The most common mistake we see before AI training is skipping the question of what the team needs. A company books a generic session, the sales team finds it too basic, the finance team finds it irrelevant, and two people who were already using AI on personal accounts sit quietly at the back. A needs assessment takes a week or two of light effort and prevents all three.
This guide gives you a checklist table you can copy into a spreadsheet, a set of survey questions you can paste into Microsoft Forms or Google Forms, and a simple way to score the tasks that are worth training on. It is written for a company of roughly 10 to 200 people with no dedicated learning and development team.
What is an AI training needs assessment?
An AI training needs assessment is a structured look at the gap between how your people use AI today and how the business needs them to use it, done before you design or buy training. It answers four questions: what work could AI help with, who does that work, what are they allowed to put into AI, and what do they already know.
Training needs analysis traditionally works at three levels: the organization (goals, rules, tools), the task (what the work requires), and the person (who needs what). The same structure works for AI, with one addition: an inventory of the AI tools already in use, because in most companies the answer is "more than leadership thinks."
Why run a needs assessment before AI training?
Because AI use is already happening unevenly, and training that ignores that wastes the time of both the beginners and the experienced users. A needs assessment lets you group people sensibly and build exercises around real work.
Microsoft's 2024 Work Trend Index found that 75% of knowledge workers were using AI at work, 78% of those users were bringing their own AI tools, and only 39% had received AI training from their company. A small company is unlikely to be different. The assessment surfaces the personal accounts, the tasks people have already moved to AI, and the ones they are nervous about, which is exactly what a trainer needs to know. If personal accounts turn up, our guide to shadow IT and shadow AI explains how to handle it without turning it into a crackdown.
How do you run an AI training needs assessment?
Run it in seven steps over one to two weeks. Most of the effort sits with one internal owner and the team leads; employees spend about ten minutes on the survey.
- Name an owner and a goal. One person runs the assessment. Write one sentence on why the company wants AI training now: a new tool rollout, uneven use, a policy, a client asking about it.
- Inventory tools and accounts. Which AI tools does the company pay for, on which plans, and who has a seat? Which free or personal tools are people using? Include AI features built into software you already own.
- Classify your data. Agree on three tiers: fine to use in an approved business AI tool, only with care or redaction, never. Client confidential information, health information, card numbers and passwords usually sit in the last tier.
- Send the survey. Use the questions below. Make it anonymous or at least clear that answers about current AI use will not be held against anyone.
- Interview team leads. Twenty to thirty minutes each. Ask for the five tasks that eat their team's week and where errors would be costly.
- Build and score a task list. Combine survey and interview results into one list of candidate tasks per role and score them with the matrix further down.
- Turn it into a plan and a baseline. Decide groups, formats and labs, and record starting confidence and task times so you can measure later.
AI training needs assessment checklist
Copy this table into a spreadsheet and tick each row off as you go. The "who answers" column keeps the work from landing on one person.
| Area | Question to answer | Who answers | Done |
|---|---|---|---|
| Goal | Why AI training, why now, and what would make it worth doing? | Owner or leadership | ☐ |
| Approved tools | Which AI tools and plans does the company pay for, and who has seats? | IT or office manager | ☐ |
| Unapproved tools | Which personal or free AI tools are people using for work? | Survey | ☐ |
| Built-in AI | Which existing software (CRM, email, accounting, design) has AI features switched on? | IT or system owners | ☐ |
| Data tiers | What may go into AI, what needs redaction, what never goes in? | Leadership with counsel if needed | ☐ |
| Policy | Is there a written AI acceptable-use policy? Does staff know it? | HR or owner | ☐ |
| Roles | Which roles and how many people in each? | HR | ☐ |
| Tasks | The five recurring tasks per role that take the most time | Team leads | ☐ |
| Risk | Which tasks involve regulated data, client advice or decisions about people? | Team leads and HR | ☐ |
| Current skill | Self-rated confidence, 0–10, and how often people use AI now | Survey | ☐ |
| Barriers | What stops people: rules, access, time, trust, not knowing where to start? | Survey | ☐ |
| Champions | Who already uses AI well and could help colleagues? | Survey and team leads | ☐ |
| Illinois rules | Is AI used, or planned, in hiring, promotion, discipline or other employment decisions? | HR | ☐ |
| Logistics | On-site or online, room, laptops, sign-in on work accounts, group sizes | Office manager | ☐ |
| Funding | Is ETIP reimbursement worth applying for? Pre-approval must come before training. | Finance | ☐ |
| Baseline | Starting confidence and times for a few named tasks, recorded | Owner of the assessment | ☐ |
What questions should an AI skills survey ask employees?
Ask about current use, tasks, confidence, data and barriers, in that order, and keep it under ten minutes. These are the questions we use, grouped so you can drop a section if it does not apply.
