OptiMax Partners
← All articles

No AI Policy? Your Employees Are Already Writing One

2026-09-03

More than half of companies (55%) still have no official AI policy. That vacuum is not staying empty. According to CNBC and SurveyMonkey's Q3 2026 AI and Jobs Survey, one in four employees who use AI already redirect colleagues to "ask AI before asking me." Teams are writing the rules before their companies do.

What Is Happening While Companies Stay Silent?

In the survey, 55% of workers said their employer has no official AI policy. Another 34% said use is left to individual judgment, while only 6% work where AI is required and 5% where it is banned outright. Nine out of ten workplaces have effectively set no direction at all and are relying on employees to figure it out.

The trouble is that informal rules are settling in fast. Among AI users, 23% (34% of daily users) have redirected a colleague to ask AI before offering their own help. Listen to the reasons and there is no malice in it: they want colleagues to become more self-sufficient (50%), they want to protect their focus time (44%), or AI's answer was simply good enough (34%). For 12%, asking AI first has already become their team's standard practice.

It is a practice born of good intentions, but it reads differently once it spreads with no shared standard for accuracy checking or confidentiality. A policy vacuum does not stay disorderly; it gets filled by an order the company has never reviewed.

How Much Are Employees Already Using AI?

More deeply than most leaders assume. 29% of workers use AI daily (21% multiple times a day, 8% once a day). Another 34% use it occasionally, and 37% never use it at work. Taken together, two in three workers have already brought AI into their jobs.

What separates daily users from everyone else is depth, not just frequency. Research and information gathering is the most common use for everyone (56% overall), but daily users range much wider:

Use case Daily users Weekly users
Researching and gathering information 64% 57%
Brainstorming and generating ideas 51% 39%
Writing emails and messages 49% 33%
Creating documents and reports 47% 26%
Analyzing data 44% 25%
Summarizing long documents 41% 24%
Planning and organizing tasks 37% 21%

The harder the task, the closer the gap gets to doubling. For daily users, AI is no longer a search box; it is a working tool that fits the hand. And while companies postpone policy, that depth of use keeps growing.

Does Anxiety Change With How Much You Use AI?

Yes, and this is the survey's most striking finding. Daily users are the only group whose positive feelings about their own job security outweigh the negative: 27% feel more secure because of AI versus 22% who feel less secure. In every other group, the relationship flips.

Net job-security sentiment due to AI (% more secure minus % less secure) +5 Daily users -15 Weekly -21 Occasional -6 Non-users Daily: 27% vs 22%. Weekly: 10% vs 25%. Occasional: 3% vs 24%. Never: 4% vs 10%.

Sentiment about the job market overall follows the same line. Daily users split almost evenly (35% optimistic, 33% pessimistic), weekly users lean negative by 19 points, and occasional users hold the darkest view of all: 7% optimistic versus 52% pessimistic.

One insight matters most here. The people most anxious about AI are not the ones who know nothing about it; they are the ones who know enough to feel the threat but have not used it enough to feel capable. This is exactly why policy and training matter: unsupported, halfway exposure is what feeds an organization's anxiety.

Should You Worry About Overreliance, Then?

The data says not yet. Only 6% of AI users go to AI first when they hit a problem. A clear majority (58%) try to work it out themselves before asking AI, and 30% do not use AI for problem-solving at all.

The felt benefits, meanwhile, are unmistakable. 68% of users say AI has saved them time (84% of daily users), 56% feel more confident in their skills (73% of daily users), and 45% say their critical thinking has actually improved (60% of daily users). For employees who use it well, AI is working as a tool that builds capability rather than erodes it.

How Far Should Junior Employees Be Allowed to Use AI?

This is where opinion splits hardest. 42% believe junior workers should use AI with clear guidelines, 14% only under supervision and approval, and 9% freely without restrictions. Strikingly, 35% believe junior workers should be prohibited from using AI at all.

The question underneath is a single one: is AI a shortcut that builds a junior employee's capability, or a detour around the reasoning practice that builds it? The labor market itself has not settled the answer, which means every company needs to set its own standard. And as we saw above, where there is no standard, informal practice fills the space instead.

How Is AI Reshaping Employees' Career Plans?

Broadly and deeply, already. 32% of workers have changed or reconsidered which skills to develop because of AI (45% among students), 24% have rethought which jobs or companies to apply to, and 23% have changed or considered changing their target industry. One in five workers (20%) at least sometimes considers moving into a trade or hands-on profession as a more AI-resistant path.

