AI has reached nearly every worker, and the hours and the rules have not
AI use at work is close to universal and state legislatures have begun writing the disclosure and human-review rules, while employers fund AI training without giving staff the work hours to use it.
Coverage window: September 18–24, 2026
Most of the people doing the work now use AI to do it, and few of them have been told what rules apply. In the UK, 86% of employees say they use or experiment with AI at work, and 57% say it has raised how much they are expected to get through in a day. Those figures come from Owl Labs' tenth annual State of Hybrid Work report, published on 23 September and drawn from a survey of 2,014 full-time UK employees, part of a global sample of more than 8,000. 1
The rule-writing is happening outside the companies using the tools. MultiState, a firm that tracks state legislation, published its review of the 2026 sessions on 21 September and counted 155 bills on AI in the workplace, covering employee monitoring, automated hiring, and bias assessments. Colorado and Connecticut have now enacted laws that require employers to tell people when automated tools are used to decide things about them. 2
Employers more than doubled the AI training on offer in a year, and employees report no work hours in which to take it. The largest employer-funded training commitment of the week arrived on the same day as that survey, with a $70 million price tag. Official August data, published on 18 September, showed payrolls essentially flat across the country.
| Signal | What changed this week | What to watch |
|---|---|---|
| AI in the workday | 86% of UK employees use or experiment with AI at work; 27% use it daily and 26% several times a week. Managers are ahead of everyone else at 92%, against 70% of individual contributors. 1 | Whether the expectation gap closes: 57% say AI has raised how much they are expected to finish in a day, and 34% say it has made them less productive. |
| Unapproved tooling | 26% of AI users work with tools their employer has not approved, rising to 32% among Gen Z. 1 | Whether an approved route exists before the next client document goes into a personal account. |
| Monitoring and disclosure | 59% believe their employer monitors them, most commonly through login and logout times (32%), meetings attended (22%), and keyboard activity (15%). 82% say employers should be legally required to disclose it. 1 | Whether disclosure arrives as a legal duty before it arrives as a policy. |
| State law | 155 bills on AI in the workplace were considered in the 2026 sessions. Colorado and Connecticut enacted notice and explanation duties for automated decisions; California sent the governor a ban on surveillance tools that read neural data or infer emotional states. 2 | Whether bias-assessment requirements survive committee in 2027, and how California's governor acts on the surveillance bill. |
| Employer-funded training | The share of companies offering AI-specific training rose from 24.8% to 58.3% in a year, while 56.4% of employees say no work hours are set aside to build those skills. Verizon committed $70 million to free AI training. 34 | Whether training bought becomes capability, and whether verified skills start counting in promotion decisions. |
| Measured labor market | August payroll employment was essentially unchanged in 46 states and the District of Columbia; the national unemployment rate held at 4.1%. 5 | Whether the fall hiring season reaches the state data before the year ends. |
Almost everyone uses AI at work. The rules are being written in state capitals.
Owl Labs makes hybrid-meeting hardware, and its report is the company's own annual survey, now in its tenth year. The workplace it describes has settled into a rhythm: 60% of UK employees would choose a hybrid arrangement if they had the choice, and 31% say their employer changed its remote or hybrid policy in the past year, up from 24% in 2025. 1
AI has moved faster than the policies around it. Among the 86% who use the tools, 68% say AI has made them more productive and 60% say AI skill improves their chances of promotion. The costs sit beside the gains. 39% worry AI will make their role obsolete, a figure that rises to 43% among Gen Z, and 35% say their work-related stress has increased since last year. 37% describe themselves as disengaged, with burnout the leading reason they give. 1
Monitoring is the part employees can see and cannot verify. 59% believe their employer uses software to track them, and 82% say employers should be legally required to disclose when they do — a higher share than support almost any other measure in the report. 1
MultiState's review shows the disclosure question moving from preference to statute. Colorado's SB 189 requires the deployers of a covered automated decision-making tool to notify applicants and to provide a detailed explanation after an adverse outcome; it also lets applicants correct inaccurate personal data and request a meaningful human review. Connecticut's SB 5 requires employers to notify employees and applicants when automated decision tools are used, explain adverse decisions, and offer a chance to correct the data behind them. 2
The broader wave is thinner than its bill count suggests, and that matters to anyone planning against it. Proposals to require bias or impact assessments were introduced in North Carolina, Michigan, New Jersey, New York, Massachusetts, and Vermont, and every one died in committee. New York's bill to make large employers report annually on AI's effect on their hiring passed both chambers and was not signed into law. A New York proposal to levy a 2% corporate income surcharge where AI caused terminations, and a New Jersey plan for tax credits to employers who retrain displaced staff, both failed to advance. California's AB 1883, which would bar workplace surveillance tools that collect neural data or infer an employee's emotional state, is the one still live, sitting with Governor Gavin Newsom. 2
The review covers the 2026 sessions as a whole, and its own state map is current as of 1 September. Two enacted state laws and one bill awaiting a signature are the concrete obligations so far. The direction of travel is clearer than the current requirements.
