AI adoption challenges: six common barriers and what helps

Your team has access to AI, but making it useful still takes work. Some employees can’t see where it fits. Others have tried it and come away with more questions about accuracy, information handling, or whether the effort is worthwhile.

Start by identifying what’s getting in your team’s way. Practical training can help employees build AI skills and judgment, but they’ll also need clear rules, support from leaders, and a workable process. The six challenges below can help you work out where your team needs help and what to do next.

Common AI adoption challenges and what helps

Training can help employees create and review AI-assisted work. Your organization still needs to decide what information they can use and address problems connecting AI with existing systems. Teams may face several of these challenges at once, so plan for both the learning and the changes needed to put it into practice.

Employees can't see where to apply AI in their day-to-day

An impressive demonstration can still leave someone wondering what they’d use AI for on Monday. If that sounds familiar, look for training that helps employees connect the tool’s capabilities to a task they already understand. They need a reason to try it and enough guidance to make a useful attempt.

This difficulty shows up in the research. A UK government survey of 3,500 businesses with at least five employees, conducted in H125, found that the lack of an identified need and limited skills or expertise were the most commonly cited barriers to adoption. Separately, the OECD’s late-2024 survey across seven countries found that 57.3% of SME non-users considered generative AI unsuited to their work. That’s a perception worth exploring: employees may need help finding a suitable application, or their proposed task may genuinely be a poor fit.

Consider a hypothetical project team learning to draft an internal status update from approved, non-sensitive source notes. Useful training would connect the exercise to the work the team needs to produce:

• Clarify who will read the update and what they need to know.
• Explain how that purpose and the source notes shape the instructions given to AI.
• Let employees try the task and receive feedback on the draft.

When assessing a training option, look for that connection between explanation, practice, and feedback. The intended result is an employee who can identify a suitable task, attempt it, and explain why AI might help.

Ziplines helps business teams build practical AI skills through expert instruction and hands-on application. Contact us to discuss training for your team.

Familiar routines feel easier to rely on

An employee’s current method has one advantage: they know how to get the work done. Learning an alternative takes time, and the first attempts may feel awkward. Give people a clear reason to invest that effort, with room to practice and someone to ask for help.

Gallup’s February 2026 survey of 23,717 US employees found that non-users and infrequent users in organizations offering AI cited preferences for established methods and concerns about usefulness. Among employees with access to AI, manager support was associated with more frequent use. Those findings give you useful questions to explore with your team: what makes the current approach dependable, and what would make trying an alternative worth their time?

The training itself can provide demonstrations, guided practice, and feedback. Your organization needs to make that learning commitment workable.

Agree with managers on:

• The work problem employees will explore and why it matters.
• Time to practice alongside their existing responsibilities.
• A contact for questions during and after learning.
• Opportunities for experienced colleagues to assess the approach and contribute their judgment.

Access to a course leaves the question of time unanswered unless managers help resolve it. Employees shouldn’t have to guess whether learning is a priority while their normal workload stays fully booked.

Keep the conversation open to questions about fit. Someone who can explain why the familiar method works well is contributing useful knowledge. Use that comparison to decide whether the team needs more practice, a different application, or a reason to continue with the current approach.

Employees aren't sure if they can trust the output

Even a well-written response can be wrong. If employees are hesitant to use AI because they don’t trust the output, give them practice checking it and deciding whether it’s reliable. That helps them recognize when a result is useful and when it needs correction.

The UK government study identified output accuracy as a deployment challenge. The BCG experiment also illustrates why success on one task doesn’t settle whether AI is suitable for another: on a task outside the model’s tested capabilities, the two GPT-4 groups combined were about 19 percentage points less likely than the no-AI group to produce correct solutions.

For the hypothetical project team, preparing the status update would include checking it against the approved source notes. Look for learning that helps employees answer questions such as:

• Do the statements match the source notes?
• Is confirmed progress clearly distinguished from unresolved questions?
• Can the reviewer explain why the update is ready to share?

Feedback on those judgments is part of the learning. An instructor should be able to discuss why a draft needs correction, when a different approach would help, and when the output should be rejected. That gives you more to assess in a training proposal than a demonstration of how quickly the tool writes.

The person accountable for the work still decides whether it’s ready to use. Training can help employees develop a reliable review habit; it can’t guarantee error-free output or make every application appropriate.

Employees don’t know what information they’re allowed to use

An employee may understand the task perfectly well and still be unsure what material they’re allowed to put into an AI tool. Clear company guidance gives them a basis for deciding, and training helps them apply it to situations they recognize.

