How much time does AI save in your job?
I turned Anthropic’s public data on how people use Claude into something you can search. Guess first, then find your job. Then read what the number leaves out, because that part matters more.
Guess first
The average task people brought to Claude in May would take a person an estimated 4 h 44 min alone. How long do you think it takes with AI?
Find your job
Type a job title. There are 717 here, from fundraisers to roofers. Each one is built from conversations that look like that job’s tasks.
Pick one from the list. Use the arrow keys, then Enter.
Jobs people ask me about
| Job | Alone | With AI | Faster |
|---|---|---|---|
| Executive Secretaries and Executive Administrative Assistants | 7 h 59 min | 32 min | 15.2× |
| Training and Development Specialists | 10 h 7 min | 43 min | 14.2× |
| Meeting, Convention, and Event Planners | 9 h 7 min | 45 min | 12.2× |
| Accountants and Auditors | 9 h 29 min | 47 min | 12.1× |
| Fundraising Managers | 6 h 41 min | 34 min | 11.8× |
| Office Clerks, General | 3 h 22 min | 19 min | 10.9× |
| Secretaries and Administrative Assistants, Except Legal, Medical, and Executive | 5 h 26 min | 30 min | 10.8× |
| Fundraisers | 5 h 40 min | 34 min | 10.1× |
| Chief Executives | 6 h 55 min | 42 min | 9.8× |
| Marketing Managers | 9 h 23 min | 58 min | 9.7× |
| General and Operations Managers | 10 h 10 min | 1 h 6 min | 9.3× |
| Administrative Services Managers | 3 h 30 min | 25 min | 8.5× |
| Project Management Specialists | 5 h 58 min | 43 min | 8.4× |
| Public Relations Specialists | 5 h 22 min | 41 min | 7.8× |
| Human Resources Specialists | 3 h 28 min | 28 min | 7.5× |
| Software Developers | 5 h 54 min | 48 min | 7.4× |
| Receptionists and Information Clerks | 4 h 51 min | 40 min | 7.3× |
| Bookkeeping, Accounting, and Auditing Clerks | 2 h 31 min | 22 min | 6.9× |
| Food Service Managers | 2 h 58 min | 29 min | 6.2× |
| Social and Community Service Managers | 3 h 29 min | 34 min | 6.1× |
| Real Estate Sales Agents | 3 h 27 min | 35 min | 5.9× |
| Human Resources Managers | 6 h 26 min | 1 h 12 min | 5.4× |
| Elementary School Teachers, Except Special Education Mostly personal conversations | 4 h 41 min | 56 min | 5.0× |
| Clergy Mostly personal conversations | 4 h 48 min | 1 h | 4.8× |
| Customer Service Representatives Mostly personal conversations | 47 min | 12 min | 4.0× |
| Registered Nurses Mostly personal conversations | 2 h 21 min | 37 min | 3.9× |
| Child, Family, and School Social Workers Mostly personal conversations | 3 h 20 min | 1 h 8 min | 2.9× |
| Mental Health Counselors Mostly personal conversations | 4 h 24 min | 1 h 52 min | 2.4× |
Sorted from the most times faster to the least. Job titles are O*NET® occupation titles.
Where AI saves the most time, and the least
Office work speeds up the most. In office and administrative jobs, a task estimated at 4 h 17 min alone takes about 25 minutes with AI, and more than half of those conversations are real work.
The smallest savings show up where the task is about a person: community and social service, and health care. There, a task estimated at 3 h 46 min alone still takes 1 h 26 min with AI.
One caution before you read too much into that. In community and social service, only about 1 in 13 of those conversations is work, and nine in ten are personal: people asking for themselves, not someone doing a care job. Health care looks much the same.
I still take something from it. When the conversation is about a person, it doesn’t shrink much, even with AI. At work, that’s the part I’d protect: let AI take the admin and the searching, and keep judgment, connection and trust with people.
Times faster with AI, by kind of work
Community and social service, and health care. Most of these conversations are personal, not work.
See the numbers for all 22
“Times faster” divides the average time alone by the average time with AI.
