Stanford's 2026 AI Index Is a Mountain of Charts. Here Are the Five That Matter If You Run a Small Firm.
88% of organizations use AI. Agents are in single digits. Gains are biggest where work is easy to measure. A plain-English read of the Economy chapter for owners.
Every spring Stanford's Institute for Human-Centered AI publishes the AI Index, the reference report on where AI stands. Most of it is about models, investment and research. The Economy chapter is the part an owner should read. Here's what it says, and what to do with it.
1. Almost everyone uses AI now
Organizational AI adoption rose to 88% of surveyed organizations in 2025, and generative AI is used in at least one business function at 70% of organizations. Using AI is no longer an advantage on its own. Using it well on the right tasks is.
2. AI agents are still rare
Despite the noise, AI agent deployment was in the single digits across nearly all business functions. If you feel behind on agents, you aren't. Plain automation and assistants are where most firms really are. Our do you need an AI agent? test helps sort it.
3. Gains are biggest where work is measurable
The report says productivity gains are largest in structured, measurable work where outputs are easy to monitor. It cites studies reporting gains of 14% to 15% in customer support, 26% in software development and 50% in marketing output. Gains are smaller in tasks requiring deeper reasoning.
The lesson: start where you can count the output, such as tickets closed, quotes sent or posts published. If you can't measure the task, you can't tell whether AI helped.
4. The job effects are in hiring, not mass layoffs
AI's labor market effects are showing up unevenly, concentrated in hiring pipelines and the youngest workers in exposed occupations. Employment for software developers aged 22 to 25 fell nearly 20% from 2024. One-third of organizations expect AI to reduce their workforce in the coming year, though almost half expect little or no change, and large-scale job losses haven't shown up in overall employment data.
For a small firm, the practical question is junior roles: who learns the work if AI does the entry-level part? The report also flags evidence that heavy AI reliance may carry long-term learning penalties that slow skill development.
5. AI is getting more valuable and cheap to try
Estimated U.S. consumer surplus from generative AI reached $172 billion a year by early 2026, up from $112 billion a year earlier, and the report notes most of these tools remain free or close to it. Trying AI costs little. The cost is in the time spent using it badly.
What to do with this
- Stop asking whether to use AI. Ask which measurable task to start with. Our AI use case scorer ranks candidates.
- Measure output, not usage. The firms in the studies above counted finished work.
- Protect how people learn. Keep juniors doing enough of the real work to build judgment.
- Ignore agent FOMO. Single-digit deployment means you have time to get the basics right.
For why individual AI gains often don't reach the bottom line, read the workplace AI productivity stall.
How to use the report without drowning in it
Treat the AI Index as a map of where evidence is strongest, not as a shopping list. Customer support, software and marketing show clearer measured gains because the work produces countable outputs and errors are comparatively visible. Work that depends on ambiguous judgment, relationships or rare exceptions is harder to evaluate, so a vendor's average productivity claim tells you very little about your firm.
A practical starting scorecard has only five lines: the task, the monthly volume, the current minutes per accepted result, the error or rework rate, and the person accountable for quality. Run the AI version beside the current process, then compare accepted results — not logins, prompts or drafts. If the task cannot be counted, begin with a narrower slice that can, such as summarizing support tickets before deciding whether an agent should answer customers.
The hiring finding deserves the same discipline. Saving senior people time by removing every routine task can also remove the ladder junior employees use to learn. Keep some real work, review and explanation in junior roles, and measure whether new staff are becoming able to judge the AI rather than merely operate it.
That last point matters because today's shortcut can become tomorrow's supervision problem when nobody has learned the underlying work well enough.
Frequently asked questions
What is the Stanford AI Index?
An annual report from Stanford's Institute for Human-Centered AI that collects data on AI research, technical performance, the economy, policy and public opinion.
How many organizations use AI according to the 2026 AI Index?
The Economy chapter reports organizational AI adoption reached 88% of surveyed organizations in 2025, with generative AI used in at least one business function at 70%.
Where does AI improve productivity most?
The report says gains are largest in structured, measurable work, citing studies with gains of 14% to 15% in customer support, 26% in software development and 50% in marketing output.
