Two founders can open the same ChatGPT account on the same Monday. One walks away with a finished client report and a cleaner pipeline. The other walks away with forty half-useful drafts and a bigger to-do list. The tool is identical. What's different is whether anyone decided what the tool was for. On a recent episode of Real with Ritesh, "ChatGPT for Business: Powerful AI Tool or Bigger Chaos?", I made the case that this gap is now the whole game. Here's what the data says, what ChatGPT can actually do in 2026, and how to add it without adding noise.
- OpenAI launched ChatGPT Work on July 9, 2026: an agent that works across apps and files and produces finished documents, spreadsheets, and slides (BNN Bloomberg, 2026)
- 88% of organizations use AI regularly, but only 39% report any EBIT impact and about 6% report 5% or more (McKinsey, State of AI 2025)
- Workflow redesign has the strongest link to EBIT impact, yet just 21% of adopters have fundamentally redesigned one (McKinsey, 2025)
- Adopt in three stages, Assistance, Collaboration, then Execution, and only automate a workflow once it's documented and has approval rules
- Start with one repeatable workflow, name the approver, and measure the result before adding a second
ChatGPT Stopped Being a Chat Window
On July 9, 2026, OpenAI launched ChatGPT Work, an agent that can act across a user's apps and files and produce finished documents, spreadsheets, presentations, reports, and websites (BNN Bloomberg, "OpenAI launches ChatGPT Work, deepening race for workplace AI tools," July 9, 2026). It's no longer a tool you ask a question. It's one you hand a project.
That's a real shift for a founder. The old pattern was prompt, read, copy, paste, fix. The new one is closer to delegation: describe the outcome, let the agent gather context from connected tools, and review what comes back. It can spend hours on a single project and keep going while you're away from your screen.
OpenAI says you can follow its progress, answer questions, change direction, and approve important actions as it works, and that actions requiring approval pause until you've reviewed them (OpenAI, "ChatGPT is now a partner for your most ambitious work," July 2026). That detail matters more than the headline. The approval step is where your business rules live.
So here's my take from the episode: the feature list isn't the story. Every founder will have access to the same agent. The advantage goes to whoever has the clearer goal, the cleaner workflow, and the sharper approval rules to point it at.
Where Does the Chaos Come From?
In its State of AI 2025 survey, McKinsey found 88% of respondents say their organizations regularly use AI in at least one business function, yet only 39% report any EBIT impact, and roughly 6% qualify as high performers with 5% or more of EBIT attributable to AI (McKinsey QuantumBlack, "The state of AI in 2025: Agents, innovation, and transformation," November 2025). Use is everywhere. Results aren't.
The third bar is the one to stare at. McKinsey found that fundamentally redesigning workflows has the strongest link to EBIT impact among all the organizational changes it tested, yet only 21% of gen AI adopters had done it for even one workflow (McKinsey, 2025). Most companies bolted a new tool onto an old process and hoped.
That doesn't mean ChatGPT doesn't help people. OpenAI's December 2025 enterprise report, drawing on a survey of 9,000 workers across nearly 100 enterprises, found workers save an average of 40 to 60 minutes a day, with heavier technical roles reporting 60 to 80 (OpenAI, "The state of enterprise AI," December 2025, via People Management). Personal time savings are real. The chaos starts when nobody converts them into a business result.
Read those two numbers together and the problem is obvious. An hour saved per person that gets spent producing more unreviewed output isn't leverage. It's volume. More summaries, more drafts, more decks, and the same unclear priorities underneath. If you want the deeper version of why context beats prompting, I covered it in the context engineering breakdown.
Context In, Judgment Stays Yours
ChatGPT Work is built to pull from your connected apps and files, and that's exactly why context is the biggest lever you have. An agent working from last quarter's pricing sheet will produce a beautiful, confident proposal at last quarter's price. Connected, current context cuts that risk. Stale or messy context multiplies it.
That makes data governance a founder topic, not an IT footnote. Decide what the agent can see, who can change what it sees, and which folders are the single source of truth. I walked through the governance side for another agent in the Claude workflow governance post, and the same rules apply here.
The other half is judgment. ChatGPT can organize, draft, and execute. It can't decide which risks matter most, what "good enough" looks like for your brand, which offer is actually profitable, or which customer you should stop serving. Those calls are yours, and they're the reason a founder is in the room.
Voice and file handling make working with it feel more natural. You can talk through a problem on a walk or drop in a messy spreadsheet. Nice. But a smoother interface only makes the input easier. It doesn't replace the decision about what to do with the output.
What Are the Three Levels of ChatGPT Business Adoption?
Adopt ChatGPT in three stages, and don't skip one. In McKinsey's 2025 survey, 62% of organizations were experimenting with AI agents but only 23% were scaling them (McKinsey, State of AI 2025), which tells you most teams are jumping to automation before the groundwork exists. Each level below has a prerequisite, and the founder's role changes at every step.
| Level | What It Does | Prerequisite | Founder Role |
|---|---|---|---|
| Assistance | Thinking, writing, research, summarizing | Basic AI literacy | Overseer of output quality |
| Collaboration | Working with projects, files, and team context | Organized information and workflows | Context provider and editor |
| Execution | Moving work through a process with approvals and outputs | Clear, documented workflows and rules | Workflow strategist and approver |
Assistance is where most of us started, and it's low-risk because a human reads everything. Collaboration adds shared context, so it only works if your information is organized. Execution is the one that scares people, and it should. It's only safe once the workflow is written down and the approval rules are explicit.
The mistake I see is founders trying to buy their way to Level 3. They switch on the agent, connect everything, and hope the process reveals itself. It doesn't. Automation doesn't create clarity. It just runs whatever you had at faster speed, mess included.
