On this page
TL;DR:
- Yes, for most teams. Staff need judgement and communication at the moment they use AI, so it makes sense to practise them together on real work.
- AI moves the human job from writing first drafts to briefing, checking and taking responsibility. Those are soft skills.
- Kenyan employers already report gaps here. In a 2023 Federation of Kenya Employers survey, 49.1% of firms named communication as the social skill most lacking in applicants.
- The evidence shows a risk: people who trust AI more report thinking less critically about its work.
- Keep them separate in some cases, such as deep technical AI skills or serious team conflict. The decision table below will help you choose.
Who this guide is for
This guide is for HR managers, learning and development leads and team leaders who are planning staff training in Kenya. You may have one budget line for "communication skills" and another for "AI". You are wondering whether to run them apart or together.
The short answer: combine them when the work needs both
Combine soft skills and AI training when staff will use AI for work that goes to another person: emails, reports, proposals, feedback, customer replies, meeting notes. In all of these, the tool produces the words and a person is responsible for whether they are right, kind and useful.
Run them separately when the goal has nothing to do with the other skill. Those cases are listed further down.
What actually changes when teams use AI
Every soft skill still matters, and several now show up in new places.
| Skill | What it looked like before | What it looks like with AI |
|---|---|---|
| Written communication | Writing from a blank page | Briefing the tool clearly, then editing hard for the reader |
| Critical thinking | Weighing an argument or a report | Asking "how do I know this is true?" of a confident answer |
| Judgement | Deciding what to do | Also deciding when not to use AI at all |
| Empathy | Reading how a person feels | Noticing what a polite draft leaves out |
| Giving feedback | Telling a colleague what to improve | Doing that when the work may be partly AI |
| Teamwork | Dividing up the work | Agreeing who has checked what, and who owns it |
| Leadership | Setting direction, setting an example | Setting AI rules, using AI openly, protecting time to check |
A 2025 study of 319 knowledge workers from Carnegie Mellon University and Microsoft Research describes the same shift. Participants gave 936 examples of using AI at work. The authors found that AI moves critical thinking toward checking information, fitting AI's response into the work, and overseeing the task.
The evidence, in plain terms
Four pieces of research make the case. All have limits, and most were not done in Kenya.
People who trust AI more may think less. In that same study, higher confidence in AI was linked to less critical thinking. Higher confidence in one's own ability on the task was linked to more. It is a survey of what people reported, so it shows a link and not a cause. It still gives a clear training goal: build people's confidence in their own judgement along with their AI skills.
AI can make good people wrong. A pre-registered experiment with 758 Boston Consulting Group consultants found that those using AI on tasks it handled well completed 12.2% more tasks, worked 25.1% faster and produced results rated more than 40% higher in quality. On a task selected to be outside AI's ability, consultants using AI were 19 percentage points less likely to reach the correct answer. The hard part is that nobody is told which kind of task they are holding. That takes judgement.
Poor AI work costs colleagues time. BetterUp Labs and the Stanford Social Media Lab surveyed 1,150 full-time US desk workers in September 2025. They use the word "workslop" for AI content that looks good but lacks substance. 40% said they had received some in the previous month, and each case took about two hours to deal with. These are self-reported US figures, so treat them as a warning and not as a Kenyan estimate.
AI can change how people work together. A field experiment with 776 Procter & Gamble professionals found that individuals using AI matched the performance of teams working without AI. It also found that AI use was linked to more positive feelings during the work. This is good news, and it raises a real question for managers: if people can get more done alone, what do you do to keep them talking?
The Kenyan picture
Kenyan employers were raising these concerns before AI became common at work. The Federation of Kenya Employers' Skills Needs Survey, reported by The Star in November 2023, listed the social skills firms found most lacking in applicants:
| Skill | Share of responding firms |
|---|---|
| Effective communication | 49.1% |
| Critical thinking | 41.7% |
| Teamwork | 25.7% |
| Time management | 23.4% |
That survey predates wide workplace AI use, and it covers job applicants. Its relevance is simple. If communication and critical thinking were already short, AI gives people a way to produce polished text without building those skills. A manager reading it may not notice the gap until a client does.
Local context matters too. A 2024 study tested seven AI models on code-mixed WhatsApp chats in Swahili, English and Sheng, and found all of them had difficulty with cultural nuance. A message to a senior client in Nairobi needs a different opening from a message to a peer, and a person has to judge that. For more on what employers look for, see the skills Kenyan employers screen for in 2026.
