Solution: I summarize down to about 8,000 words before asking it to write.
2. n8n (workflow automation) , ~$20-50/month depending on tier
n8n is my workflow brain. Anywhere I'd previously hire a VA to "do this every time that happens," n8n does it now. Lead intake. Follow-up sequences. Content distribution. Cross-posting from one platform to another.
I'm running self-hosted (well, my team is, I'm not touching the server) but n8n Cloud is fine for most people. Their pricing page has the current tiers, I'd just check those rather than trust a number I wrote down 3 months ago.
Where it's still rough. The community nodes for newer platforms can be flaky. We've had 2 workflows break because a third-party node updated and changed the schema. Stick to the official nodes when possible, or build in error handling.
3. Fathom (meeting AI) , free tier or paid plans
Every Zoom I'm on, Fathom is in there taking notes. It records, transcribes, summarizes, identifies action items, and lets me search across every call I've ever had. So when a client says "remember when you mentioned that thing about lead magnets 4 months ago," I find it in 8 seconds instead of guessing.
The free tier covered me for a long time. We upgraded to paid because the team plan lets us share recordings across team members and includes a few features like CRM sync. I'd check their pricing page for what's current.
Why Fathom and not Fireflies or Otter? Honestly, all three work. I picked Fathom because the summaries were the cleanest of the three when I tested. Fireflies is fine. Otter is the OG and has its fans.
Where it's still rough. Like all meeting AI, it occasionally hears something wrong and the summary inherits the error. I scan every summary before I trust it. Takes 30 seconds, worth it.
4. Sales follow-up AI (Lindy or similar) , pricing varies
I'm being deliberately vague here because I'm still actively testing this category. Lindy is one of the platforms I've evaluated for AI agents that handle inbound sales conversations, follow-ups, scheduling. There are several competitors emerging.
What I want this category to do: pick up after my team's first reply, follow up on a cadence, qualify the lead, book the call. Stop the "lead goes cold because nobody followed up on day 7" problem.
Where it's still rough. Honestly? This whole category is the roughest part of my stack. The platforms can sound robotic if you don't tune them carefully. I've also seen one tool send a follow-up to a lead AFTER they'd already booked, which is the kind of thing that erodes trust. So I'm hedging. I'm not telling you "go buy Lindy today." I'm telling you this is a category to watch and to pilot carefully.
If you want my candid take when you're in Level Up, I'll tell you exactly where each platform stands in my testing. The category is moving fast enough that I'd rather give you in-the-moment intel than something I wrote down 6 weeks ago that's already outdated.
5. Notion (SOPs and operations brain) , ~$8-15/month per user on team plan
Notion is where the business lives. Every SOP, every process, every project, every client note, every offer template. If a process exists, it lives in Notion. If it lives nowhere else, it doesn't exist.
The reason I keep Notion specifically (and not ClickUp, Asana, Monday, or any of the others) is that Notion plays unusually well with AI. They've put real effort into MCP integration. So my AI agents can read and write Notion pages, which means my operations brain is also my agents' brain. That matters more every month.
Where it's still rough. Notion's search is okay, not great. Their built-in AI is fine for inline tasks but I still prefer dropping content into Claude for serious thinking.
6. Gmail with MCP triage , Workspace pricing, plus AI usage
This isn't a separate product, it's a setup. Gmail (via Google Workspace) connected through an MCP server so my AI agents can read my inbox, draft replies, label, search, and triage. The result: my morning inbox review went from 45 minutes to about 12. The AI does the first pass. I confirm or edit.
Google Workspace pricing is a few dollars per user per month for the lower tiers, last I checked. The MCP server is free open-source, you just need to be willing to set it up (see my separate article on MCP servers if you haven't).
Where it's still rough. Permissions and OAuth scopes. Granting an AI agent write access to your inbox is the kind of thing that should make you slightly nervous. I started with read-only for 3 weeks, then graduated to draft-only, then graduated to "can send if marked auto-send" with a human review queue for anything else.
The total monthly cost, honestly
Adding it up rough monthly numbers for a small team: - Claude: $20-200 depending on whether you're on Pro or paying for API agents - n8n: $20-50 if cloud, less if self-hosted - Fathom: $0-50 - Sales follow-up AI: $50-200 (testing category) - Notion: $30-90 for a small team - Google Workspace: $30-100 for a small team
So somewhere between $150 and $700/month total for a small operation, depending on team size and how much you push usage. Compared to one mid-level operations hire at $5,000-8,000/month, the stack is a steal.
What I cut
A short list of things I tested in the last 12 months and removed: - 3 different "all-in-one" AI marketing platforms. Every one of them did 5 things mediocrely instead of one thing well. - 2 separate CRMs trying to replace what Notion already does for me. - One "agent builder" that promised the moon and shipped a thin wrapper around OpenAI.
This isn't shade. It's that my business doesn't need everything that exists. It needs the 6 tools above doing their jobs.
The stack will change. The principles won't.
By 2027, half this list might look different. New tools will replace some, mergers will eat others. What won't change: I run my business on tools where the AI can see my data and act on it, not tools where I copy-paste forever between siloed apps.
If you want me to walk you through this stack on real screens, with the exact setup, the actual templates, and what to skip, that's the whole reason Level Up exists. I show you the system, not just the names of tools.