How to adopt AI across a business: 15 questions answered
By the Day Seven team · Updated 3 November 2026
Short answer: AI adoption works when people are trained on their own repeated tasks, with clear rules on data, and when someone measures whether the work actually changed. Buying licences and running an awareness session is where most organisations stop, and it is the main reason use fades.
This guide answers the 15 questions leaders ask us most about AI adoption, each under its own heading with the short answer first. Where we quote research, it is from the UK government's employer guide, What works for AI upskilling in the UK, published with Skills England, and we link every source. Where we describe how Day Seven works, it is the method set out on our own pages.

The seven steps at a glance
Adopt AI in seven steps: check readiness, prioritise a few high-value tasks, run a short pilot, put light governance in place, train the people who will use it, measure the result against a baseline, then scale what worked. Most stalled programmes skip the middle steps: they buy licences and run an awareness session, but never pick specific tasks, set rules or measure.
- Start from tasks people repeat every week, not from a tool.
- Pilot with a small group and a fixed end date before buying at scale.
- Agree data rules and human checks before the pilot starts, not after.
- Train on the real tasks in the pilot, then measure time, quality and sustained use.
The 15 questions below go into each step in more detail.
How do you get employees ready to use AI at work?
Train them on the tasks they already do every week, in the tools they are allowed to use, with the data rules built into the exercises. Start from what they already do with AI informally, rather than assuming everyone is a beginner.
The government's employer guide puts it plainly: "Training should teach individuals how to work effectively with AI systems, not just how to access tools." It adds: "Training should also recognise and build on existing informal or self-taught AI use, rather than assuming a zero-starting point."
In practice that means three things before anyone is trained:
- Know what people can reach. Which approved tools, and which features, each person can actually sign in to.
- Agree the rules. What data may and may not go into each tool, and when a person must check or sign off the output.
- Pick real tasks. Two or three tasks per team that are repeated weekly, such as a weekly report, a type of customer email or a first-draft document.
Confidence matters as much as skill. People who are worried about getting it wrong, or about what AI means for their job, need a safe place to practise and ask questions. Our AI upskilling page sets out how we build this into a programme.
What is the best way to introduce AI into a business?
Start small: one team, two or three tasks, a short pilot with a fixed end date and a baseline recorded before training. Expand only what worked.
The seven steps above set out the order: readiness, a few high-value tasks, a short pilot, light governance, training, measurement against a baseline, then scaling what worked. A first pilot of four to eight weeks is typical.
The government guide makes the same point about pace: "Progress does not need to be fast. Small, practical steps can help build confidence, improve skills, and support steady progress."
The most common false start is choosing a tool before choosing the tasks. If people cannot name what they will use AI for on Monday morning, the tool will sit unused.
How do businesses move from AI experimentation to implementation?
By turning informal, individual use into agreed ways of working: named tasks, shared prompts, clear rules and a person who owns the decision to expand. Implementation is less about new technology and more about routine.
The government guide describes where most organisations are: "Many are experimenting with AI tools, but few have embedded them into everyday work or developed structured approaches to workforce capability." It also notes that linking AI training to existing frameworks "shifts AI adoption from informal experimentation to consistent, governed practice."
Steps that help:
- Collect what people already do. A short, blame-free survey of informal AI use usually shows where the value is.
- Standardise the best of it. Turn the prompts and approaches that work into a shared prompt library for each team.
- Write it into the process. If a task now starts with an AI first draft, say so in the process notes, along with the checking step.
- Give it an owner. One person with authority to approve tools and expand use, supported by IT and data protection.
- Measure it. Compare time, quality and continued use against the baseline, so the decision to expand rests on evidence.
How can a business create an AI roadmap?
Work out where you are now, decide which skills each role needs, then plan a sequence of small pilots that build towards wider use. A roadmap is a list of tasks, people and checkpoints, not a list of tools.