The training budget arrived, and the hours did not
Workera sells skills assessment, and it surveyed 1,000 full-time salaried employees at US organizations with 5,000 or more staff in July 2026, comparing the answers with a March 2025 survey of 800 similar professionals. It published the comparison on 23 September. 3
The adoption numbers doubled in twelve months. Employees using AI tools several times a week rose from 39.9% to 67.8%. Companies offering AI-specific skills training rose from 24.8% to 58.3%. 6
The conditions around that training stayed where they were. 56.4% of employees say no time is allocated during work hours to develop AI skills. 84.3% spend five hours a week or less on development. 42.5% say they have been passed over, held back, or mis-rated because someone misjudged what they could do. Only 26.6% get regular coaching from a manager, and 21.5% get none at all. 3
Loading chart…
Both bars are shares of the group named on the axis: the employee figure is a share of the 1,000 surveyed employees, the training figure a share of the employers they work for. The two moved sharply between 2025 and 2026 while the time available to use either stayed where it was. 3
Two further answers in the same survey explain why the spending has not yet bought much. 46.5% of employees are building skills on tools their employer never provided, which keeps that learning invisible to the business. And when asked who judges their skills more fairly, 69.1% name a human manager while 14.7% name an AI tool — employees trust the assessment they already have, and 41.1% say they would accept continuous measurement if they controlled the data. 3
Employers are adding supply regardless. Verizon announced Verizon AI Skills for America on 23 September: $70 million in total, made up of $50 million in new funding and an existing $20 million reskilling fund for departing workers. The program aggregates free training from IBM, Google, Microsoft, Anthropic, Coursera, and OpenAI, and pairs it with local coaching through the nonprofits LISC, NACCE, and Goodwill Industries International. Verizon puts the value of comparable premium training at more than $700 per person per year. 47
The offer is aimed outward, at job seekers, early-career workers, displaced employees, educators, and small businesses. The gap Workera measured sits inward, at employers who have bought the courses and not the hours in which their own staff can finish them.
The measured labor market stayed flat in August
The Bureau of Labor Statistics published state employment and unemployment data for August on 18 September, and the picture it draws resists the anecdote. Nonfarm payroll employment rose in four states and was essentially unchanged in 46 states and the District of Columbia. The gains were California at 39,400, Wisconsin at 11,800, South Carolina at 10,800, and New Mexico at 5,500. 5
Over the year, payroll employment rose in eight states, fell only in the District of Columbia, and was essentially unchanged in 42. The largest annual gains were Texas at 159,400, California at 138,500, and North Carolina at 65,600; the District lost 27,000, or 3.6%. 5
The national unemployment rate held at 4.1%, unchanged over the month and little changed from a year earlier. Unemployment rates fell in eight states and the District and held steady in 42. South Dakota recorded the lowest rate at 2.0%, Ohio's 3.3% set a new series low, and the District of Columbia recorded the highest at 5.7%. 5
Read as a national number, the labor market shows no break that would correspond to the tooling arriving in nearly every job. State and metro figures are where the differences live, and the metro release for August follows on 30 September.
What to do next
- Inventory the automated decisions and the monitoring tools, then write the notice. The people-ops or legal owner should list every system that screens, ranks, schedules, or evaluates employees, alongside every system that logs activity, and draft a plain-language notice covering each one. Colorado's and Connecticut's laws already require that notice and an explanation after an adverse decision. 2
- Put protected hours on the calendar. The learning or operations lead should assign time inside the working week for the AI training already licensed, since 56.4% of employees report no work hours set aside for it and 84.3% spend five hours a week or less on development. A course nobody has time to open shows up as a completed purchase, not a capability. 3
- Make demonstrated skill count in assignment and promotion decisions. The HR lead should require evidence of capability when roles are filled internally. 42.5% of employees say they have already been mis-rated, and 46.5% are building skills on tools the company never supplied, which means a large share of what people can do is currently invisible. 3
- State at rollout whether the expectation changes. The manager introducing an AI tool should say plainly whether the same output is now expected in less time, or more output in the same time. 57% of employees say AI has already raised daily expectations, and 34% say it has made them less productive. 1
The week's five readings point to one shared constraint: the tools have reached almost every worker, while the time to learn them, the notice about how people are being measured, and the statutes governing all of it are still catching up. Of those three, only the statutes are arriving on someone else's schedule.
References
- 1State of Hybrid Work 2026: UK report
owllabs.co.uk
- 2
- 32026 State of Skills Intelligence
workera.ai
- 4
- 5
- 6
- 7AI Skills for America
verizon.com
This story was produced automatically by a channel. One sentence is all it takes for Neodrop to keep producing for you.