Which Small Businesses Are Most at Risk from AI?

There is a simple way to think about the new business environment. If customers mainly pay you to produce information, text, research, recommendations or digital content, AI can probably perform at least part of your work. If customers pay you because you physically do something, take responsibility for an important decision, understand complicated local regulations or have a trusted relationship with them, replacing you becomes much harder.

This is why someone running a translation business may feel much more pressure from AI than someone operating a hair salon. Both can use AI. But only one has a core product that AI can already produce digitally. The same pattern appears across many businesses that are particularly relevant to immigrant and Persian-speaking communities.

Translation has long been a natural business for bilingual immigrants. A Persian speaker living in Germany might translate between Persian, German and English. Someone in Canada might work with Persian-speaking newcomers who need help understanding documents. Others may provide subtitles, website translations or business communication. AI has changed this market dramatically.

For ordinary translation, a customer can now upload a document and receive a usable translation almost instantly. The quality is not always perfect, especially when cultural knowledge, technical language or Persian-specific context becomes important, but the basic service has become much cheaper. That means charging simply for converting Persian text into English or German is becoming a weaker business model.

The stronger opportunities are moving toward certified translations, legal documents, medical communication, live interpretation, technical material, cultural adaptation and human verification. In other words, the translator’s value increasingly comes from being responsible for the accuracy of the result rather than simply producing the first translation.

Research into Persian-language AI also shows why human expertise has not suddenly become irrelevant. Persian remains less extensively represented in many AI benchmarks than English, and models can struggle with cultural knowledge, specialized terminology and context. For a professional translator, AI may therefore become part of the workflow rather than the end of the profession.

For years, digital marketing has been one of the easier service businesses to start with relatively little capital. A small agency could offer social-media posts, blog writing, graphics, advertising campaigns and search-engine optimization. Almost every one of those activities is now being affected by generative AI.

A restaurant owner who previously paid someone to write Instagram captions can generate twenty variations in seconds. A small shop can create product descriptions automatically. Images, advertising concepts and basic videos can also be generated much faster than before. This does not mean marketing agencies have no future. It means that selling “ten Instagram posts per month” or “four blog articles per month” is becoming less defensible.

The valuable part moves higher up the chain. Understanding customers, creating a brand strategy, deciding what should be communicated, measuring whether advertising actually produces sales and understanding cultural differences between markets remain much more valuable.

For a Persian marketing agency abroad, cultural knowledge can also become part of the product. Knowing how to communicate with both Persian-speaking customers and the wider local population can be more valuable than simply knowing how to create social-media content.

Accounting is another field where it is important not to confuse tasks with entire professions. AI and modern accounting software can already categorize transactions, extract information from invoices, prepare reports and explain financial data. Routine bookkeeping is therefore under significant pressure.

A business that makes most of its money manually entering transactions may find that customers increasingly expect these tasks to be included cheaply or automatically. But accounting is much broader than data entry. Taxes, financial planning, company structures, compliance and complicated business decisions still require knowledge of the local system. In regulated areas, professional responsibility is also extremely important. This is especially relevant to immigrants.

A Persian entrepreneur opening a business in Germany may understand business perfectly well but struggle with German tax terminology. Someone starting a company in Canada may need help understanding local accounting rules. A Persian-speaking accountant who understands both the local system and the client’s cultural and linguistic background can provide something more valuable than bookkeeping alone. The business changes from “I enter your numbers” to “I help you understand your business and make sure important financial decisions are handled correctly.”

Immigration is one of the most interesting cases. Years ago, access to immigration information itself had significant value. A consultant might explain which visa existed, what documents were required and how an application worked. Today, someone can ask an AI system many of those questions. This reduces the value of simply repeating publicly available information. But immigration cases are not always simple.

Applications can involve previous refusals, unusual employment situations, family circumstances, documentation problems or changing regulations. In several major immigration destinations, paid immigration advice is also regulated. That creates an important distinction. AI can help someone understand the basics. A qualified professional can take responsibility for an individual case.

For Persian-speaking immigration professionals, the future is therefore probably less about selling information and more about strategy, verification, case preparation and representation. Trust also matters enormously when people are making decisions that can affect where they are allowed to live.

Travel planning is another business where AI can directly perform work that people traditionally paid for. Ask an AI assistant to plan ten days in Italy with a specific budget and it can immediately suggest cities, hotels, restaurants and an itinerary.

That makes the traditional model of simply helping someone find a hotel and flight less attractive. But difficult travel is different. Families traveling together, older travelers, group tours, complicated routes, cultural tours and travelers who want support in Persian may still value a human agent.

There is also a big difference between planning a trip and solving a problem when a flight is cancelled, a hotel reservation disappears or a family gets stranded somewhere. Travel agencies that survive may increasingly sell convenience and responsibility rather than information.

Now consider a Persian restaurant. AI can write the menu descriptions. It can generate advertisements. It can analyze customer reviews, suggest promotional campaigns and possibly help forecast demand. But it cannot cook a proper ghormeh sabzi for tonight’s customers.

This is the important difference between replacing a business and improving a business. Restaurants, catering companies, Persian grocery stores, beauty salons and many repair businesses have relatively strong protection because customers ultimately need something to happen in the physical world. That does not make them easy businesses. Restaurants can be expensive and difficult to operate. Grocery stores need inventory and premises. Salons require skill and local customers. But their main product cannot simply be generated on a screen.

For these businesses, AI may be much more of an opportunity than a threat. A restaurant with five employees could use AI for marketing, scheduling, translation, responding to reviews, drafting menus and organizing administrative work. The food still comes from the kitchen.

