Hi — I'm Lydia, one of the co-founders of ivee. People keep asking me which AI tools I actually use day-to-day, so I wrote it all down. The full guide goes deep on the stack, the prompts and the workflows. This post is the highlights reel. A quick note before we start: stacks are personal. Mine fits how I work — running a small, scrappy startup with a tiny team and a lot to ship. Yours will look different. Steal what's useful, ignore what isn't. Here's the shape of it: the brain (the LLMs), the wiring (the automation layer), and the everything-else (notes, docs, calls). Plus one last bit — how to pick your own.
The brain: picking your LLMs
The thinking layer is the most important call you’ll make. Get this wrong and the rest of the stack ends up doing the wrong work, faster.
I use three:
Claude — for writing, editing, structured thinking, anything that needs a long memory and a steady hand. It’s my default.
ChatGPT — for image generation, voice mode and the odd second opinion when Claude and I are stuck in a loop.
Perplexity — when I want sourced answers from the live web rather than vibes.
The trick isn’t picking one and being loyal. It’s knowing which one to reach for. We compared Claude and Perplexity head to head a while back, and the takeaway holds: ask Claude when you want to think; ask Perplexity when you want to cite.
If you’re starting out and want one to learn on, pick the one with the better mobile app for your phone. The friction matters more than the leaderboard.
The wiring: connecting the tools up
The wiring layer is what turns “a clever chat” into “a thing that runs without me”. It’s the boring middle bit, and it’s where most of the time-saving lives.
n8n — my main automation home. Open-source, self-hostable, plays nicely with Claude and Gmail. I use it for inbox triage, lead routing and weekly digests.
Zapier — when I need a fifteen-minute fix for a one-off problem and I don’t want to think.
Make — for the visual people on my team who want to see the flow.
Claude Cowork mode — for things that need to read files on my desktop or talk to my actual apps. It’s the one that finally made local agents feel sensible.
I picked n8n as the home base because it scales with the business — but if you’re just starting, Zapier is faster to learn. We’ve covered the top use cases for Claude Cowork if you want a sense of where Cowork fits in.
Anthropic has a useful primer on building effective agents that I send to anyone curious about going further.
Notes, docs and the small stuff
Once the brain and the wiring are in place, you need somewhere to put everything. This is the bit that looks unglamorous on paper but quietly makes or breaks your week.
Notion — wiki, doc store, project tracker. Where the company brain lives.
Google Docs — for anything I’ll share outside ivee or want commenting on.
Granola — meeting transcripts, summaries, action lists. Hands-down the smallest tool with the biggest weekly time-save.
Slack — internal comms, plus the channel where my agents drop their daily updates.
Loom — for things that are faster to record than to type.
If your team is bigger than a handful, the McKinsey piece on why agents are the next frontier of generative AI is worth the read — it’s a useful argument-starter when you’re trying to convince a sceptical exec why this stack matters.
How to pick your own stack — three rules
Don’t copy mine. Build yours from these:
One brain, one wire, one notes app. That’s the floor. Anything more is a luxury until you’re using all three daily.
Pick what you’ll actually use. The best tool is the one that opens when you tap your phone. Friction beats features.
Add slowly. A new tool every fortnight at most. Anything faster and you’ll spend more time configuring than working.
If you’re rolling this out across a team rather than yourself, the post on getting your employees to actually use AI is the one I send people first. Adoption is a different problem from selection, and most teams skip it.
Once you’ve got the basics in place, write yours down. Even a one-page version helps. You’ll catch the gaps the moment you try to put it on paper.
FAQs: my AI tech stack
Do I need every tool Lydia uses?
No. Pick one in each layer — one LLM, one wiring tool, one notes app — and use those daily before adding anything else.
Why pick Claude over ChatGPT?
For me, Claude wins on writing, editing and long-context thinking. ChatGPT is the better fit for image generation, voice mode and a quick second opinion. Most weeks I use both.
How much does this stack cost per month?
A starter version (one paid LLM plus free tiers of n8n/Zapier and a notes app) runs around £20–30 a month. The fuller version with paid tiers across the board lands around £80–120.
Is this the same stack ivee teaches?
Most of it, yes. Our programmes lean on Claude, ChatGPT and the no-code wiring layer. The full stack guide goes deeper on prompts and exact workflows.
Where can I see the full version?
The complete tech stack guide — every tool, prompt and workflow — is available via the link at the bottom of this post.
Conclusion and next steps
Tools are downstream of habits. Pick a stack you’ll actually open, write it down, and review it every few months as new things arrive.
If you want the full version — every tool, every prompt, every workflow — the complete guide is yours to grab below. It’s the one I share with new hires on day one.