When to combine and when to separate
| Your situation | Combine? | Why |
|---|---|---|
| Staff write to customers, clients or senior leaders | Yes | Tone and accuracy are the whole job |
| Managers give feedback or handle difficult conversations | Yes | Good practice for preparing, with people deciding |
| Team is new to AI and needs a confidence boost | Yes | Safe practice on real work builds both |
| Finance team needs advanced Excel and Copilot skills | Mostly separate | Add a short session on checking and explaining numbers |
| Developers are learning to use AI in code | Separate | Technical depth needs its own time |
| Team has serious conflict or low trust | Separate first | A skilled facilitator is needed, and AI is a distraction |
| Leaders need an AI policy and governance view | Separate | A different audience with different decisions |
| Staff have poor basic writing | Foundation first | They need something to judge drafts against |
A 90-minute session you can run
Here is a session plan for a group of up to 20. It uses one task that every team recognises: a difficult email to a client.
Use an invented or fully cleaned example. Nobody should paste real client details into a public tool.
The scenario (invented). A supplier has delivered late for the second time, and your largest client is threatening to leave. You must write to the client today.
| Time | Activity | What people practise |
|---|---|---|
| 10 minutes | Everyone writes the first three lines by hand | Finding out what they actually think |
| 15 minutes | Write a prompt using the four Cs, generate a draft | Briefing clearly |
| 15 minutes | In pairs, mark up the draft: what is untrue, cold, risky or missing? | Critical reading |
| 15 minutes | Rewrite the email, then swap | Editing for the reader |
| 10 minutes | Act it out: one person plays the client and reads it aloud | Empathy |
| 15 minutes | Group discussion: what would you never let AI decide here? | Judgement |
| 10 minutes | Each person writes one rule for their own team | Leadership |
We put the handwriting first on purpose. Someone who starts from the AI draft has only that draft to react to. Someone who first writes their own view has something to compare it with.
Prompts to practise with
These use the four Cs we teach: Clarity, Context, Constraints and Customisation. Replace the square brackets, and keep real names and personal details out.
Have AI critique your draft, without rewriting it
Below is a draft email I wrote. Do not rewrite it. Context: it goes to a long-standing client in Nairobi who is unhappy about repeated late deliveries. I am the account manager. Constraints: give me no more than five comments. For each, quote the exact sentence, say how the client might feel reading it, and ask me one question to help me fix it. Customisation: be direct, as a senior colleague would be. [paste draft]
Asking for questions keeps the person doing the thinking.
Rehearse a hard conversation
Act as a team member who has missed three deadlines and feels I never explain priorities clearly. I am their manager, and this is a private conversation. Constraints: respond one turn at a time and wait for me. Be realistic, a little defensive, and do not make it easy. Do not give advice during the conversation. After eight turns, stop and tell me: where I listened well, where I talked over them, and one thing to try differently.
Rehearse with invented details. The real conversation stays between two people.
Test your own reasoning
I am about to recommend this to my director: [one-paragraph recommendation with no personal or confidential details]. Give me the three strongest objections a sceptical finance director would raise. For each, tell me what evidence would answer it. If you are unsure of a fact, say so. Do not make up figures.
Prepare to give feedback on AI-assisted work
A colleague has given me a report that looks polished, but it has generic points and no real recommendation. I do not know if AI wrote it. Help me prepare a respectful five-minute conversation. Constraints: focus on the work and do not accuse them of using AI. Give me two opening lines, three specific questions about the content, and one clear request for what to change. Customisation: we work in a Kenyan office where the colleague is older than me.
The editor's checklist
Print this and give it to everyone who sends AI-assisted work to another person.
- Is every fact, name, number and date correct, and did I check it myself?
- Did I ask what is missing?
- Does this sound like me, or like nobody in particular?
- Is it the right length for the person reading it?
- Would I be comfortable explaining every sentence if asked?
- Is there anything here I would not say to this person's face?
- Have I made a clear recommendation or ask?
- Should I tell the reader that AI helped?
What this means for leaders
Team leaders set the standard for AI use, whether they intend to or not. Three habits matter:
- Say openly how you use AI. If leaders hide it, staff will too, and errors stay hidden.
- Judge the work, and ask how it was checked. "Who checked this, and how?" is a better question than "Did you use AI?"
- Protect time for review. If AI makes the first draft faster and you fill the saved time with more work, nobody will check anything.
Managers are often the group that decides whether training lasts. See our guide to getting middle managers to use AI.
What usually goes wrong
| Mistake | What it looks like | How to avoid it |
|---|---|---|
| Two separate courses, months apart | Staff never connect the tool with the judgement | Use one programme built on real tasks |
| Soft skills become a lecture | Slides on "emotional intelligence", then back to work | Practise on actual drafts and conversations |
| AI session is only tool tricks | People learn features and send weak work faster | Add a checking step to every exercise |
| AI is used to avoid hard talks | A manager sends an AI-written message instead of having the meeting | Agree which conversations must be in person |
| Everyone sounds the same | Team emails lose their voice | Write first; use AI to critique |
| Using AI to rehearse with real names | Staff details go into a public tool | Use invented details only |
| No follow-up | Skills fade without practice | See why corporate AI training does not stick |
How to measure the result
Test the skill itself. Attendance and happy feedback forms will not tell you much.