| Kind of work | Alone | With AI | Faster | Share of conversations | Handed off |
|---|---|---|---|---|---|
| Office and Administrative Support | 4 h 17 min | 25 min | 10.3× | 7.9% | 56% |
| Business and Financial Operations | 6 h 20 min | 43 min | 8.9× | 5.8% | 43% |
| Educational Instruction and Library | 4 h 42 min | 34 min | 8.4× | 12.8% | 43% |
| Management | 6 h 30 min | 48 min | 8.2× | 5.9% | 44% |
| Computer and Mathematical | 5 h 21 min | 42 min | 7.7× | 23.8% | 64% |
| Protective Service | 1 h 45 min | 14 min | 7.5× | 0.33% | 74% |
| Building and Grounds Cleaning and Maintenance | 5 h 10 min | 42 min | 7.4× | 0.13% | 47% |
| Legal | 6 h 19 min | 52 min | 7.3× | 1.0% | 35% |
| Construction and Extraction | 3 h 2 min | 25 min | 7.2× | 0.10% | 66% |
| Arts, Design, Entertainment, Sports, and Media | 4 h 58 min | 43 min | 6.9× | 13.6% | 36% |
| Life, Physical, and Social Science | 5 h 26 min | 47 min | 6.9× | 4.5% | 40% |
| Architecture and Engineering | 5 h 38 min | 50 min | 6.8× | 3.6% | 45% |
| Personal Care and Service | 2 h 13 min | 21 min | 6.3× | 1.2% | 56% |
| Transportation and Material Moving | 2 h 38 min | 26 min | 6.0× | 0.34% | 52% |
| Installation, Maintenance, and Repair | 2 h 38 min | 29 min | 5.5× | 0.62% | 69% |
| Food Preparation and Serving Related | 1 h 38 min | 18 min | 5.4× | 0.48% | 49% |
| Farming, Fishing, and Forestry | 1 h 25 min | 17 min | 5.1× | 0.04% | 65% |
| Production | 5 h 16 min | 1 h 2 min | 5.1× | 1.0% | 64% |
| Sales and Related | 2 h 29 min | 30 min | 4.9× | 9.1% | 44% |
| Healthcare Practitioners and Technical | 2 h 31 min | 42 min | 3.6× | 3.3% | 42% |
| Healthcare Support | 1 h 20 min | 23 min | 3.4× | 0.62% | 39% |
| Community and Social Service | 3 h 46 min | 1 h 26 min | 2.6× | 2.6% | 40% |
Where Wisconsin stands
Wisconsin is tied with Ohio for 34th of 51 (the 50 states and DC) in how much people use Claude for the size of their working-age population. It uses 59% of what its population would predict, down from 61% in April.
Less of it is work, too. In Wisconsin, 36% of these conversations are about work, compared with 41% across the US. Is that a gap, or a head start for whoever closes it first?
Use for the size of the working-age population, every state, May 2026
WisconsinThe state you compare1.0: use in line with population
See every state
| Rank | State | Usage index | About work |
|---|---|---|---|
| 1 | District of Columbia | 3.32 | 44% |
| 2 | California | 1.62 | 42% |
| 3 | New York | 1.55 | 40% |
| 4 | Washington | 1.51 | 36% |
| 5 | Massachusetts | 1.42 | 39% |
| 6 | Colorado | 1.41 | 44% |
| 7 | Utah | 1.21 | 44% |
| 8 | Hawaii | 1.14 | 42% |
| 9 | Nevada | 1.13 | 43% |
| 10 | Oregon | 1.10 | 38% |
| 11 | Maryland | 1.09 | 40% |
| 12 (tie) | Connecticut | 1.07 | 39% |
| 12 (tie) | Virginia | 1.07 | 40% |
| 14 (tie) | Delaware | 1.05 | 36% |
| 14 (tie) | Florida | 1.05 | 48% |
| 16 | New Jersey | 1.02 | 42% |
| 17 | Illinois | 1.00 | 38% |
| 18 | Rhode Island | 0.96 | 40% |
| 19 (tie) | Arizona | 0.95 | 44% |
| 19 (tie) | Texas | 0.95 | 43% |
| 21 | Georgia | 0.93 | 43% |
| 22 (tie) | North Carolina | 0.88 | 42% |
| 22 (tie) | New Hampshire | 0.88 | 40% |
| 24 | Vermont | 0.87 | 41% |
| 25 | Minnesota | 0.81 | 38% |
| 26 | Pennsylvania | 0.80 | 38% |
| 27 | Missouri | 0.70 | 41% |
| 28 | Tennessee | 0.69 | 44% |
| 29 | Idaho | 0.68 | 44% |
| 30 | Maine | 0.65 | 40% |
| 31 | Nebraska | 0.63 | 38% |
| 32 (tie) | Michigan | 0.61 | 40% |
| 32 (tie) | Indiana | 0.61 | 37% |
| 34 (tie) | Ohio | 0.59 | 40% |
| 34 (tie) | Wisconsin | 0.59 | 36% |
| 36 | Montana | 0.58 | 46% |
| 37 | Oklahoma | 0.53 | 39% |
| 38 (tie) | South Carolina | 0.51 | 47% |
| 38 (tie) | Alaska | 0.51 | 42% |
| 40 | New Mexico | 0.49 | 38% |
| 41 (tie) | Louisiana | 0.48 | 41% |
| 41 (tie) | Kansas | 0.48 | 37% |
| 43 | Alabama | 0.47 | 41% |
| 44 | Wyoming | 0.43 | 39% |
| 45 | Iowa | 0.41 | 38% |
| 46 (tie) | Arkansas | 0.39 | 41% |
| 46 (tie) | Kentucky | 0.39 | 40% |
| 48 (tie) | South Dakota | 0.37 | 46% |
| 48 (tie) | North Dakota | 0.37 | 38% |
| 50 | Mississippi | 0.28 | 41% |
| 51 | West Virginia | 0.25 | 41% |
The usage index is Anthropic’s: a state’s share of Claude use divided by its share of the working-age (15 to 64) population. 1.0 means use in line with population.