The BAY Framework for Founders Using ChatGPT
In the episode I use a simple lens, BAY, to point ChatGPT at the work that actually moves a founder's business: Build, Amplify, Elevate. It maps neatly onto the three pillars this whole site is organized around, and it keeps the tool tied to goals instead of novelty.
B: Build systems, documentation, and decision frameworks
Use ChatGPT to turn what lives in your head into written systems: SOPs, checklists, onboarding docs, and decision rules. This is the highest-return use because it fixes the very thing McKinsey says matters most, the workflow itself. Write the process first, then let the tool help you keep it consistent.
A: Amplify communication, content, and authority
Use it to draft, adapt, and repurpose your ideas across email, proposals, posts, and talks so your voice shows up more often with less friction. The rule: it drafts, you decide. Your authority is built on your judgment, and a founder who publishes unreviewed output is spending that trust for nothing.
E: Elevate your own role
The point of all this is to move you up. If ChatGPT absorbs routine drafting, your hours shift toward strategy, relationships, and the high-stakes calls only you can make. If your calendar looks the same after adopting it, you've added a tool without changing the job.
How Do You Start Without Creating Chaos?
Start with one workflow, not the whole company. Pick something repeatable and visible that ties to a real goal, such as weekly client reporting, proposal drafting, or lead follow-up. McKinsey's finding that only 21% of adopters have redesigned even one workflow (2025) means a single well-run workflow already puts you ahead of most of the market. Then run it through five steps:
1. Find the pain
Look for inconsistent information, slow decisions, or work that stalls in someone's inbox. Those open loops are where an agent earns its keep.
2. Document the workflow
Write down each step, the inputs, and what "done" looks like. If you can't describe it to a new hire, you can't hand it to an agent.
3. Define the approval rule
Decide what gets checked, by whom, and what's never sent without a human. ChatGPT Work is designed so actions that require approval pause for review (OpenAI, 2026), so use that gate deliberately.
4. Measure before and after
Track time to completion, error rate, and consistency. Watch cost too: ChatGPT Work usage is reportedly metered rather than flat-rate, with no published per-task price at launch (Digital Applied, "ChatGPT Work: OpenAI's Agent That Ships Finished Work," 2026), so an unmeasured workflow can quietly become an unmeasured bill.
5. Expand only after it works
Add the second workflow when the first one is stable, reviewed, and paying for itself. Not before.
Looking ahead, the next layer is research: source-backed, evidence-first decisions, where tools like Perplexity help founders check claims before they act. That's a topic for another episode, but the principle is the same. Better inputs and clearer rules beat more output.
Want AI Working Inside Your Business, Not Around It?
I work with founders on picking the first workflow, writing the approval rules, and building systems that make AI a measurable part of the business instead of another tab. If you're ready to move from experimenting to a system that runs, let's talk.
Book a Strategy Call →Frequently Asked Questions
What is ChatGPT Work and how is it different from regular ChatGPT?
ChatGPT Work is an agent OpenAI launched on July 9, 2026. Instead of answering one prompt, it gathers context from connected apps and files, works through a multi-step project, and produces finished documents, spreadsheets, and presentations. You can follow its progress and approve important actions as it goes (OpenAI, July 2026).
Does ChatGPT actually save businesses time?
For individuals, yes. OpenAI's December 2025 enterprise report found workers save an average of 40 to 60 minutes a day. At the company level the picture is weaker: McKinsey found only 39% of organizations report any EBIT impact from AI (McKinsey, State of AI 2025), so the gain depends on how the work is structured.
Where should a founder start with ChatGPT in the business?
Start with one repeatable, visible workflow tied to a business goal, such as weekly client reporting or proposal drafting. Document the steps, define who approves the output, and measure the result. McKinsey found workflow redesign has the strongest link to EBIT impact, yet only 21% of adopters have redesigned even one (McKinsey, State of AI 2025).
Can ChatGPT replace a founder's judgment?
No. It can organize, draft, and execute, but the founder still decides which risks matter, what quality looks like, which offer to sell, and who the ideal customer is. MIT's NANDA project found 95% of generative AI pilots showed no measurable P&L impact (July 2025), and missing direction, not missing capability, is a common reason.
ChatGPT in 2026 is a powerful tool, and it will produce as much chaos as clarity depending on what you point it at. The founders who win won't be the ones with the most output. They'll be the ones with a clear goal, a documented workflow, and a person who approves the result.
Pick one workflow this week and write it down. For the governance layer, read how Claude AI is changing business workflows, or book a strategy call and we'll map yours.
Sources
- BNN Bloomberg, OpenAI launches ChatGPT Work, deepening race for workplace AI tools, retrieved 2026-09-21, July 9, 2026
- OpenAI, ChatGPT is now a partner for your most ambitious work, retrieved 2026-09-21 (via search summary; page blocked automated fetch), July 2026
- Digital Applied, ChatGPT Work: OpenAI's Agent That Ships Finished Work, retrieved 2026-09-21, 2026
- McKinsey QuantumBlack, The state of AI in 2025: Agents, innovation, and transformation, retrieved 2026-09-21 (figures confirmed via secondary summaries, e.g. CX Today), November 2025
- People Management, AI tools save workers up to an hour a day, OpenAI report finds (citing OpenAI, The state of enterprise AI), retrieved 2026-09-21, December 11, 2025
- Yahoo Finance, MIT report: 95% of generative AI pilots at companies are failing (citing MIT NANDA, The GenAI Divide: State of AI in Business 2025), retrieved 2026-09-21, July 2025
