The 6 AI tools I actually pay for and use every week in my Chick-fil-A

Every week I get asked some version of “what AI tool should I buy first?” The honest answer is that the right tool depends on the specific problem you are trying to solve. But the more useful answer is to see what an actual operator with a working AI stack pays for.

These are the six tools I use every week inside my Chick-fil-A. Real subscriptions, real use cases, real frustrations.

Heads up before you keep reading: three of these six are Claude products. I am heavily Anthropic-leaning right now because they shipped the operator-friendly features (computer access, browser control, agent workflows) earlier than the alternatives. If that changes, my stack will change. I am writing this in May 2026.

1. Claude (claude.ai, $20/month Pro or $100/month Max)

What I use it for: My primary thinking partner. Long-form analysis, comparing SOPs across documents, drafting performance reviews, restructuring meeting agendas, debugging operational problems by talking through them.

Why this one: Claude handles long context better than the alternatives in my experience. I can paste a 30-page document and ask specific questions and get useful answers without losing fidelity. The output also lands closer to my voice with less prompt engineering.

What frustrates me: I have to remember to start new chats for new topics. If I let one conversation sprawl across five unrelated tasks, the responses get muddier. Discipline of fresh chats matters.

What I would do differently: Used Projects (Claude’s persistent context feature) earlier. I wasted six months retyping operational context every session before I figured out that I could save it once in a project and let Claude carry it forward.

2. Claude Cowork (included with Claude Max)

What I use it for: Anything that requires Claude to actually do something on my computer. Editing my website code, drafting documents directly into my files folder, running a workflow that touches my filesystem and a browser at the same time.

Why this one: This is the unlock that turned Claude from a thinking partner into an operational partner. Most of my “AI work” stopped being chat and started being: tell Claude what I want, give it access to the right tools, let it execute. The website you are reading was largely built and edited in Cowork sessions.

What frustrates me: Still in research preview, which means some workflows hit guardrails I have to work around. Permissions reset between sessions for security reasons, which is correct but adds friction.

What I would do differently: Started using it for my real operational documents sooner. I was using it for one-off experiments for months before I realized I should be running my actual weekly work through it.

3. Claude Browser Extension (free with Claude)

What I use it for: Driving Chrome on autopilot. Logging into vendor portals to pull invoices, navigating through HR systems to update records, posting reviews and responses, taking screenshots of analytics dashboards for my weekly summaries.

Why this one: Most of my repetitive admin work happens inside web browsers. The browser extension lets Claude handle that work without me building a custom integration for every site. Saves me roughly four hours a week of clicking through portals I do not enjoy.

What frustrates me: Some sites detect automation and block it. Sites with strict bot detection or weird auth flows still need me to do them by hand. The list of sites that work is most of them, but not all.

What I would do differently: Set up the extension on my office computer first, my floor laptop second. The use case is heavier in the office than on the floor and I should have prioritized accordingly.

4. ChatGPT Pro ($20/month)

What I use it for: My second opinion. When Claude gives me an output I am unsure about, I run the same prompt through ChatGPT and compare. Also for image generation (training graphics, internal posters, mockups for team meetings) where ChatGPT’s tooling is still ahead.

Why this one: Cross-checking matters when the output is going to drive a real decision. The two models have different strengths and different blind spots. Catching disagreements between them has saved me from multiple bad calls.

What frustrates me: Context fragmentation. Anything I taught Claude has to be re-taught to ChatGPT. Most of the time I am willing to pay that cost. Occasionally I am not, and I skip the cross-check, and occasionally I regret it.

What I would do differently: Cancelled this for the first three months while I was getting Claude usage dialed in. The second-opinion value is real but it is a premium use case, not a starter use case.

5. Gemini (Gemini Advanced, $19.99/month)

What I use it for: Anything Google Workspace integrated. Summarizing long Gmail threads, pulling insights from Google Sheets, drafting directly inside Google Docs without copy-pasting. My entire operation runs on Google Workspace, so Gemini sitting natively inside those apps removes friction.

Why this one: The Workspace integration is the entire reason. If I were on Microsoft 365, I would pay for Copilot for the same reason. The model is good. The placement is the unlock.

What frustrates me: It is the weakest pure reasoning of the three LLMs I use. For complex analysis I default to Claude. For Workspace-native work I default to Gemini. That split is the right call but it means I have to think about which tool to grab.

What I would do differently: Used the Sheets integration sooner. I was pulling data into Claude manually for months before I realized Gemini could analyze the same data inside the sheet without me moving it anywhere.

6. Manus ($39-99/month depending on plan)

What I use it for: Autonomous multi-step research. “Find me the top 10 catering competitors within five miles, pull their pricing for similar-sized orders, summarize their differentiators, send me a comparison.” Manus goes off, browses the web, gathers the data, and returns a report. That kind of task used to take me half a day.

Why this one: It is purpose-built for agentic work that needs to run for 30 to 60 minutes without supervision. Claude can do similar work but Manus is designed for it from the ground up. Different category of tool.

What frustrates me: Sometimes it goes down rabbit holes. Without good guardrails it will spend an hour on the wrong sub-task and come back with low-value output. Discipline in the initial prompt matters more here than with any other tool.

What I would do differently: Started with smaller scoped tasks. The first big task I gave it (a sweeping competitive analysis) was too broad. The first three tasks should have been single-step research jobs to learn what the tool does well.

What this stack costs me

Roughly $180-260/month total, depending on which plans. Plus the operational tools I mentioned in my “10 systems” post (inventory tool, review management tool, etc.) that round out my full AI spend to about $300/month.

The return on those tools is meaningfully six figures annually in margin improvement, time saved, and revenue recovered. The math is not close.

What none of these tools are

None of them are “AI for restaurants” or “AI for operators.” None of them were built specifically for my industry. The lesson there is important: the right AI stack for your operation is almost certainly a combination of general-purpose tools applied to specific operational problems, not a single vertical platform that claims to solve everything.

Vertical “AI for [your industry]” tools tend to be overpriced and underbuilt because their addressable market is small. General-purpose tools have more competition, faster improvement cycles, and lower pricing. Adapt the general tools to your specific work with master prompts and automation. That stack wins.

What to do this week

1. Audit your current tool stack. Write down every AI tool you are paying for right now. Next to each, write the use case and the monthly cost. If you cannot articulate a specific use case for any tool, cancel it this week.

2. Pick one of the six tools above to trial. Not all of them. One. Run it for a full week on real work. The shopping cart approach where operators add six tools at once almost always fails because no single tool gets enough attention to produce results.

3. If you are on Google Workspace and not using Gemini, fix that today. Twenty bucks a month for native integration with the tools your team already lives in. Fastest immediate win on this list.

4. Try Cowork or the Claude Browser Extension for one repetitive web task. Pick the most boring weekly thing you do in a browser. Let Claude handle it once. See what happens.

5. Cancel one tool you are not actually using. Every operator I know has a subscription they forgot to cancel. Recover the cash and use it on something that pays off.

The right AI stack is not the longest stack. It is the smallest stack that covers your highest-leverage workflows. Six tools sounds like a lot. Half of them are interfaces to the same underlying model, which is the point. The interfaces matter as much as the model.

Build the minimum that works for your operation. Run it for six months. Then add or change. Most operators never stop at the right number because they get excited about every new tool launch.

Discipline beats novelty.