The 5 AI Tools I Actually Use Every Day (And the 20 I Stopped Using)

I have tested more than 20 AI tools in the past 12 months.

I use 5.

Not because the others were bad. Some of them were genuinely impressive.

But there is a difference between impressive and useful. Between a great demo and something that earns a permanent place in how you actually work.

This is my real stack. What I use to run AtheonX, my AI content agency for CEOs, alongside Funnel Duo Media, which I run with my brother Reeve.

And at the end, I will tell you why so many tools did not survive. That part might be more useful than the list itself.

Why Most AI Stacks Do Not Last

Here is what keeps happening.

A new AI tool launches. The internet goes loud. Founders post their stack updates. Threads promise this one changes everything.

You sign up. Spend an afternoon setting it up. Try it for three days.

Then it sits in a browser tab, untouched, until you finally close it.

I have done this more times than I would like to admit.

The problem usually is not the tool. The problem is adding tools before you have defined exactly what job they are supposed to do.

A tool without a clear job is just noise.

The 5 tools that stayed in my stack all share one thing: I know precisely what problem they solve and when I reach for them. No overlap. No confusion. No "wait, should I use this or that?"

That clarity is underrated.

The Stack

1. Claude (Anthropic)

I use Claude for almost everything that involves thinking or writing.

Not because it produces the smoothest output. ChatGPT is cleaner for certain tasks. Other tools have better integrations.

But Claude handles nuance differently.

When I am working through a client brand positioning, or writing a framework piece that actually needs to sound like the person, or designing an agent workflow where the logic has to hold up under real conditions, Claude is where I end up.

What made me stay: give it a messy, half-formed brief and it does not just complete the task. It processes the idea first. That distinction matters when you are running creative work at scale.

At AtheonX, a core part of what we deliver is content that sounds like the CEO, not like an agency. Claude is the tool that makes that achievable.

What I use it for: client strategy docs, long-form content drafts, agent system design, anything that needs a thinking partner when the team is offline.

2. Perplexity

My replacement for Google, at least for research.

For finding recent data, checking a claim, understanding a market quickly, Perplexity is faster and more useful than a standard search.

The reason I trust it: it cites sources. Every answer comes with links. When I am writing content that makes specific claims, for clients who are public-facing executives, I can verify instead of hoping the AI got it right.

I have seen too many AI-written pieces get embarrassed by a fact that turned out to be fabricated. Perplexity does not eliminate that risk, but it reduces it significantly.

What I use it for: pre-writing research, market sizing, fact-checking claims before publishing, quick competitive scans.

3. Make (formerly Integromat)

The automation layer that connects everything else.

I have tried Zapier. I have tested n8n. For complex, multi-step workflows running in production daily, I keep coming back to Make.

One reason: the visual workflow builder makes debugging fast.

Every automation breaks eventually. The question is how long it takes to figure out why. With Make, I can see exactly which step failed and what the error was. That visibility alone has saved hours I do not want to think about.

At AtheonX, a major part of our value to clients is building content operations that run without constant oversight. Make is how we deliver that.

What I use it for: connecting publishing platforms, client reporting pipelines, content distribution workflows, AI output routing.

4. ElevenLabs

Voice and audio generation.

This one surprises people. But audio content is underbuilt in most business content strategies.

Most CEOs can write. Many can record video with coaching. Almost none have a consistent audio presence, even as audio consumption grows. Edison Research 2024 Infinite Dial report found that nearly half of Americans listen to podcasts monthly.

The problem for busy executives: recording takes time and creates a dependency on their schedule.

ElevenLabs removes that bottleneck. We clone the CEO voice once, in a proper session, and can produce audio content without scheduling around their calendar.

The quality has reached a point where listeners engage the same way they engage with human-recorded content. I have tested it with real audiences. Retention holds.

What I use it for: audio versions of long-form client content, voice-over for short video, podcast experiments.

5. Notion

My operating system.

Strategy docs, content calendars, client onboarding, meeting notes, ideas. Everything lives in Notion.

I have tried replacing it twice. Coda first. Then a combination of Linear and Confluence. Both times I came back.

Notion is not the most exciting tool on this list. It does not have a new feature that changes how I work.

What it has is structure that stays organized without constant maintenance. After years of use, the shape of my thinking lives there. Rebuilding that elsewhere costs more than any performance gain from switching.

What I use it for: knowledge management, project tracking, content planning, client workspace.

Why 15 Tools Did Not Make It

This is more useful than the list above. Here is why tools do not survive.

They solved a problem I did not actually have.

A lot of AI tools are built for pain points that feel real in a demo but do not match your actual workflow. "AI email management" sounds good until you realize the issue is not volume. It is that every email requiring attention requires a judgment call no tool can make.

I have downloaded tools, used them carefully for a week, and realized I had spent six hours solving a two-hour annual problem.

They needed constant maintenance to stay useful.

Some automation tools require regular adjustment as data sources shift. The time spent maintaining them exceeded the time they saved. That is a net negative.

The tools worth keeping get more useful over time. Not more demanding.

The output quality was not consistent.

For internal brainstorming, 70% accuracy is fine. For client deliverables and published content, it is not.

Some tools I tested had spectacular highs and bad lows. That variance is worse than slightly lower average quality. Consistent and good beats unpredictable and occasionally brilliant.

They duplicated something that already worked.

There are dozens of AI writing assistants. I needed one. I tested several, picked the one that fits my workflow, and stopped testing the rest.

At some point, the ROI of testing alternatives drops below zero.

What to Take from This

The goal is not a big stack.

The goal is a small stack that runs so quietly you forget it is there.

Every tool you add is something that can break, demand updates, and require your attention when it misbehaves. Add them deliberately or do not add them.

My criteria now: does this solve a problem that costs me real time today? Can I validate that with one week of honest testing? Does it introduce a maintenance burden I am willing to carry?

Most tools fail those questions.

The five that passed are the ones I still use.

If you want to build a content operation that runs on AI infrastructure without breaking every other week, that is what we do at AtheonX.

Book a call with my team and we will walk through what a system like this looks like for your business.

Related: how Jackson runs AI agents as an executive team and work with Jackson on AI systems.

Jackson