On the afternoon of November 4, 2026, OpenAI’s status page posted an incident about error rates climbing across paid ChatGPT plans. If you were refreshing the tab that afternoon with a half-finished email going cold, I’m sorry, that’s annoying. OpenAI had a fix going within minutes and gave the all-clear about 2 hours later, so that version of ChatGPT not working sorts itself out.
When a business owner tells me ChatGPT isn’t working for them, they’re usually talking about their day-to-day. It loads fine and the status page is green. The subscription renews every month too. Meanwhile the follow-ups still get written manually, proposals get rebuilt from scratch each time, and somebody on your team is pasting the same 3 paragraphs of company background into a blank chat for what has to be the 40th time.
I was stuck there myself for a long time. I spent 18 months building Claude projects that wrote like a polite stranger. It took me that long to figure out what they were missing, which was my context, my voice and one proper job to do every day. As of November 2026, my laptop time sits at about 3 hours a week, down from 90. You can build the same setup inside ChatGPT, so I’ve laid it all out below.
Is ChatGPT down, or is it your setup?
Checking which problem you’ve got takes about 30 seconds. Open status.openai.com in a new tab and look at the Conversations and Login rows. A yellow or red mark there means it’s on OpenAI’s side, so the only fix is a coffee and a few minutes.
Everything showing green? Work through these on your end, top to bottom.
| What you’re seeing | Try this first |
|---|---|
| The page won’t load or keeps spinning | Do a hard refresh, then open ChatGPT in a private window to rule out a browser extension |
| An error appears halfway through a reply | Regenerate once, then copy your prompt into a fresh chat |
| It works on your phone but not your laptop | Clear the site’s cookies or switch networks, since office VPNs and firewalls can block it |
| A message says you’ve hit a limit | Wait for your plan’s usage cap to reset or switch to a lighter model |
| It loads, it answers, and your week looks exactly the same | That’s the problem the rest of this post is about |
ChatGPT makes a brilliant advisor and a forgetful employee
OpenAI studied 1.5 million conversations for its September 2025 usage report. It found 49% of messages are people asking for something, 40% are people getting it to do something and 11% are people simply expressing themselves. OpenAI’s own conclusion was that people value ChatGPT most as an advisor.
That makes sense for how most people use it. You ask a question, get your answer and close the tab. Running a business needs something else from it, though. You need it to know who you are, what you sell and who your clients are. You also need it doing the same job the same way every Tuesday.
MIT’s NANDA team was blunt about this in its July 2025 report on enterprise AI. ChatGPT “forgets context, doesn’t learn, and can’t evolve.” The same research found workers at over 90% of the companies surveyed regularly using personal AI tools for work, while only 40% of those companies had bought an official subscription.
Chances are your team is already using it, each person in their own tab with their own prompts. None of that gets shared with the rest of the team or carried over into the next chat.
I mean, would you keep paying an assistant who forgot your prices and your best client’s name every single morning?
Boston Consulting Group puts numbers on this in its February 2026 piece on AI and the workforce. BCG says about 10% of the value from AI comes from the algorithms, 20% from the technology needed to run them and the remaining 70% from the people side.
ChatGPT sits in that first 10%. Most of the owners I talk to spend almost all their time and energy picking the right tool, which is the smallest slice of the 3, then wonder why their week still looks the same.
5 signs your ChatGPT subscription has turned into shelfware
Shelfware is software you keep paying for after everyone’s quietly stopped using it. Go through the 5 below and see how many sound like your business. Even one of them is usually enough to explain why the subscription has started to feel like a monthly donation.
Every chat starts from a blank box
Try a test this afternoon. Ask 3 people on your team to open their last 5 ChatGPT chats and count how many start with pasted background about your business, your offer or your clients.
2 out of 5 or more tells you ChatGPT is mostly being used as a notepad you pay for. All that pasted background should be written down once and loaded in for good.
Only one person still logs in
Most AI rollouts start with a kickoff meeting where everyone’s excited and nodding along. 3 months later the login history tends to show 1 or 2 names on it, quite often including yours.
MIT found 60% of organizations evaluated enterprise-grade AI tools, 20% got as far as a pilot and only 5% reached production. From what I’ve seen, most of that drop happens without anyone noticing, because the team was never shown what the tool was meant to do for them.
Everything it writes needs a full rewrite
Time yourself on the next draft it hands you. An edit that takes longer than writing it fresh means ChatGPT has no idea how you sound.
This was my problem for the longest time. The polite-stranger drafts only stopped after I loaded 100,000+ words of my own thinking into the system, which I call Voice DNA. I walk through a full year of running it in my AI employee breakdown.
You can’t name one number it moved
Ask yourself what ChatGPT changed last month and answer with a number. Any number works, hours saved on follow-up or the days it takes to get a proposal out the door. My AI consulting ROI guide shows how to measure it and when the return tends to show.
Most people I ask can’t answer that, usually because no starting number was ever written down. Write today’s number down before you change anything.
You bought ChatGPT but you’re still doing everything manually
MIT found the best returns often sit in the functions everyone ignores, operations and finance, while most AI budgets chase sales and marketing tools. My own system summarizes my inbox and pre-writes my drafts every 2 hours. Those 2 jobs are a big part of how my laptop time got down to about 3 hours a week.
Follow-up is the friendliest place to start. For the wider picture of which processes can run without you, read my guide to AI business automation.