The government's tools give a useful structure. The employer guide suggests you "Use the AI Skills Adoption Pathway to understand your current stage of AI use". The pathway has nine stages: awareness, exploration, assessment, experimentation, reflection, upskilling, integration, strategy and scaling. According to the same guide's survey, "AI adoption is still concentrated in early stages, particularly awareness (21%) and exploration (19%), with far fewer organisations reaching more advanced stages such as integration (7%), strategy (4%), or scaling (1%)."
A simple roadmap has four parts:
- Starting point: your current stage, and where people already use AI.
- Skills by role: what leaders, managers and everyday users each need. The government's AI Skills Framework groups these into technical, non-technical, and responsible and ethical skills.
- Sequence: which teams and tasks go first, and what result would justify moving on.
- Checkpoints: when you will measure and decide.
Our AI consulting work covers this, from a short discovery and roadmap sprint to longer adoption programmes.
How do leaders build an AI adoption strategy?
Tie AI use to a small number of business goals, give it a named owner, set clear rules, and fund the training and time people need. Then lead by example, because staff watch what leaders actually do.
Leadership comes up repeatedly in the government guide. Among its common pitfalls: "Training is less effective when leaders do not prioritise it, link it to organisational goals, or model its use". Another is that "Unclear rules, responsibilities, and goals limit adoption and reduce confidence".
A workable strategy answers five questions:
- Which business goals should AI help with first, and which tasks sit under them?
- Who owns AI decisions, and who supports them in IT, data protection and HR?
- What are the rules on tools, data and human checking?
- How will people get time to learn and practise?
- How will you measure whether it worked, and when will you decide to expand?
Leaders need their own training too, focused on setting expectations, judging output and managing risk rather than on clicks. See our executive and leadership AI training.
Why do AI licences go unused in large organisations?
Usually because people were given access without being shown how AI fits their own work, given time to practise or told what is allowed. A licence answers the question "can I?", not "how, for what, and is it safe?".
The government guide lists the pitfalls it sees most often, and several describe unused licences exactly. "Training is not linked to real work", so it is "too general or tool-focused, making it hard for employees to apply in their roles". "Employees lack time, support, and access to learn". And "Existing skills and informal AI use are not recognised".
Other causes we see:
- No named tasks. Nobody has agreed what the tool is for in each team.
- Unclear rules. People are unsure what data they can use, so they avoid the tool altogether.
- Features they cannot reach. People are trained on features their licence does not include, then find them missing.
- One session, no follow-up. Enthusiasm fades within weeks if nothing in the working week changes.
- Nobody measuring. Licence counts are reported, but nobody checks who is using the tool for real work.
How can a business assess its readiness for AI?
Check five things: whether there is a plan and an owner, which tools people can actually reach, how skilled and confident they are, how much real use there is today, and whether there are rules people follow. Most organisations are strong on some of these and weak on others.
Our free AI readiness scorecard asks ten questions, two in each of those five areas (strategy, tools, skills, adoption and governance), takes about three minutes, and names the two changes likely to make the biggest difference.
The government also publishes a free Employer AI Adoption Checklist, described as "a self-assessment tool for organisations to evaluate readiness to adopt AI, identify skills gaps and plan inclusive adoption strategies."
Before Microsoft Copilot training, we run a readiness check with you covering which tools and Copilot features your people can access, which use cases to focus on, and the data and permissions points to be aware of. Licence provisioning and tenant configuration stay with your IT team or Microsoft partner. See Microsoft Copilot training.
How do you increase AI adoption after rolling out licences?
Pick a few tasks per team, train people on exactly those tasks, and follow up within a few weeks to see who is still using the approach. Reinforcement after training matters more than the size of the launch.
The government guide recommends follow-up rather than one-off sessions: "Follow-up check-ins and short refresh sessions help reinforce learning." It also suggests that "Revisiting training after 3 and 6 months can help staff reflect on how they are using AI and make changes to their way of working."
What helps after the licences are live:
- Task-based sessions for each team on the two or three tasks agreed in advance.
- A shared prompt library for each team, so good prompts are reused rather than rebuilt. An empty library a month on is a useful warning sign.
