What the latest AI usage charts say about your organization.
Every few months a new batch of AI usage charts goes around, and someone forwards one to the leadership team with a question mark. The latest is the seventh edition of a16z's Top 100 Gen AI Consumer Apps, published in early October by the venture firm Andreessen Horowitz. It runs to 21 charts built from card-spending, mobile and web-traffic data.
The report is about consumers, so it measures what people do with their own money and their own phones. That is still useful to an organization, because your staff are those consumers, and the habits they form at home walk in the door with them on Monday. We picked the handful of charts that say something about how a team should plan, and left the vendor horse race alone.
Most people are still early.
The headline figure is modest. According to a16z, nearly half of U.S. consumers say they use AI, but only about a quarter use it daily, and in August 4.5% of eligible U.S. consumers paid for one of the three big chat assistants. a16z notes that share has more than doubled from 2.1% a year earlier.
For a department head who feels behind, this is a reasonable thing to hear. Most of the people you hire and serve are still at the trying stage. There is still time to build careful habits before the tools are everywhere, and careful habits are much easier to set early than to fix later.
People pay for one job done well.
a16z added a new ranking this year: AI products ordered by what U.S. consumers actually spend on them, using YipitData's card panels. Some names rank far higher on spending than on traffic. An email app is fourth by spending in a16z's list, a music generator is seventh, a voice tool tenth and a hardware note-taker sixteenth. a16z also reports that 29 of the top 50 products by spending appear on neither its web nor its mobile traffic list.
The pattern is familiar from any software budget. People open the general chat window often, but they reach for their wallets when a tool does one specific task they used to dread. Inside an organization the same logic holds. The assistant everyone has access to will get used a little by many people, while the narrow tool that clears the intake queue or drafts the monthly board report is the one someone will defend at renewal time.
A few people get most of the value.
The sharpest chart in the report is about concentration. a16z finds the top 1% of AI spenders account for 19.5% of tracked consumer spending, more than the bottom half combined at 16.6%, and that the top 10% account for roughly half. The median payer, by a16z's figures, spends about $25 a month.
The habits your staff form at home walk in the door with them on Monday.
You can see the same curve in most offices without a survey. One coordinator has built herself a drafting routine for every grant report, a facilities manager uses AI to sort work orders, and everyone else opens the tool when they remember. a16z also lists what its heaviest spenders over-index on, and it is mostly automation, agents and image and video generation. The people at the deep end have stopped chatting and started building small systems. That is where your own heavy users are probably heading, and it is worth knowing who they are before they get there on their own.
Pick one tool and teach it properly.
Two figures from a16z point the same way. Only 8% of U.S. subscribers to the most popular assistant also pay for its closest work-focused rival, and a16z reports that 13% of people who pay for one AI tool pay for a second. Very few people split their attention across several general assistants.
That supports a plain decision for most teams: choose one sanctioned assistant, write down what it is for, and spend the training time on it. Keep the instructions, templates and examples your team builds in your own shared folder, not only inside the tool, so a later switch costs a week of adjustment instead of a fresh start. a16z's subscription charts show leadership between vendors moving month to month, which is another reason to own your materials.
Agents are real and still mostly technical.
The report spends several charts on personal agents, AI that acts for you instead of answering. a16z cites an early-user study of seven group chats and 1,502 messages in which coding and technical automation were the most discussed agent tasks. That suggests most of the people using agents today are technical.
For a school district or a clinic, that is a signal to watch and prepare, without rebuilding anything yet. The preparation is ordinary: know which systems hold your records, who can grant access to them, and which tasks you would trust to software that clicks buttons on its own. a16z's monetization chart is also worth a look before any purchase. Of 44 AI-native web products it reviewed, 84% charge subscriptions and 64% charge for usage or credits, many of them both, so budgets should assume a monthly fee plus a meter that rises as use grows.
Five questions to take to your next staff meeting.
Reading the charts against your own team
- Roughly how many of our people use an AI tool daily, and how many tried it once and stopped? A show of hands is enough to start.
- Which single task, if a tool did it well, would someone here gladly pay for out of their own budget?
- Who are our heavy users, what have they built, and has anyone else seen it?
- Have we named one assistant as the default, and do the templates people make with it live somewhere we control?
- When a tool starts acting on our systems instead of answering questions, which tasks would we allow first, and who would approve the access?
Write the answers on one page and date it. When the next round of charts comes out, compare your page with theirs before you compare yourself with anyone else.