Among SME non-users, the OECD found concerns about copyright, legal or regulatory issues, and what happens to information fed into models. Within organizations that offer AI, Gallup found more frequent use among employees who reported having clear AI policies. Clearer rules can help employees understand what they’re allowed to do. But that finding doesn’t show whether those organizations are protecting information or meeting their legal obligations.

Before expecting employees to work independently, make sure the responsible people in your organization have answered the questions training will depend on:

• Which tools are approved for this kind of work?
• What information is permitted, and what is restricted?
• Who can answer a question about an uncertain activity?
• Who keeps the guidance current?

With those decisions in place, a trainer can use approved, non-sensitive examples to teach employees how to recognize the relevant rule. If an activity remains uncertain, pause it until the responsible policy, privacy, or security contact has answered the question.

When choosing training, ask how your team will practice applying the guidance. Employees should come away knowing how to identify what’s permitted and where to get an answer when it’s unclear. The organization sets the boundaries; the instructor helps people understand how to work within them.

AI adds friction to an already difficult workflow

If using AI means hunting for records, comparing conflicting versions, and copying information between disconnected systems, employees have a fair reason to question the effort. Your goal is to find out where that extra work comes from. Some difficulties need better instruction; others need the people responsible for the process or systems to make changes.

Gallup found more frequent AI use among employees who reported that it integrated well with their existing systems and processes in organizations that offer AI. That association makes fit worth examining, though it doesn’t guarantee a productivity gain. Consider where AI would enter the task and what employees need before they can use it responsibly.

In the hypothetical status-update example, the team needs approved notes it can find and rely on. If nobody knows which version is current, another lesson on prompting won’t settle the question. Bring the issue to the people who can resolve it:

• The information owner can establish who maintains the source notes and where they belong.
• The process owner can clarify who provides inputs, reviews the draft, and authorizes sharing.
• The relevant technical teams can assess access or integration needs.

Those responsibilities may sit with the same person in a smaller organization. What matters is that the necessary decisions get made, so employees have usable inputs and a clear process to follow.

Training can then help the team practice that process, including its review responsibilities. Keep listening to employees as they apply it. When someone says the new approach adds work, ask which part: the answer helps you decide whether to provide more learning support or fix a problem in the way the work is organized.

Leaders can’t tell if AI is worth the investment

You may know how many employees completed training and how often they use AI, yet still struggle to explain what improved. That’s a separate question worth answering. Look at the work employees produce, the effort involved, and the cost of continuing the application.

McKinsey’s survey published in August 2026 found that 37% of respondents reported a positive AI contribution to their organization’s earnings before interest and taxes, or EBIT. About one in five said operating costs constrained AI use. The survey collected 1,719 responses across 97 nations in May–June 2026; those self-reported organizational measures describe something different from an employee’s experience of saving time.

For your team, compare a chosen task with how the work was done before AI was introduced. Training can help employees produce work for review and explain the choices they made. Using the hypothetical status update, you might assess:

• Whether the finished update meets the team’s quality expectations.
• How much correction it needs before sharing.
• The time spent preparing inputs, drafting, and reviewing.
• The operating costs of continuing to use AI for the task.

Choose measures that reflect what your business needs from the work. A draft produced quickly may still take substantial time to prepare or correct, so include that effort in the comparison. Equally, a useful improvement needs to be explained in terms the business can assess, rather than inferred from course completion or tool activity.

For a step-by-step approach, see our guide to measuring AI training ROI and business impact.

That evidence gives you a basis for deciding what comes next: further learning, a process change, or ending an application that isn’t earning its place. Employees can demonstrate what they’ve learned and how they apply it; leaders decide whether the resulting work justifies the investment.

You don’t need the same approach to every AI adoption challenge. Once you know what’s getting in the way, it’s easier to choose the next step. Relevant training, supported practice, and the organizational fixes that let people use their skills give your team a credible way forward, with progress you can see in their work.

Ready to help your team put AI to work? Ziplines offers practical AI training for enterprises and small and medium-sized businesses. Contact us to discuss your goals and training needs.

Frequently asked questions

Does every employee need the same AI training?

No, employees may benefit from shared fundamentals while needing different practice for their roles. Choose the level and applications around their tasks and responsibilities, including the judgment they’ll need to exercise. The technical depth can vary with the work.

How long does it take to overcome AI adoption challenges?

There’s no universal timeline established by the research discussed here. Start with the barrier and define useful progress, such as employees being able to explain and evaluate an appropriate application. A course’s duration alone won’t tell you how quickly your organization will adopt AI.

Can a team make progress without using AI in every process?

Yes. Assessing a task and deciding the existing approach remains appropriate can be a sound outcome. Low AI use alone doesn’t tell you whether employees have made that judgment or are still facing a barrier that needs attention.

‍