The direction of travel is clear all the same. Asked which skills will last, workers put critical thinking, creativity, and communication at the top. On education the verdict was sober: 33% say AI has made a college degree less worth it, well ahead of the 18% who say it is now worth more.

What Belongs in an AI Policy?

There is no shortage of writing about AI governance. Most of it stops at principles and never tells you what to actually put in the document when you open it on Monday morning. To close the gap this survey exposes, an AI policy needs at minimum the following eight components.

Component The question the policy must answer What fills the gap if you leave it blank
1. Purpose and scope Who does this policy cover, and for which tools and tasks Every department invents its own rules
2. Approved tools and accounts Which tools are company-approved, and are personal accounts allowed Shadow AI spreads outside your control
3. Data input rules What can never be entered: client personal data, trade secrets, employee records Confidentiality breaches and privacy-law exposure
4. Output verification and accountability Who owns the final output, and what review does it pass through Unchecked errors flow straight to clients
5. Use in hiring and HR decisions Is AI used to screen, assess, or promote, and are candidates told As shown below, this can already be a legal violation
6. Transparency and disclosure Do you tell clients when AI contributed to a deliverable Damaged trust and contract disputes
7. Training and proficiency support How do you move halfway users to fluency The anxiety and pessimism this survey documents
8. Policy owner and review cycle Who owns the document, and when is it revisited A file nobody opens six months later

The fifth component is no longer optional. In Ontario, as of January 1, 2026, employers with 25 or more employees must state in any publicly advertised job posting whether AI is used to screen, assess, or select applicants, and must retain postings and application forms for three years after takedown. New York City's Local Law 144 requires an annual independent bias audit of automated employment decision tools used in hiring or promotion, a public summary of the results, and notice to candidates at least 10 business days before use. Whether you hire in Canada or the U.S., using AI in recruitment without a policy and disclosure is already a compliance gap.

You do not need to launch with a thick document covering all eight. What matters is taking a deliberate position on each component at least once. Any component you leave blank gets filled, as we saw above, by your employees' informal rules.

So Where Should You Start?

  • Write the policy down before informal rules harden. It does not need to be long: start with three of the eight components above (scope, data input rules, and output verification), and "ask AI before asking me" gains its guardrails.
  • Remember that the most anxious employees are the halfway users. Shallow exposure without support is worse than none. Aim training not at introducing the tool but at getting people past the awkward middle stage.
  • Choose your junior-talent stance deliberately. Guided use (the 42% plurality position) protects both skill development and productivity; an unexamined ban or a free-for-all protects neither.
  • Center development plans on critical thinking, creativity, and communication. Employees already believe these are the skills that last; align reviews, promotions, and training accordingly.

FAQ

What does having no official AI policy actually risk?

Informal rules fill the vacuum. Nearly one in four AI users already sends colleagues to AI first, yet there is no shared standard for accuracy checking, confidentiality, or when human judgment is required. Even a short written policy turns that improvisation into something you can train against and audit.

Why are occasional AI users the most pessimistic group?

Halfway exposure appears to be the cause. They have seen what AI can do, so they feel the threat, but they have not used it enough to make the tool their own. Daily users convert that same awareness into a sense of advantage; 27% of them feel more secure in their jobs because of AI.

Should companies ban junior employees from using AI?

35% of workers think so, but the plurality view (42%) favors use with clear guidelines. A guided approach lets junior staff build AI fluency, which in this survey travels together with confidence and security, while protecting the reasoning practice early-career growth depends on.

What components does an AI policy actually need?

Eight are the baseline: scope, approved tools and accounts, data input rules, output verification and accountability, rules for AI in hiring and HR decisions, client disclosure, training and proficiency support, and a policy owner with a review cycle. If you use AI in hiring, start with the legal requirements already in force, such as Ontario's job-posting disclosure duty and New York City's bias-audit law.

Are daily AI users actually more productive, or do they just feel that way?

The survey measures perception, but the size of that perception is substantial: 84% of daily users say AI saves them time, 73% feel more confident in their skills, and 60% say their critical thinking has improved. Confidence and a sense of security shape engagement and retention on their own, so perception is a real outcome.

AI policy is one piece of how your organization runs. If you are curious where your teamwork, leadership, or organization as a whole could grow, book a free consultation. We'll look at where you are today and point you to the changes that matter most. Book your free consultation

Disclaimer: This article summarizes publicly reported survey findings for general informational purposes. Figures reflect the survey's US worker sample as published and may not generalize to every workforce. Source: