TheAI-readinesschecklistforanyKurdistanbusiness

TL;DR: Most Kurdistan businesses I talk to do not need an AI strategy. They need ten boring things in place first. Once those ten exist, an AI project goes live in weeks and pays back in a quarter. Without them, AI is an expensive way to discover what was always wrong with the data. This is the checklist I run before quoting any project.

How to use this list

Score each item 0, 1, or 2. Zero means it does not exist or nobody owns it. One means it exists but is half-broken. Two means it actually works on a normal Tuesday. If your total is below 12 out of 20, fix the gaps before spending anything on AI. Below 8 and you are not ready — the AI build will surface problems faster than it solves them.

Data foundations (4 items)

1. There is one source of truth for customers

One sheet, one table, one CRM — somewhere a row says "this is the same person." If customer info lives in three different WhatsApp threads and a paper notebook, AI cannot help yet. It can help you set this up, but the AI assistant on top will not start paying off until the source of truth exists.

2. There is one source of truth for orders or bookings

Same idea. If yesterday's bookings are partially in WhatsApp, partially in a spreadsheet, partially in the head of the front- desk person — AI can read the messages but it cannot reconcile them with reality, because there is no reality.

3. Inventory or service catalog is current

AI replies to customers as confidently about a stocked item as an out-of-stock one. If your inventory drifts more than 20% from reality, an AI assistant will lie to customers an embarrassing percentage of the time. Spend the week to fix inventory hygiene first.

4. Pricing is documented and consistent

If the price of the same item is different on the menu, the delivery app, and what staff tells customers — AI will pick one and stick with it, and it will be wrong some of the time. One canonical price list, accessible to whatever AI you ship, is the cheapest fix you will ever do.

People and process (4 items)

5. One named owner for the AI project

Not "we as a team." A specific name. The person who decides what the AI does, what it must not do, and signs off on changes. Without this, the project drifts and three months later nobody is sure if it is helping.

6. Staff understand what AI will and will not do

A 20-minute team meeting before launch. What the assistant handles. What it routes to a human. How a customer reaches a human if the AI fails. Without this, frontline staff sabotage the rollout (rationally — they are protecting their jobs and their customers).

7. There is a defined "escalate to human" pattern

Every AI deployment needs an off-ramp: a clear keyword, button, or condition that hands the conversation to a human within minutes. Customers tolerate AI that occasionally fails as long as a human is one tap away. Without an off-ramp, one bad interaction loses a customer for a year.

8. There is a weekly review of failed cases

Twenty minutes a week looking at the conversations the AI did badly on. Not a dashboard, just reading the actual transcripts. That is the single highest-leverage habit for keeping an AI product good. Without it, the model decays silently and you only find out when a customer complains in public.

Technical baseline (4 items)

9. WhatsApp Business is set up properly

Not personal WhatsApp. The actual Business app, with a profile, catalog, and quick replies. AI plugs into this through the Cloud API. If you are still on personal WhatsApp, fix that first — it is a Saturday afternoon's work and unblocks everything else.

10. Phone numbers, addresses, and IDs follow one format

+964 prefix or 0 prefix or no prefix? Arabic digits or Latin? Pick one and clean the data. AI is forgiving but inconsistent formats double the build cost because every prompt needs normalization rules.

11. There is a backup of the system being automated

Whatever AI is plugging into — POS, booking system, CRM — has a working export and a recent backup. AI does not eat data but the integrations sometimes do. A 30-day backup window is the floor.

12. Internet at the location is good enough

AI assistants live in the cloud. If your shop's internet drops out twice a day, build assumes offline-tolerant flows from day one. Fixing internet is cheaper than building offline resilience into every workflow.

Business clarity (4 items)

13. The metric you want AI to move is named

"Reduce missed bookings by half." "Cut customer-service hours from 4 to 1 per day." "Stop losing CoD reconciliation disputes." Specific. Measurable. If the metric is "use AI more" — abandon ship and figure out the metric first.

14. There is a baseline number for that metric

You cannot tell if AI helped if nobody wrote down where you started. One week of measurement before launch beats six months of arguing afterwards.

15. The budget exists and is not borrowed from this month's payroll

AI projects in Kurdistan in 2026 land between $750 (audit) and $6,000 (build). Plus monthly runtime. If that money is precarious, the project will be killed mid-build over an unrelated cash-flow issue and you'll lose what you spent.

16. There is a kill criterion in writing

Under what circumstances do you turn the AI off? After two customer complaints? After the cost crosses a threshold? After a privacy incident? Decide this before launch when you are rational, not after launch when you are panicked.

Four bonus items that move you from ready to fast

1. A staging environment that mirrors production, so you can test AI changes without breaking live traffic.
2. Customers messaging mostly in WhatsApp text, not voice. Voice is solvable but adds 30% to build time.
3. A web presence the AI can reference — your site, a price list, a FAQ. The AI does not invent answers; it cites the source you give it.
4. A culture where staff bring up problems early. AI surfaces edge cases by the dozen in the first month, and a defensive team will hide them.

Want me to score yours?

Send a one-paragraph description of your business — what you sell, how customers reach you, what the busiest day looks like. I will reply with a score and the cheapest order to fix the gaps. Free. About fifteen minutes. Use the contact page or WhatsApp.