Elon Musk made a point recently that's been rattling around in my head.
He compared early computing to human calculators. One laptop with a spreadsheet replaced entire rooms of people doing math by hand. Then he went further: if even a few cells in that spreadsheet still required a human to fill them in, you couldn't compete against a spreadsheet that ran entirely on the computer.
That's the part worth sitting with. Not the AI-replaces-everything framing. The bottleneck math.
One manual step in a digital workflow doesn't just slow things down. It becomes the ceiling for everything around it.
The spreadsheet analogy in your actual business
Think about quoting a job. Your CRM captures the lead. An AI assistant pulls in the relevant specs from an intake form. Pricing logic runs against your rate sheet. A draft quote gets generated and formatted.
But if one person still has to eyeball every quote before it goes out, the speed of that entire pipeline is limited to however many quotes that person can review per hour. Doesn't matter how fast the upstream steps got. The bottleneck sets the pace.
Now picture a competitor whose quoting process runs end to end without a human pause. They respond in minutes. You respond in hours, maybe the next morning.
That gap compounds. Not just in speed, but in close rates, customer perception, and how many leads you can afford to chase in a week.
This isn't theoretical
You can find the same pattern everywhere in a typical small business:
Invoice follow-up. An AI drafts the reminder email, but someone still has to decide when to send it and whether the tone is right. Meanwhile, a competitor's system sends graduated follow-ups on a schedule, adjusting language based on payment history.
Customer support routing. Your chatbot handles the easy questions, but anything unusual goes into a queue for a person to triage. A fully automated system classifies, routes, and responds to 80% of tickets without waiting for anyone.
Scheduling. Your team uses an AI tool to suggest times, but a coordinator still confirms every booking. Somewhere, a competitor's calendar just handles it.
None of these are dramatic, Terminator-level automation stories. They're small operational frictions. But stack enough of them and you get a company that moves at the speed of its slowest human checkpoint.
Musk's conclusion, and where I partly disagree
Musk's broader claim is that companies running entirely on AI will demolish companies that aren't. I think he's right about the direction. The gap between fully automated and partially automated workflows is real and it's widening.
Where I'd push back is on the timeline. Most SMBs don't get to "fully AI" in one step. The technology isn't there yet for every workflow, and ripping out human judgment from processes that involve real stakes, like hiring or contract negotiation, is not something you should rush.
The actual risk isn't that AI replaces your company overnight. It's that a new competitor, one without legacy workflows, without approval chains, without meetings about meetings, starts operating at a tempo you can't match. Not because they're smarter. Because they just have fewer manual steps between input and output.
What to actually do about it
The move isn't to panic or to try automating everything by next quarter. It's to get honest about where your manual steps are and which ones are genuinely adding value versus which ones exist out of habit.
Start by mapping your most common workflows end to end. Quote-to-close. Lead intake to first response. Support ticket to resolution. Look for the places where work stops moving and waits for a person.
Then ask two questions about each pause:
- Is this human step actually improving the outcome, or is it just tradition?
- If a competitor removed this step entirely, would their customers notice a difference?
Some of those steps will turn out to be important. Quality checks on complex custom work, relationship calls with high-value accounts, judgment calls that require context an AI can't access yet. Keep those.
But some will turn out to be habits dressed up as necessities. The weekly review that could be a daily automated scan. The manual data entry that exists because two systems don't talk to each other. The approval that nobody actually reads.
Those are your ceiling. Every one you remove makes the entire workflow around it faster.
The transition window for this kind of operational tightening is real. It's also shorter than most people think. The companies that use it well won't be the ones who automated the most. They'll be the ones who figured out which manual steps were worth keeping and got rid of the rest.