What people keep using it for
Time saved is only half the story. OpenAI’s economists asked what happens after someone first uses AI for a task outside their own job: do they come back to it the next month?
Customer communication sticks. Legal research doesn’t. More than half came back to customer communication the next month, and 1 in 10 came back to legal research.
Came back to the task the next month
They found something else. When people work outside their own job, they ask the AI to explain things less often. The authors suggest they may be borrowing the expertise without learning it.
That’s the part I watch for. If only the tool understands the work, nobody owns it, and it breaks the day the person who set it up leaves.
What the number leaves out
I’d rather you trust this page for the right reasons. Here’s what these numbers are, and what they aren’t.
- It’s an estimate. An AI read each sampled conversation and estimated how long the task would take a person alone and with AI. Nobody ran a stopwatch.
- It’s about the task, not the person. “Fundraisers” means conversations that look like fundraising tasks, not people with that title.
- It’s one AI, on its individual plans. Claude’s Free, Pro and Max plans, in chat and Cowork. Not company accounts, Claude Code or other AI tools, and only people who already use it.
- It ends when the conversation ends. It can’t see whether the work was used, checked by someone else or done over.
- “Times faster” is a ratio of averages. It divides the average time alone by the average time with AI, so a few very long tasks can move it.
- It’s consistent, which isn’t the same as right. April and May match closely, job by job. That means the estimate is steady. It doesn’t prove it’s accurate.
The biggest thing it leaves out is you. It averages thousands of strangers. Your number lives in one task, with your tools and your people.
Want your real number?
Show me one thing you redo every week. We’ll find out together what’s actually possible, with the people who do it and the tools you already have.
That’s Better Way: one recurring task, $500, online or in person when we can. If you don’t leave with a clear, practical better way forward, I refund the $500.
Or call Sarah, my AI assistant, at (920) 645-0459. She can answer questions and book the same call.
Keep going
- One SOP, written four waysBy hand, in a chat, set up with skills, and with AI agents. Who does each step.
- Prepare the HumansWhere I write about the people side of AI at work.
- The free process assessmentTwelve questions about one process, and a one-page brief.
- Try what I builtAn AI phone assistant, a quote builder and a working dashboard.
- Download the tableAll 717 jobs as a CSV, credits included. Use it in your own deck; credit Anthropic.
Sources and credits
Time by job, kind of work and state. Anthropic Economic Index, data release of June 26, 2026: Claude chat and Cowork conversations, May 2026 (April for Wisconsin’s earlier rank). Maxim Massenkoff, Eva Lyubich, Szymon Sacher, Zoe Hitzig, Shaoyi Zhang, Ryan Heller and Peter McCrory, “Anthropic Economic Index report: Cadences,” Anthropic, June 26, 2026. The data is released by Anthropic under a CC BY license.
What I changed. I kept the occupation, kind-of-work and US state rows, worked out “times faster” as the average time alone divided by the average time with AI, and wrote the plain-language labels. The CSV keeps Anthropic’s values as published; the page rounds them for reading and ranks tied states together. The table I share here is under CC BY 4.0; credit Anthropic when you use it.
Job titles. This page includes information from the O*NET® database by the U.S. Department of Labor, Employment and Training Administration (USDOL/ETA), as used in the Anthropic Economic Index. Used under the CC BY 4.0 license. O*NET® is a trademark of USDOL/ETA. Build With You AI has modified all or some of this information. USDOL/ETA has not approved, endorsed, or tested these modifications.
What people keep using it for. Alex Martin Richmond and Caroline Chin, “Work at the Frontier: How workers are unlocking new ways of working,” OpenAI Economic Research, September 2026, pages 3 to 11. The chart is my own drawing of their published numbers. Their sample is work messages sent on the individual ChatGPT accounts of U.S. users who also have ChatGPT Business accounts, April to July 2026, which they say is not representative of the U.S. workforce.
No affiliation. Build With You AI isn’t affiliated with Anthropic, OpenAI or the U.S. Department of Labor, and none of them reviewed or endorsed this page. Anthropic and Claude are trademarks of Anthropic. OpenAI and ChatGPT are trademarks of OpenAI.
Updated October 31, 2026. Spot a mistake? Email me and I’ll fix it.