I broke down the daily, weekly, and monthly tasks of 15 professions — tax accountants, lawyers, doctors, architects, chefs, professors, developers, designers, marketers, sales reps, real estate agents, logistics managers, content creators, accountants, and executive assistants. The question for each task was simple: does performing it require the credentials or specialized training of that profession? If yes, professional task. If no, common task.
Seventeen out of 25 recurring tasks are common across all 15 professions. That's 68%.
Frequency
Total tasks
Common
Professional
Common %
Daily
10
8
2
80%
Weekly
8
5
3
63%
Monthly
7
4
3
57%
Total
25
17
8
68%
Daily is the most lopsided. Doctors see patients every day. Developers write code every day. But the eight tasks wrapping that core work — replying to emails, coordinating schedules, searching for information, keeping records, writing documents — have nothing to do with the profession. Weekly, professional tasks like case research and design reviews start appearing, but 63% is still common. Monthly, 57%.
Experts spend less than a third of their time on expert work.
That might be uncomfortable to hear. People spend years building expertise, and then someone says two-thirds of their job doesn't actually require it. But look at today honestly. How many hours required your specialized knowledge? For a tax accountant: the time interpreting tax law and building a strategy. For a doctor: diagnosing a patient and choosing a treatment. For a developer: designing architecture and writing core logic. Subtract those from the day. What's left is email, scheduling, admin, documentation. That's not tax accounting or medicine or engineering. That's just work.
The 68% AI is coming for is exactly this territory. It's not being taken — it's being handed back so the other 32% gets more of you.
Most conversations about AI adoption get this wrong because they happen at the job level. "AI for accountants." "AI for marketers." "AI for developers." Job-specific guides, job-specific prompt templates. Not wrong, but limited. A job title is an external label. The actual work is a bundle of tasks underneath it. A tax accountant's day includes interpreting tax law, but also replying to emails. A chef's day includes cooking, but also ordering ingredients and coordinating with suppliers. People who look like they're in completely different worlds at the job level overlap remarkably at the task level.
The data confirms this. The common tasks are the same 17 tasks across all 15 professions.
But handing off 68% to AI isn't as simple as saying "sort my email" and "summarize my meeting notes." Even for the same common task, the context differs by profession — and how that context gets designed determines whether AI's output is generic or actually useful.
Take email replies. Both a tax accountant and a real estate agent write them every day. Same task. But tell AI "write a reply" without context and both get something mediocre. For the tax accountant, the variables that matter are: filing type, stage in the process, sensitivity from prior audit history, and tone calibrated to the client relationship. Set those up, and AI doesn't produce "Thank you for your submission, it is under review." It produces "I've reviewed the revenue documents you sent. Working on the VAT interim filing — looks like three purchase tax invoices may be missing."
For the real estate agent, the variable set is completely different: client type, transaction stage, urgency, and whether the tone should be ROI-focused for investors or lifestyle-focused for residents. Same task, same AI — but once the variables are designed, it operates like a tax accountant's assistant in one context and a real estate agent's in another.
I call this framework AX — AI Experience. Not which AI tool to use, but how to design the context variables that make the same tool work differently for every profession. Traditional automation was fixed input, fixed output: every profession needed its own system built from scratch. AX design builds the structure once per task and swaps in context. Seventeen common tasks, designed once, cover fifteen professions.
A tax accountant can write a good email. But if that hour went to designing one more tax-saving strategy, the client gets more value. Handing off 68% to AI isn't a concession. It's the only way to go deeper in the 32% that actually requires you.