- Internal champions who can answer questions in the moment. We can train your champions and hand over materials so the capability stays with you.
- Regular check-ins at set points to see who is still using it, and for what.
- Usage data from the tool's own admin reports, read alongside task measures rather than instead of them (see the Copilot question below).
How do organisations identify useful AI use cases?
List the tasks people repeat every week and score each one on frequency, value, risk, data sensitivity and ease. The best first use cases are frequent, low risk, use data you are allowed to process, and already have a review step.
A simple scoring grid:
Informal use is often the best source of ideas. Ask people, without blame, where they already use AI. The government guide recommends "use-case libraries that show safe and practical uses" as part of practical training. A short AI hackathon can also help a cross-functional group test ideas against real tasks quickly.
| Criterion | Question to ask | Good first candidate |
|---|---|---|
| Frequency | How often is the task done, and by how many people? | Weekly or more, by most of the team |
| Value | What does a faster or better result unlock? | Time freed for client or revenue work |
| Risk | What happens if the output is wrong? | Errors are caught by an existing review step |
| Data | Does the task need personal or confidential data? | No, or only within an approved enterprise tool |
| Ease | Can people do it in tools they already have? | Yes, with no integration work |
What should a company do before implementing AI?
Agree which tools are approved, what data is off limits, when a person must check the output and who owns the decision to expand. Then record how long your chosen tasks take now, because a baseline collected after training is not a baseline.
A short checklist before you start:
- Approved tools and the account type for each (consumer, business or enterprise).
- Data rules: what is public, internal, confidential or personal, and what may go into which tool.
- Human checks: where a person reviews or signs off AI output.
- An owner for AI decisions, even if part time.
- A baseline for two or three tasks: time, frequency and how often the work comes back for rework.
- Training first. Among the government guide's criteria for good training: "baseline AI training is mandatory before staff use workplace AI tools, particularly where AI use involves organisational data, confidential information, regulated activity, safety, or professional judgement".
If personal data is involved, UK GDPR still applies, and the ICO's guidance on AI and data protection is the starting point. Our employee AI policy guide covers the rules in more detail.
What are the biggest barriers to AI adoption in organisations?
The biggest barriers are usually training that is not linked to real work, lack of time to learn, unclear rules and weak leadership support, rather than the technology itself. Cost and availability of training matter too.
Figures from the government guide's employer survey: "Employers report cost (42%) and limited availability (37%) as key challenges to participation." Employers also identified wider gaps, "including lack of clear AI skills frameworks (35%), limited focus on ethics and governance (29%), and weak leadership or organisational buy-in (22%)." The guide notes that while "nearly all organisations report providing AI training (97%), many still identify key gaps in provision, particularly in flexibility (51%) and practical, contextualised learning (34%)."
Each barrier has a practical answer:
- Not linked to real work: train on the tasks agreed with each team.
- No time: give people protected time to learn and practise, and keep sessions short and focused.
- Unclear rules: write a short, plain set of rules before the pilot starts.
- Weak leadership support: brief leaders first and ask them to use the tools visibly.
- No measurement: agree what you will measure before training, not afterwards.
How can a leadership team prepare for AI adoption?
Leaders should get hands-on with the tools themselves, agree who owns AI decisions, set the rules and decide what success would look like before anyone else is trained. Their job is to give direction, time and permission.
Practical steps for a leadership team:
- Use the tools. A short session for leaders on their own work, so they can judge output and set realistic expectations.
- Name an owner with authority to approve tools and expand use.
- Set the rules on data, tools and human checks, in plain language.
- Choose the first teams and tasks, and the measure that would justify expanding.
- Protect time for people to learn, and say so publicly.
- Model it. Staff take their cue from what leaders do. The government guide warns that training is less effective when leaders do not link it to organisational goals or use it themselves.
Our executive and leadership AI training is built for this stage.
Why aren't our employees using ChatGPT despite having licenses?
Usually because they have not been shown how ChatGPT helps with their own tasks, are unsure what data they can put into it, or tried it once, got a poor answer and stopped. Access on its own rarely changes how people work.
