AI Tools I Trust — and How I Actually Use Them

Tatiana Shemiakina
Jun 25
6 min read

Updated: Jul 1

Curious how this applies to your team? Grab a 15-minute call

I'm often asked which AI tools I actually use. Not the tools I tested once. Not the tools that show up in big "top 50 AI tools" lists. The ones I return to again and again when I need to think, write, reAI Tools I Trust — and How I Actually Use Themsearch, design, or prepare.

So here is my honest list.

1. ChatGPT and Claude: my thinking partners

I use ChatGPT and Claude as conversational partners — not just for "write this for me," but for thinking with me.

The quality depends enormously on how you prompt. I rarely ask a simple question and accept the first answer. I usually ask the model to take a role: be my critical thinking partner, an OD consultant, a skeptical reviewer, a research assistant, a strategy advisor, a professor who won't let me get away with vague logic.

I often ask:

"Don't agree with me too quickly."
"Where is my reasoning weak?"
"What am I missing?"
"What would a skeptical reviewer say?"
"What evidence would support or challenge this?"

This is where these tools become genuinely useful. They help me slow down, question my assumptions, and make decisions based on clearer logic.

I also use them against each other. When one gets stuck, starts hallucinating, or gives me something too generic, I move the same task to the other. Sometimes Claude is better for long, thoughtful writing; sometimes ChatGPT is better for practical structuring, images, brainstorming, and step-by-step work. I don't treat either as the truth. I treat them as thinking partners with different strengths.

Claude feels more careful and elegant to me, especially for nuanced writing and long documents — though it's also expensive and usage-limited, even when you pay. ChatGPT feels more flexible for everyday work, and it's easier to use with people who aren't advanced AI users yet.

And ChatGPT has become surprisingly useful for images. I know there are more specialized image tools — Midjourney, Adobe Firefly, Ideogram, Canva — but for my work, ChatGPT is often enough. I use it for thumbnails, event cards, banners, presentation visuals, and concept images when I need something quickly and can revise it conversationally.

Example: when I prepare conference visuals, I don't just ask "make a flyer." I say what the event is, who the audience is, what emotional tone I want, what text should be prominent, what should be subtle, and what to avoid. Then I keep refining.

2. Perplexity: when I need research trails

For research, I often prefer Perplexity. Not because it's perfect — it isn't, and I still check the sources. But I like it when I need a fast research trail: what exists, what's being cited, what terms are used, what recent conversations are happening, and where I should look next.

For me it beats asking a general chatbot to "research this" when I need to see the sources immediately. It's especially useful at the start of a topic, when I'm trying to map the field. I use it for questions like:

What are the current conversations about AI and early-career talent?
What are companies saying about AI adoption and employee trust?
What research exists on AI, work design, and skill development?

I don't use Perplexity as the final authority. I use it as a research scout.

Deep Research or Deep Dive tools can be stronger when I need a more comprehensive, synthesized report — but they can also be slower, heavier, and sometimes too polished. Perplexity is lighter. It helps me find the trail quickly, then I decide what's actually credible.

3. NotebookLM: when I already have the sources

NotebookLM is one of my favorites when I already have materials. If Perplexity helps me find the world outside, NotebookLM helps me work with the world I've already collected.

I use it when I have articles, reports, transcripts, class materials, or proposals and I want to understand what's inside them without losing the connection to the source. It's very helpful for summaries, briefing notes, short explanations, mind maps, study guides, questions to ask — and now even short video-style explanations.

I especially like it for teaching and learning. Upload several articles on AI and organizational behavior, and I can ask it to identify themes, contradictions, definitions, and practical implications — or to create a mind map so I can see the structure of the material visually.

This isn't the tool I use for beautiful writing. It's the tool I use when I need to understand source material faster.

4. Gamma: when I need reports and presentations to look decent quickly

For reports and presentations, Gamma is still one of my favorites. I like it because it understands structure: it can take a lot of content and organize it into cards or slides that make sense. It doesn't always get everything right, but it gives me a far better starting point than a blank slide.

I use it to turn a memo, workshop outline, report, or set of notes into something visual — a consulting report, a workshop deck, a proposal, a summary of findings, a short educational presentation.

What I like most is how easy it is to rearrange. If a section is too long, I split it. If the order feels wrong, I move cards. If the design isn't right, I adjust it without fighting the tool for two hours. Gamma isn't my final designer; it's what gets me from "I have a document" to "I have a presentable structure."

5. ChatGPT and Claude for designing tools and chatbots

I use both when I design tools, prompts, or chatbots — one of the most valuable uses for me.

If I want to build a chatbot for a workplace wellbeing program, I don't start by writing random prompts. I use AI to help me define the purpose, the user, what it should never do, the tone, the framework behind it, the questions it should ask, and how it should respond thoughtfully when someone is hesitant, frustrated, or overwhelmed.

ChatGPT is especially useful when I want something scalable and easy to share with people who aren't expert users — simple structures, instructions, scripts, and workflows others can actually use. Claude is strong when I need nuance: tone, boundaries, careful language, and thoughtful dialogue. I like it for refining the personality of a tool so it doesn't sound robotic, too cheerful, or too directive.

My rule: ChatGPT helps me build and operationalize; Claude helps me refine and humanize. Both can do both — but that's how I tend to use them.

6. Granola: for meeting notes that don't make the meeting weird

For meeting notes, I prefer Granola. I've used Otter, and it can be very useful — but I don't always like how meeting bots join the call and become visible in the room. Sometimes it changes the feeling of the conversation. Sometimes it creates awkwardness. Sometimes it just feels too present.

Granola feels different. It doesn't try to become another participant; it just helps me keep notes. Quieter, more modest, less intrusive. I use it when I want to stay present in the conversation but still leave with a useful summary, action items, and a record of what was discussed.

This matters because so much of my work is relational. I don't want the note-taking tool to become the center of attention — I want it to support the conversation, not change it. (I still think people should be car

eful with privacy and consent when using any meeting tool. But in terms of experience, Granola fits how I like to work.)

My overall rule

I don't trust AI tools because they're "smart." I trust them when I understand what role they play in my workflow.

ChatGPT and Claude are thinking partners.
Perplexity is a research scout.
NotebookLM is a source-grounded study partner.
Gamma is a structure and presentation helper.
Granola is a quiet meeting-memory tool.

And none of them replaces judgment. That's probably the most important point. The best use of AI, at least for me, isn't to outsource thinking — it's to create better conditions for thinking: faster drafts, clearer structures, better questions, more perspectives, and more time for human judgment.

So when I choose an AI tool, I ask three things: