Most case studies about AI in construction start with a Fortune 500 company, a big-bang transformation programme, and a budget that exists nowhere in the real Middle Eastern market. So let's talk about something more useful: a 30-person contracting firm in Riyadh, the kind I see all the time, and what they could realistically do with AI starting Monday morning.
This is a composite — pieces of several real firms blended together so I'm not pointing at anyone in particular. Call them Al-Bina Contracting. Thirty people, ten years in the market, mid-size residential and small commercial projects, two ongoing jobs and three more in tender. Three engineers, four site supervisors, a small admin team, the rest are field staff. The owner-manager makes most decisions. No dedicated IT person.
What could they do?
Week one: drafting and translation
Spend nothing. Sign up for a paid ChatGPT or Claude account — call it 90 SAR a month for the firm. Train two people to use it well: the owner-manager (for client correspondence) and the head engineer (for technical responses to consultants).
Use it for:
- First drafts of letters in formal English to consultants and clients
- Translating internal memos from Arabic to English when needed
- Cleaning up tender clarification responses before they go out
- Summarising long emails into action items
Do not use it for:
- Anything from clients with strict NDAs
- Pricing or estimation
- Anything where the firm's own contractual position is at risk
Estimated time saved across the team: 5–8 hours a week. Estimated cost: 90 SAR a month.
Week two to four: document handling
Set up a workflow for tender documents. When a new tender arrives, before anyone starts marking it up, run it through an AI tool with a standard prompt:
Summarise this tender in one page. List all deadlines. List all scope items. List any unusual clauses (liquidated damages, penalties, performance guarantees). Flag anything that looks risky for a small contractor.
The result is a one-page brief that the owner-manager can read in five minutes before the team begins detailed work. Bid/no-bid decisions get faster. The team doesn't waste days on tenders that were always going to be a bad fit.
This single change could be worth more than everything else combined. Small contractors lose money mostly on bad bids, not on bad delivery.
Month two: site documentation
Train the supervisors to use voice-to-text on their phones to dictate observations during site walks. At the end of the day, paste the transcripts into an AI tool with a prompt to format them as a daily report. Owner-manager reviews, signs, sends.
This used to be a 30-minute task at the end of an exhausting day. Now it's a five-minute review, which means it's actually done consistently. Better records, fewer disputes, less stress.
Important: be explicit about safety incidents. The model will calmly summarise everything unless told otherwise. Always have the supervisor flag safety items before AI sees them.
Month three: a custom tool, finally
By now, the firm has used AI enough to know its limits. Time to build one small custom thing.
A reasonable first build: an internal tool, hosted on a single laptop, that takes a project's specifications and produces a draft material order list. Built with off-the-shelf components — a small interface, a language model called via API, a database of previously-ordered items to ground the suggestions. Maybe a weekend of an engineer's time, plus 200 SAR a month of API credits. The tool turns a half-day task into a two-hour task and helps avoid forgetting items that always get forgotten.
Don't try to make this product-grade. It's a tool for one team, in one office, used by one engineer. Internal tools get to be ugly, fragile, and useful.
What to skip
In the first six months, don't:
- Try to roll out AI to the whole team. Two power users beat thirty reluctant ones.
- Buy "AI for construction" SaaS that costs thousands per month. Almost everything those products do, your engineers can do directly with a paid ChatGPT or Claude account, with full data control.
- Promise the client AI-driven anything. Quietly use it where it helps; let the client see better outputs, not flashier marketing.
The point
The headlines about AI in construction promise robots and drones and computer-vision-driven safety systems. Those exist, and some of them work, but they're not where a 30-person firm in Riyadh starts. The starting line is much closer than that — paid chatbot, smart prompts, two trained people, an honest data policy. From there, real value compounds quickly.
The firms that win this decade in regional contracting won't be the ones that bought the most AI. They'll be the ones that figured out, faster than their competitors, which 20% of AI was actually useful for them, and ignored the rest.