Common causes we see:
- Generic training. A demo of what ChatGPT can do in general does not show someone how to use it for their weekly report.
- Unclear data rules. If people do not know what is allowed, many avoid the tool, and some carry on using personal accounts instead.
- Weak first results. A vague prompt gives a vague answer. People need to learn how to brief the tool: the goal, the audience, the source material and what good looks like.
- No checking habit. People who do not know how to check output either distrust it or trust it too much.
- Nothing changed in the process. If the team's way of working stays the same, the tool stays optional.
The fix is the same as for any tool: agree tasks, set the rules, train on real work, and check use a few weeks later. See our ChatGPT training.
Why is our Copilot adoption rate so low across teams?
Copilot adoption is often low because people are trained on features their licence does not include, are shown generic demos rather than their own work, or are not sure what Copilot can see in your files. Adoption also varies between teams because the useful tasks differ.
First, check what the number means. Microsoft's Copilot usage report describes its active users rate as the number of active users divided by the number of enabled users. It tells you who tried Copilot, not whether any task got faster or better.
Then look at the usual causes:
- Features people cannot reach. What Copilot can do depends on the licence each person has. Training should match the features they will see on Monday.
- Content Copilot can reach. People may hold back if they are unsure what Copilot can see through your existing file-sharing permissions.
- Generic sessions. Finance teams working in Excel need different practice from marketing teams drafting content.
- No follow-up. One session, then nothing, and use drops off.
This is why our Microsoft Copilot training starts with a readiness check on the tools and features your people can access. We do not provision licences, set up your tenant or configure Copilot; that stays with your IT team or Microsoft partner.
How do we measure ROI from corporate AI training programs?
Measure two or three real tasks before and after training: how long they take including checking, how often they happen, and how often they come back for rework. Then report hours released separately from money saved, because time freed is only cash if a real cost is removed.
This is the method on our measuring AI training ROI page:
- Choose the tasks. Repeated weekly, measurable, and done by most of the people being trained.
- Agree a baseline before training. A good programme records a baseline before anyone is trained: minutes per task, how many times it happens per person per week, and how often it needs rework.
- Define good and safe. The quality bar a reviewer would apply, and the safe-use checks: what must be verified, what data must never be pasted in, and where a person signs off.
- Check again after training. A good programme repeats the same measures at follow-up checkpoints it sets in advance, using the same method as the baseline, and reads the vendor's usage data (such as Microsoft's Copilot usage report) alongside them, since usage data shows who is active but not whether the work improved.
- Turn minutes into capacity. Capacity hours = (baseline minutes minus post-training minutes) × task volume ÷ 60. Post-training minutes must include the time spent checking the output.
- Keep capacity and cash apart. Hours released become cash only if a cost such as overtime, contractor spend or an outsourced task is removed.
- Count all the costs once. Participant time, training fees, AI tool licences attributable to the group and internal support time.
Treat attendance, satisfaction scores and completion rates as measures of the event, not of the return. Note what else changed at the same time, such as a new tool version or a process change, and do not credit the training with all of it. The government guide lists "Impact is not measured and training is not sustained" among its common pitfalls.
What this means for your team
Adoption is a habit, not a launch. Pick two or three real tasks per team, agree the rules on tools and data, train people on exactly those tasks and record how long the work takes now. Check the same measures again in the weeks after training, and expand only what worked. Ask leaders to go first and to give people time to practise. If you do only one thing this month, record a baseline for one task before anyone is trained, because you cannot go back for it later.
Sources
- GOV.UK: Employer guide: What works for AI upskilling in the UK
- GOV.UK: AI skills tools package
- ICO: Guidance on AI and data protection
- Microsoft Learn: Microsoft 365 Copilot usage report
Talk it through
If you would like help applying this to your own teams, book a free 60-minute discovery call with Qasim or get in touch. Tell us the tools you have and the tasks that take the most time, and we will suggest where to start. There is no obligation to book anything.