Explainers

How work is changing with AI, and what the next few years may look like

Work is shifting from doing every step yourself to directing AI helpers and checking their results. Here is what is changing now, what the data says for Kenya, and how to prepare your team.

Two smiling African businesswomen working together on laptops in a modern office.

The short answer

Work is moving from doing every step yourself to telling an AI what you want, then checking and deciding. The AI handles first drafts, sorting, searching and routine steps. People keep the judgment, the relationships and the final say.

This is not a prediction for 2040. The first pieces arrived this month. The rest of this post explains what is changing, what the numbers say, and how a Kenyan business can get ready without panic.

Three shifts happening now

1. From tools you use to helpers that work for you

Until recently, AI meant a chat box. You typed, it answered, and the work stopped when you closed the tab.

That is changing. On 29 September, OpenAI launched Dots, ChatGPT agents that run on their own cloud computers and keep working toward a goal while you are away. Meta's Muse Spark 1.3, released on 2 September, is aimed at coding and agent tasks too. An agent is an AI program that works through several steps on its own, instead of answering one question at a time.

2. From doing the work to directing and reviewing it

When a helper can draft the invoice, the reply or the report, your job moves up a level. You decide what to ask for, check what comes back, and fix what is wrong. OpenAI itself says its agents can still make mistakes, and advises people to review anything that matters.

Reviewing is a real skill. It means knowing what a good result looks like and spotting what is missing.

3. From expensive to cheap

The best models keep getting cheaper to run. Anthropic says its new Opus 5.5 costs about 40% less to run than Opus 5. OpenAI cut API prices for GPT-6 Sol and Luna by 50% compared with GPT-5.6 promotional pricing. Cheaper models mean more jobs become worth handing to AI, including small ones that would never have justified the cost before.

What the data says

Nobody knows exactly how this plays out, but two sources give a useful picture.

Globally. The World Economic Forum's Future of Jobs Report 2025 expects 170 million new roles and 92 million displaced by 2030, a net gain of 78 million. It also expects nearly 40% of the skills used on the job to change, and says 59 of every 100 workers will need new training. Employers' plans point both ways: 77% plan to upskill their staff, and 41% expect to reduce their workforce where AI can do the work.

In Kenya. A November 2025 analysis by the Overseas Development Institute found that about 2.5 million Kenyans work in roles highly exposed to generative AI, out of roughly 16.7 million workers. Most of those are in formal office work: clerks, administrators, bookkeepers, shop assistants and cashiers. The same analysis says most of the workforce, including farmers, boda boda riders and construction workers, faces little immediate disruption.

Two things to keep in mind: "exposed" means parts of a job could be automated, not that the job disappears; and these are estimates made before this month's releases.

What it looks like in a Kenyan business today

Here are examples of how the shift can look in practice. They are illustrations, not forecasts.

A small agency uses AI for first drafts of proposals and social captions. Staff spend their time on client calls, taste and quality checks instead of typing from a blank page.
An accounts office lets AI read supplier invoices and flag mismatches. The accountant reviews exceptions rather than keying in every line.
A retail shop uses AI to draft replies to WhatsApp enquiries and a weekly stock summary. The owner still handles complaints and negotiates with suppliers.
A solo consultant has an AI helper that sorts the inbox overnight and drafts follow-ups. They approve and send.

In each case, the person is still accountable for the result. The AI removes the repetitive part.

What the next few years may look like

This section is our reading of the trends above, not a forecast with data behind it.

More people will manage AI helpers. Many staff may supervise two or three agents, each doing routine tasks, much like a team lead supervises junior colleagues today.
Small teams will be able to do more. A five-person business may handle work that used to need ten, mainly in admin, research, content and customer follow-up.
Judgment and relationships will matter more. When anyone can produce a decent first draft, what stands out is knowing the customer, making the call and having taste.
Entry-level work will change shape. Routine tasks that used to teach beginners the ropes may be handled by AI, so businesses will need to design new ways for young staff to learn.
Trust and data rules will matter. If an agent can read your email and files, you need clear rules on what it can touch and who is accountable when it gets something wrong.

Skills that keep their value

Kenyan employers are already asking for a mix. BrighterMonday's list of in-demand skills for 2026 puts AI and data skills at the top, alongside creativity, problem solving, interpersonal skills and adaptability. The pattern matches the WEF's: technical know-how plus human skills such as analytical thinking, resilience and collaboration.

Common mistakes

Waiting until it is "settled". The tools change every few weeks. Waiting for things to calm down means falling behind people who learn as they go.
Handing over work without checking it. An unchecked AI mistake sent to a customer is still your mistake.
Treating it as a staff-cutting tool only. Teams that use AI to take the boring work off people, then move them to customer-facing work, tend to get more out of it than teams that only cut costs.
Giving agents access to everything. Start with read-only access and keep payments, banking and sensitive customer data out of reach until you trust the setup.
Skipping the training. Buying a tool without teaching staff how to ask, check and correct leads to unused subscriptions.

What to do this quarter

List your team's repetitive tasks. Pick the three that eat the most hours each week.
Try AI on one of them. Use a normal chat assistant first, with real examples from your business.
Write down what "good" looks like for that task, so anyone checking the output has a standard.
Set data rules. Decide what staff may and may not paste into AI tools.
Teach the team to review. Practise spotting errors in AI drafts, not just producing them.

If your team would benefit from hands-on practice with your own work, Glala Studios runs AI training for teams in Kenya.