Notes from the Field · October 2026 · 4 min read

AI That Does Not Speak the Client’s Language

Arabic is spoken by more than 500 million people and makes up about 0.7 percent of global URLs. In AI-enabled business development services, language is part of the design.

Most micro and small business owners in this region do not run their businesses in formal written Arabic. They price, negotiate and complain in dialect, and more and more of it happens in voice notes. That is the language any AI tool built for small business support has to understand. New evidence from the World Bank says most of these tools do not.

What the World Bank Found

The World Bank’s October 2026 Economic Update for the Middle East, North Africa, Afghanistan and Pakistan, titled “From Divide to Opportunity: AI, Jobs, and Growth,” names three gaps that decide whether AI pays off in the region: localization, usage and foundational capital.

The localization gap matters most for small business support. AI learns from local data, context and language, and the region’s languages are largely missing from what these systems learn from. Arabic is spoken by more than 500 million people, yet it accounts for only about 0.7 percent of global URLs. Urdu accounts for 0.03 percent. Local dialects, the Bank notes, are even less present.

The headline of the report is cautiously hopeful. AI is more likely to complement than replace workers across the region, and fewer than 10 percent of jobs face near-term automation risk. The second gap is the warning that program designers should read twice: access to AI tools does not automatically translate into productive use.

Why This Lands on Business Development Services

AI is arriving in Business Development Services (BDS) through chat assistants that draft business plans, automated diagnostics, and generators for marketing content. Each of these tools learns from text. When that text is mostly English, and the Arabic in it is mostly formal and written, the tool becomes fluent in the language of donors and consultants and far less fluent in the language of the client.

The advice can be correct and still miss. A cash-flow tip phrased for a bank officer. A sales script that no customer in Upper Egypt would say out loud. A diagnostic question the owner answers politely and wrongly because she did not follow it.

This is where the two gaps meet. The program reports access: accounts opened, tools deployed, users registered. The entrepreneur tries the tool twice, finds that it does not understand her, and goes back to her supplier and her cousin for advice. The indicator says reached. Her behavior says nothing changed.

Treat Language as Design, Not Translation

The one point of this piece: in AI-enabled BDS, language is part of the design, not a translation step at the end. Four practical moves follow from it.

  1. Test with the client’s own words before launch. Use real voice notes, real dialect and the questions owners actually ask, not a clean script written by the program team. This belongs in the diagnostic that comes before design.
  2. Measure use, not access. Track repeat use after 30 days and whether the advice changed a decision. Accounts created tells you very little.
  3. Keep the advisor in the loop. The human advisor who speaks the client’s dialect is the bridge between the tool and the business, not the cost the tool removes.
  4. Contribute to shared language data. The World Bank’s answer is regional collaboration, with middle-income countries contributing talent and local data. BDS networks and programs hold exactly the kind of everyday business language that AI training data lacks. With consent and proper data protection, they can help close the gap rather than wait for someone else to close it.

The Business Development Institute (BDI), which I co-founded, built its own AI-enabled platform for the people who grow businesses, so this is not an outside view. Any platform serving entrepreneurs in this region has to answer the same question: does it work in the language the client actually uses?

Across more than 16,800 practitioners trained in 18 countries, the advisors that small business owners trusted were the ones who could explain a cash-flow statement in the client’s own words. AI tools will be judged by the same standard.

Work With Me on This

If your program is bringing AI into training, coaching or advisory work, I help teams design it around the people who will use it.

See the Service →

Sources

← All Perspectives

Organizations I Have Worked With
USAIDGlobal Affairs CanadaWorld BankCentral Bank of EgyptCommercial International BankIFCKfWEuropean UnionHabitat for Humanity InternationalAga Khan Agency for MicrofinanceDrosos FoundationUN WomenUN-HabitatLFS AdvisoryGoss GilroyEnvironment and Development GroupAECOMFINBI