Persian supermarkets and specialty-food businesses are also interesting because they combine physical products with cultural knowledge. People do not visit a Persian grocery store simply because they need “food.” They may be looking for a particular brand of rice, dried herbs, saffron, sweets, pickles or ingredients that are difficult to find in ordinary supermarkets. AI cannot substitute for the product.

What it can do is change how people discover and buy it. A Persian grocery store can become more digital, offer local delivery, build an online store, create Nowruz packages, prepare gift boxes or use customer purchasing patterns to decide what should be stocked. In this case, refusing to use AI may eventually become a bigger risk than AI itself.

A similar principle applies to hairdressers, beauticians, electricians, plumbers, mechanics and many repair businesses. AI may eventually become very good at identifying what is wrong. You might photograph a damaged appliance and ask an AI assistant what the problem is. But diagnosis is not the same as repair.

Someone still has to come to the property, open the machine, replace the part and make sure everything works safely. The same is true in a salon. AI might show a customer what a certain hairstyle could look like, manage bookings and create advertising content, but it cannot currently cut someone’s hair.

For someone considering entrepreneurship today, businesses based on real-world skills deserve more attention than they often receive. In recent years, digital businesses looked especially attractive because they required very little capital. AI is changing that calculation. Some of the businesses that are cheapest to start are also the easiest for AI to disrupt.

For Persian entrepreneurs abroad, there is another important change happening. Speaking Persian used to create a strong barrier in certain markets. If a newly arrived Iranian needed someone who could translate a document, understand Persian and explain the local system, there might only be a small number of available professionals.

AI is weakening the language part of that advantage. But it is not eliminating cultural understanding or trust. There is a big difference between translating someone’s sentence and understanding what the person is actually trying to achieve.

Diaspora businesses often understand both sides. They understand the expectations of Persian customers, but they also understand how the host country works. That combination can remain extremely valuable.

A Persian accountant can understand why a newly arrived business owner finds the local tax system confusing. A Persian event organizer can understand what is expected at an Iranian wedding while also knowing local venues and suppliers. A Persian grocery owner knows which products matter to the community. A Persian-speaking tradesperson may make an older customer feel much more comfortable when discussing an expensive repair.

These advantages are difficult to reproduce with software alone.

One type of business looks increasingly vulnerable:

A business where the customer is mainly paying because the owner has access to information the customer does not have.

AI is making information extremely cheap. Basic translation, simple research, travel recommendations, generic content, standard advertising text and routine explanations are all moving in that direction. Businesses built around these services will need to evolve.

The stronger business model is increasingly based on something AI cannot provide cheaply: responsibility, trust, physical execution, regulated expertise, local relationships, specialized knowledge, original products or genuine human experience.

That is a much better question for anyone thinking about becoming self-employed. Do not only ask, “Can AI do some of this work?” Almost every business will eventually have tasks AI can perform. Ask instead: What is the customer ultimately paying me for, and can that part be easily replaced?

There is also an important warning. A business can be difficult for AI to replace and still be a terrible investment.

Restaurants are a perfect example. They are highly resistant to direct AI substitution, but they can require substantial capital, employees, equipment, rent, permits and constant management.

A translation service might have much higher AI exposure, but someone who already has the necessary expertise can start it with almost no capital. So choosing a business should never be based only on AI.

Startup costs, local demand, competition, qualifications, language ability, regulations and personal experience are still extremely important. AI simply adds another question that entrepreneurs can no longer afford to ignore.

One of the biggest mistakes may be assuming that everyone needs to start an “AI company.” Most entrepreneurs do not.

A Persian restaurant does not need to become an AI startup. A plumber does not need to build software. A Persian grocery owner does not need to become a machine-learning expert. They simply need to become more efficient than competitors who ignore the technology.

OECD research on small and medium-sized businesses shows that AI adoption is spreading quickly, but many companies still struggle to integrate it properly into everyday operations.

That creates an opportunity. The winning small business may not be the one selling AI. It may be the ordinary local business that answers customers faster, understands demand better, spends less time on administration and markets itself more effectively because it learned how to use AI behind the scenes.

Looking at the current direction of AI, businesses based on physical work, local presence, trust and professional accountability appear relatively resilient. Restaurants, catering, food businesses, salons, repair services and skilled trades are difficult to replace directly. Accounting, immigration services and education are likely to change significantly, but skilled professionals can remain valuable by moving away from routine information and toward judgment, responsibility and personalized service. Translation, basic digital marketing and simple travel planning face greater pressure because much of their traditional output can already be produced automatically.

That does not mean these businesses will disappear. It means their easiest work is becoming less valuable. And that may be the most important lesson for Persian entrepreneurs abroad. The question is no longer simply whether a business is online or offline, modern or traditional. The question is whether the part customers pay for is easy to automate. AI is making information cheaper.

Trust, responsibility, cultural understanding, real-world skills and human relationships are still much harder to replace.

For many small businesses, that is exactly where the next opportunity may be.

This article is based on Abhavij’s review of recent research on artificial intelligence, employment and small businesses, including work from the International Labour Organization, OECD, World Economic Forum and recent studies evaluating Persian-language AI systems. AI exposure does not mean that an occupation or business will disappear. In many cases, the more realistic outcome is that some tasks become automated while the business itself changes.

If you own a Persian-speaking business abroad, making it easier for people in your city to discover your services may become even more important as competition becomes increasingly digital. Abhavij is building a community where Persian-speaking businesses, services and people can be discovered more easily across different countries and cities.