AI agents4 min readBy Ahmed AkramPublished
The future of AI in business: what changes by 2030
Predictions about AI age badly, including ours. So we have split this one into what we're fairly sure of and what we aren't, and kept the list of things to do this year short.
Short answer
By about 2030, expect AI agents to handle routine multi-step work across your systems, better Arabic and multilingual support, lower costs per task, and more regulation. We are less sure how fast jobs change, how far agents can be trusted with money, and which vendors will last. The companies most at risk are those whose data is too messy for AI to use, whether or not they ignore AI.
What we are fairly sure of
Agents will do more routine multi-step work. Today a person reads an email, looks up an order in one system, checks stock in another and replies. Software that does those steps, asks for help when unsure and leaves a log is already working in small deployments. By 2030 it will be ordinary for invoice handling, support triage, reporting and order follow-up.
Arabic and multilingual support will keep improving. Dialects, mixed Arabic and English messages and scanned documents are still weak spots, but the gap has been closing and we expect it to keep closing. Costs per task will keep falling too, so jobs that are not worth automating today will be worth it later.
| Fairly sure | Less sure | |
|---|---|---|
| Routine work | Agents handle multi-step tasks across systems | How many jobs shrink, and how fast |
| Language | Better Arabic and multilingual support | Whether dialects reach the same quality |
| Cost | Cost per task keeps falling | Which vendors stay cheap and stable |
| Trust | More rules on how AI is used | How far agents are trusted with money |
Fairly sure
- Routine work
- Agents handle multi-step tasks across systems
- Language
- Better Arabic and multilingual support
- Cost
- Cost per task keeps falling
- Trust
- More rules on how AI is used
Less sure
- Routine work
- How many jobs shrink, and how fast
- Language
- Whether dialects reach the same quality
- Cost
- Which vendors stay cheap and stable
- Trust
- How far agents are trusted with money
Regulation will grow
More rules are coming. The EU AI Act is the best-known example, and in general terms it asks for more care, documentation and human oversight as the risk of a use rises. Other countries are writing their own versions, and they will not all agree.
If you sell into Europe, or handle personal data anywhere, assume you will need to explain what your AI does and keep records of it. That is easier if you build logging and clear approval points in from the start than if you add them under pressure later.
Which roles change, and how
We think tasks change before jobs do. Data entry, first-line email replies, report assembly and basic reconciliation shrink. Work that needs judgment, relationships, accountability or physical presence does not shrink much. People who check, correct and improve what an agent does become more valuable than people who only do the repeatable part.
We don't know how quickly this plays out, and anyone giving you a percentage of jobs lost is guessing. Plan for roles to shift, retrain people on review and exception handling, and watch what happens in your own team.
Likely to move to agents
- Copying data between systems
- First-line email and chat replies
- Assembling weekly reports
- Matching invoices to orders
Likely to stay with people
- Negotiating with key customers
- Deciding exceptions and refunds
- Owning the final decision
- Work that needs someone on site
Messy data is the main risk
An agent can only work with what it can read. If your prices live in three spreadsheets, customers are duplicated in two systems and key rules exist only in one person's head, the agent has nothing reliable to act on. A competitor with tidy data and clear processes will get useful automation sooner and cheaper.
This is the part we would worry about most if we ran a small company. Buying AI tools is easy. Getting your information into a state where they work is the slow job, and it benefits you even without AI.
What to do this year
None of this needs a large budget. Most of it needs a few weeks of attention from someone who knows how the business actually runs.
Pick one task
Choose a repetitive job with clear rules and a visible cost in hours.
Tidy its data
Fix the sources it relies on: one customer list, one price list.
Pilot with review
Run an agent with a person approving results, and measure errors.
Write simple rules
Decide what data may be shared with AI tools and who approves what.
Review and extend
Keep what worked, drop what did not, then pick the next task.
Questions people also ask
Will AI replace my employees by 2030?
Some tasks will disappear and some roles will change, mainly repetitive ones. We would not plan around mass replacement. Plan for people to spend less time on routine work and more on checking, exceptions and customers.
Is it too late to start with AI?
No. Tools get cheaper and easier every year, so you are not behind by starting with one small task now. What helps most is tidying your data and processes, which pays off whichever tools you choose later.
How does the EU AI Act affect a small company?
In general terms it applies more obligations to higher-risk uses of AI, and it can reach companies outside Europe that sell into it. Check your own situation with a lawyer, and keep records of what your AI systems do.
Related service
Want to pick the first task worth automating?
Talk to usMore articles
- AI agentsAre AI agents worth it?Where an agent pays for itself in a small business, where it doesn't, and how it compares with WhatsApp's built-in AI.Read the article
- AI agentsStop your support AI making things upWhy support bots invent policies and prices, and five guardrails that make it rare and harmless.Read the article
- AI agentsIs company data safe in ChatGPT?What happens to data you paste into AI tools, what never to paste, and a one-page policy your team will actually follow.Read the article