Current use
- How often do you use an AI tool for work? (Never, a few times a month, weekly, daily)
- Which AI tools have you used for work in the last month? (List the approved tools plus "other, please name")
- Do you use any AI tool for work on a personal account? (Yes, no, not sure)
- What have you used AI for so far? (Writing, summarizing, research, spreadsheets, brainstorming, other)
Tasks
- What are the three tasks that take the most of your week?
- Which task do you most wish took less time?
- Which tasks would you never want AI involved in, and why?
Confidence
- How confident are you using AI tools in your daily work? (0 means not at all, 10 means completely)
- How confident are you that you could tell when an AI answer is wrong? (0–10)
- Which statement fits best: I have not started; I have tried it but it did not help; I use it for a few things; I rely on it for parts of my job.
Data and rules
- Do you know what company information you may and may not put into an AI tool? (Yes, partly, no)
- Do you handle client confidential, health, financial or personnel information? (Yes, no)
Barriers and format
- What has stopped you from using AI more? (No access, unclear rules, no time, results not good enough, do not know where to start, other)
- How would you prefer to learn? (In-person workshop, live online, short videos, one-on-one help)
- Would you be willing to help colleagues once trained? (Yes, maybe, no)
- Anything else the trainer should know?
Question 8 is the one to keep word for word. It becomes your baseline, and asking it again after training is the simplest learning measure there is. Our guide to measuring AI training ROI shows how to use it, and our published results show what it looked like in our two-day program: attendees who started under 6 finished at an average of 8.2, up from 4.7.
How do you score and prioritize AI use cases?
Score each candidate task on four factors and train on the ones with high frequency and time, low risk and data you are allowed to use. A simple 1-to-3 scale is enough.
| Factor | 1 | 2 | 3 |
|---|---|---|---|
| Frequency | Monthly or rarer | Weekly | Daily |
| Time per instance | Under 15 minutes | 15 to 60 minutes | Over an hour |
| Risk if wrong (score reversed) | Client, legal or financial harm | Embarrassing but fixable | Internal only, easy to check |
| Data fit (score reversed) | Needs "never" tier data | Needs redaction | Fine for an approved tool |
Add the four scores. Tasks scoring 10 to 12 are good lab material: frequent, slow, safe to practice and easy to check. Tasks scoring 4 to 6 belong in the "not yet" or "not at all" column, which is itself useful to put in the policy. Typical high scorers in our experience include meeting notes to action items, first drafts of routine client emails, turning rough notes into an SOP, and cleaning up a weekly spreadsheet report.
How do you turn the results into a training plan?
Group people by role for the applied work and, where numbers allow, by starting confidence for the foundations. Then choose the format that matches the gap.
- Mostly beginners, mixed roles: a 2–3 hour AI workshop per group of up to about 15 people, covering foundations, privacy and two or three shared tasks.
- Clear role-specific tasks: a full day of foundations followed by a second day with three labs per role, built on the tasks you scored highest, using practice or fictional data.
- Whole company, several departments: cohorts over one to four weeks with one policy and one shared prompt library; see corporate AI training in Chicago.
- Owners and leaders: one-on-one coaching on their own reporting and correspondence.
Use the champions from question 15 as helpers in the room and owners of saved instructions afterward. And check the curriculum against our buyer's checklist for what AI training should include before you sign anything.
What compliance issues should the assessment flag?
The assessment should flag any planned AI use that touches decisions about employees, biometric data, or clinical mental-health work. Each brings Illinois-specific rules.
- Employment decisions. Since January 1, 2026, the Illinois Human Rights Act, as amended by HB 3773, prohibits using AI in recruitment, hiring, promotion, discipline and similar decisions in a way that discriminates, and requires notice to employees when AI is used for those purposes. The Illinois AI employment notice kit covers what to do.
- Biometrics. Any tool that captures voiceprints or face scans raises the Biometric Information Privacy Act; see our BIPA compliance guide.
- Therapy and behavioral health. The Wellness and Oversight for Psychological Resources Act (HB 1806), signed in August 2025, prohibits using AI to provide therapeutic decision-making while allowing administrative and supplementary support for licensed professionals.
Record the answers in your AI acceptable use policy. This is general information, not legal advice; confirm with your counsel.
Can 312 run the needs assessment for us in Chicago?
Yes. A free first call covers your goals, team and tools; then we talk to your team leads, collect the tasks that eat their week, and build the labs around them. We train on-site across Chicago and the suburbs and live online, on your own ChatGPT, Copilot, Claude or Gemini accounts. A 2–3 hour workshop for up to about 15 people typically starts from about $2,500; longer programs are quoted in writing. Our AI training cost guide explains what moves the price, and funding AI training in Illinois covers ETIP, which requires pre-approval before training starts.