Before training, ask each person to complete a short written task. Repeat a similar task six weeks later. Have a manager who does not know which is which score both.
| Criterion | 1 (weak) | 3 (good) | 5 (strong) |
|---|---|---|---|
| Accuracy | Contains errors | Facts are correct | Facts checked, and sources noted where needed |
| Reader awareness | Could be sent to anyone | Fits the reader | Answers what this reader will worry about |
| Clear ask | No next step | Next step stated | Next step, owner and date |
| Voice | Generic | Sounds like a person | Sounds like this team |
| Judgement | Used AI without thought | Checked the main points | Can explain what AI got wrong and what they changed |
Also track a few real signs: how often managers send work back, client complaints about tone or errors, and whether staff report AI mistakes. A rise in reported mistakes early on can be a good sign. It may mean people are checking.
If you are not sure your organisation is ready, start with our AI readiness assessment.
Where training will not help
A single workshop will not turn a poor communicator into a strong one. Nobody becomes an expert in a few days. Training also cannot fix unclear priorities, unfair workloads or a culture where people fear admitting mistakes. If people do not feel safe telling you that AI got something wrong, deal with that first.
Questions people ask
What are soft skills in the context of AI?
They are the human skills needed around AI work: communicating clearly, thinking critically, using judgement, showing empathy, working as a team and leading others. AI can imitate the language of these skills. A person still has to supply the judgement and take responsibility.
Will AI make communication skills less important?
Our view is the opposite for most office jobs. As more text is produced quickly, people need to judge it, shape it for a reader and say what they actually mean. Employers in Kenya already named communication as a top gap before AI became common.
Can AI be used to teach soft skills?
Yes, as a practice partner. Role-plays and feedback on drafts are useful, and are available on free tools. It cannot watch body language or know your team's history, so human coaching still matters.
Does using AI damage critical thinking?
We do not know for certain. A 2025 survey found people with higher confidence in AI reported less critical thinking. It does not prove AI causes decline. A sensible response is to build checking into the way people use AI.
How long should a combined programme be?
A useful start is a 90-minute session on one task. A stronger programme is one to two days of practice with follow-up sessions over six weeks. The right length depends on your team and how much real work you can bring in.
Who should attend first?
Start with team leaders where you can. They set the standard for what counts as good enough, and they coach everyone else after the trainer has left.
Should staff say when they have used AI?
For most internal drafts, this is a team decision. Where the reader would expect to know, or where a client contract or academic rule requires it, disclose. Whatever you decide, the sender remains responsible for the content.
Sources
- Lee et al., "The Impact of Generative AI on Critical Thinking" (CHI 2025), last checked 9 November 2026
- Dell'Acqua et al., "Navigating the Jagged Technological Frontier" (working paper, MIT Sloan copy), last checked 9 November 2026
- BetterUp Labs and Stanford Social Media Lab, workslop research, last checked 9 November 2026
- Dell'Acqua et al., "The Cybernetic Teammate", NBER Working Paper 33641, last checked 9 November 2026
- The Star, report on the Federation of Kenya Employers' Skills Needs Survey (21 November 2023), last checked 9 November 2026
- Ochieng et al., "Beyond Metrics: Evaluating LLMs' Effectiveness in Culturally Nuanced, Low-Resource Real-World Scenarios" (2024), last checked 9 November 2026
Where FuKazee fits in
FuKazee trains people in AI and in the work around it, including Microsoft 365, Google Workspace, leadership, communication and soft skills. We assume nobody in the room knows AI yet, we avoid jargon, and we show the free version of every feature. We have trained more than 3,000 people since 2022.
A combined programme with us uses your team's real tasks, with confidential details removed. People practise writing, checking, giving feedback and deciding what AI should not touch. We run sessions in Westlands, Nairobi, live online, or on-site by arrangement. See AI training in Kenya for how we work with organisations.
To talk through a combined programme for your team, email lead@fukazee.com
Want guided support instead?
Plan AI training around your team's real work. Answer a few quick questions about your sector, team size and goals, and we will recommend the right programme and format. Sessions are built on your workflows and include a usage policy.
Rodney has run FuKazee's training since 2022 and still teaches inside every programme line: corporate sessions, the postgraduate cohort and the KCSE examiner webinars. First Class Honours in Journalism and Mass Communication, University of Nairobi.
