Every owner is quietly asking the same thing, even if they won’t say it out loud: if I bring in AI, do I cut staff? The straight answer: for most businesses AI changes what your people do far more than it removes them, and the owners who win biggest use it to make good people more productive, not to cut heads. This post covers how to think in tasks rather than roles, what that looks like in a Saudi business, a worked example, where the “replace people” plan backfires, and how to start.
For most businesses, AI changes what your people do more than it removes them
In our experience, the owners getting the biggest wins aren’t shrinking their teams. They’re using AI to pull the repetitive, draining, low-value work off good people, so those people can do the work that actually grows the business. AI is a multiplier on the people you already have, not a swap for them.
Someone you already employ knows your customers, suppliers, and exceptions. AI knows none of that on day one.
Think in tasks, not roles: the three buckets
Most jobs are a bundle of tasks. AI is good at a slice of them: the repetitive, high-volume, rule-based parts. Hand those over and your people get hours back for the things AI can’t do. Judgment. Relationships. Selling. Solving the weird problem. Caring about the customer.
The practical way in: take one role, list what that person actually does in a week, and sort every task into one of three buckets.
flowchart TD
A(["One role, listed as weekly tasks"]) --> B{"Repetitive, high volume, clear right answer?"}
B -->|Yes| C(["Hand over: AI does it, a person spot-checks"])
B -->|Partly| D(["Share: AI drafts, the person decides"])
B -->|No| E(["Keep: judgment, relationships, exceptions"])
C --> F(["Same people, more of the work that grows the business"])
D --> F
E --> F
- Hand over. Data entry, first-draft replies to routine questions, copying figures between systems, filing. AI does it; a person reviews a sample.
- Share. Drafting a proposal, summarizing a long email thread, preparing the monthly report. AI produces the first version; the person edits and owns it.
- Keep. Negotiating with a supplier, handling an angry customer, deciding whether to extend credit, noticing that something doesn’t look right. These stay human, and they are where your margin lives.
Most roles come out with a real slice in “hand over” and the majority in “keep.” That is a role you upgrade, not one you delete.
The reality check: using AI and profiting from it are different things
Most companies already use AI in some form; far fewer see it move the bottom line. McKinsey’s State of AI found that only around 39% report an enterprise-level EBIT impact from generative AI, and only about 5 to 6% are genuine “AI high performers” (McKinsey, 2024).
Read that against the “cut staff” question. The gap isn’t between companies that cut and those that didn’t. It’s between companies that pointed AI at specific, measured work and companies that bought licenses and hoped. Headcount is the wrong lever; task selection is, and which work to automate first walks through it.
What this looks like in a Saudi business
- Customer support. An assistant answers the routine questions in Arabic and English (order status, opening hours, password resets) and hands the rest to a human with a summary attached. The team stops drowning in tier-one tickets and spends its time on the complex cases that build loyalty. Getting the Arabic right is its own job: how to use AI for customer service in Arabic and English.
- Admin and finance. Invoices, supplier statements, and Etimad tender documents get read into structured data automatically, and a person checks the exceptions. The same team closes the month faster and does analysis instead of keying.
- Sales. AI drafts the first version of proposals and follow-ups and keeps the CRM current. Your salespeople spend the recovered hours in front of customers, the only place they make you money.
One Saudi-specific point owners miss. The moment AI touches customer or employee data, you are inside the PDPL, which took effect 14 Sept 2023 with full compliance required by 14 Sept 2024 and SDAIA as the regulator (Morgan Lewis, 2024). A support bot that reads customer records is a data processing decision, not just a productivity one. Design for it up front.
A worked example: a ten-person support team
Say you run a ten-person customer support team. The numbers here are made up to show the shape. Each agent handles about 40 conversations a day, and when you sample a week of tickets, roughly 60% are routine: order status, delivery times, the same eight questions in different words. You put an assistant in front of the queue. It resolves most routine ones and hands the rest, plus every complaint and refund, to a person with a summary.
The “cut heads” version says: 60% of the work is gone, so cut six agents. Then the assistant misreads a complaint as a routine question, the four remaining agents are underwater, response times double, and the two who knew your product best have already left.
The “augment” version keeps the ten. Six move to the complex cases and to proactive work: calling customers whose orders are late before they call you, writing the help articles that stop routine questions arising, feeding what they hear to product and sales. Four keep the queue moving with the assistant on the front. You answer faster, handle more volume without hiring, and your best people do better work. Same payroll, bigger business.
The illustrative numbers can move. The pattern doesn’t: the saving shows up as growth you didn’t have to hire for, not as a smaller wage bill. We modeled the cost side in saving costs with AI.
Why “replace people to cut cost” usually backfires
A few hard reasons, then the honest caveat.
- AI makes mistakes, and only your people can catch them. It needs people who understand the work to notice when an answer is confidently wrong. Cut them and you’ve cut your safety net when you need it most.
- Your staff hold the knowledge and the relationships. The supplier who takes your call, the customer who trusts one specific agent, the reason a certain invoice is always paid late. None of that lives in a system, and once it walks out it’s slow and expensive to rebuild.
- The real prize isn’t doing the same work with fewer people. It’s doing much more and better work with the same people. One grows the business. The other makes it smaller.
- In the Kingdom, the talent math points the other way. AI is projected to contribute about $135bn to Saudi Arabia’s economy by 2030, the largest gain in the Middle East (PwC, via trade.gov). That is a growing economy where capable people who know your business are hard to hire back. Letting them go, then re-hiring at a premium when the growth arrives, is the expensive route.
- Saudization counts people, not productivity. A restructure that lets Saudi staff go first can change your Nitaqat classification, and what that costs in lost services is usually more than the salaries saved. Check it before you decide.
Now the caveat. For purely repetitive roles, yes, AI reduces how many people that specific work needs. If a role is almost all “hand over” tasks, it shrinks. The smart and usually more profitable move is to move those people, with training, into the “share” and “keep” work going unstaffed today. In a tight talent market, keeping capable people and making them more productive beats losing them and discovering later that you need them.
How to start: one team, one quarter, measured
The pattern that works is boring and reliable. Pick one team where the repetitive work is visible and measurable: usually support, admin, or finance. Do the three-bucket exercise on its roles with the people who hold them in the room; they know the tasks better than any consultant. Measure where the hours go today. Run a small, contained pilot on the “hand over” bucket, with a person reviewing the AI’s output. Then measure again: response times, throughput, error rate, and what the freed hours were spent on.
If the freed hours went into work that grew revenue or retention, you have your model and you scale it team by team. If they went nowhere, the problem is management, not AI, and one team is a cheap place to learn that. Tell your people the plan at the start, plainly: the point is to take the boring work off you, not to replace you. Teams that believe that adopt fast. Teams that don’t will make the tool fail.
Wondering how AI changes your team, not just your tools? SDCG helps owners use AI to make their people more productive instead of gambling on replacing them. We’ll map your roles into the three buckets, show where AI lifts your team and where humans stay in charge, and help you build the AI capability to run it. We’re independent, so there’s no product we’re quietly pushing. Book a free 30-minute review.