Why AI implementation stops working in month 2
AI implementation usually goes great in week 1. It’s somewhere around month 2 that people stop opening it. In my experience it’s nearly always one of the 3 reasons in this table.
| What goes wrong | What it looks like in your business | The fix |
|---|---|---|
| It has no owner | The person who set it up got busy, so the instructions never get updated | Name one owner and give them 30 minutes a week to tune it |
| Feedback never goes back in | Your team fixes the same mistakes in every draft and tells the AI none of it | Paste the corrected version into the project’s instructions every Friday |
| The scope keeps growing | You started with follow-ups and now it’s meant to run marketing, hiring and support | Finish one workflow before you start the next one |
I’d start with the feedback row. ChatGPT can’t learn from corrections it never sees, so whatever it got wrong in draft 1 will still be wrong in draft 400.
What changed when I stopped blaming the tool
For a long time I was sure I needed a better model. I tried every new one as it came out, switched back and forth, and kept getting the same polite stranger back.
In the end it came down to 2 changes. The first was giving it context it keeps for good. I loaded my voice, my offers, my clients and the way I make decisions into a Claude system wired to my files, one time.
The second was giving it one job to repeat. Mine summarizes my email and pre-writes my drafts every 2 hours, the same way every day, while I’m on my phone doing something else.
I sold my 100-person agency in 2021 and started again. I passed $100K a month in month 11 of the rebuild with 2 people. As of November 2026, First Movers AI runs at about $250K a month with 15 humans. The AI side of it sits on a $100 a month Claude plan.
I moved to Claude for reasons I lay out in my Claude vs ChatGPT comparison. You don’t need to switch to get this working, though. ChatGPT Projects and custom GPTs give you the same 2 things, a place for your context to live and instructions it follows every time.
If you’d rather have someone build it for you, that’s what my team does in an AI consulting build. We map it with you in week 1 and hand it over at the end of 2 months.
A 30-day plan to get ChatGPT working in your business
Start with one job, please, and leave the company-wide rollout for later. This is the month I’d run myself if I were starting from zero next Monday.
| Week | What you do | How you know it’s working |
|---|---|---|
| Week 1 | Pick one job that eats hours every week and write down how long it takes today | You have a baseline number in writing |
| Week 2 | Build a context file with voice samples, offers, prices and your 10 most common client questions, then load it into a ChatGPT Project | First drafts need only light edits |
| Week 3 | Turn the job into saved instructions with the same inputs every time, then hand it to one teammate | Someone other than you runs it start to finish |
| Week 4 | Measure against week 1, then connect the output to where the work lives, your inbox or CRM | The week 1 number has moved |
Please don’t rush week 2. Load in the emails you’re proud of and the proposals that closed, because ChatGPT picks up your voice from real examples much faster than from a description of how you write.
You can absolutely run all 4 weeks yourself.
For the do-it-yourself route, AI Labs has 65 courses and 2 live sessions a week as of November 2026, where you can work through it with us. Small teams will find a fuller version of this plan in my guide to AI consulting for small business.
The ChatGPT you already pay for can be your best hire this year
BCG puts the share of organizations getting substantial financial gains from AI at about 5%. Their software looks much like everyone else’s. The big gap BCG found is in their people, since they plan to upskill over half their staff on AI while the laggards plan for 20%.
None of that needs a new tool. It all happens in the 70% BCG keeps coming back to, your people and the way work moves through your business. Luckily that’s also the cheapest part to fix.
I love that, because it means getting into that 5% comes down to habits your team can start on Monday.
Questions about ChatGPT not working for business
Why is ChatGPT not working right now?
ChatGPT not working right now usually comes down to an outage on OpenAI’s side or something on your end, like a browser extension, a blocked network or a usage cap on your plan. Check status.openai.com first. With everything green there, try a private window, a different network and a fresh chat before assuming anything bigger has broken.
Why is ChatGPT not working for my business when I use it every day?
Daily ChatGPT use and a working business system are two separate things. Daily use usually means one person asking questions in a blank chat. A business system keeps your voice and client context loaded and runs the same job the same way each time. You’ll see it working in a number you track, such as hours saved on follow-up each week.
Should I switch from ChatGPT to Claude?
Sometimes. Claude has been the better writer for First Movers, which is why the business runs on it. Switching tools alone rarely fixes a business problem, though, because the missing context and the missing workflow come along with you to the new tool. Fix those first, then compare the two on your own real tasks.
How long does it take to get ChatGPT working for a business?
Getting ChatGPT working for a business takes about 30 days for one workflow, from a baseline number to a measured result. A full done-for-you AI build from First Movers, with Voice DNA, custom skills and the systems connected, takes about 2 months and starts with a week of mapping your tools and goals.
Can I get ChatGPT working for my business on my own?
Yes. Getting ChatGPT working in-house is very doable for 1 or 2 workflows, starting with a baseline number and a context file of your voice and offers. MIT’s 2025 research found outside partners reached deployment about twice as often as internal builds, so outside help is worth weighing for bigger projects that touch several systems at once.
What your Tuesday looks like once it works
Say it’s a Tuesday in January. Your follow-ups went out overnight in your voice. The proposal draft is sitting in your inbox with your prices already in it, which leaves you with just the part that needs your judgment.
That’s the build my team installs for clients. On a free call we map your tools and the one job worth handing over first, so you see the plan and the cost before you commit to anything. My breakdown of what AI consulting costs lays out the numbers ahead of time. Whenever you’re ready, book a free AI strategy call.
So next time you type chatgpt not working into Google, I hope it’s just OpenAI having a bad afternoon while your own setup is busy doing its job in the background.
End of content.
Final word count: 2,447
