Calculating the Cost and ROI of AI Projects

Why Start With the Return Question, Not the Technology Question?

Artificial intelligence has moved from a conference buzzword to an operational tool that now touches customer service, sales, accounting, and inventory. With so many offers around, many business owners fall into the wrong question — 'What is the newest AI tool?' — instead of the right one: 'Which problem is costing me money or time, and does AI solve it with a return that beats its cost?'

An AI project is not a subscription purchase; it is an investment with a measurable cost and a measurable return. This guide shows you how to calculate both before you spend a single riyal, so your decision rests on numbers rather than enthusiasm.

The Real Cost Components

The figure in the first quote is not the whole story. An AI project has six main cost items:

• Discovery and design: understanding the problem and defining the metrics and data sources before any code is written. Skipping this is the single biggest cause of failed projects.
• Data preparation: AI feeds on your data. Cleaning, organizing, and connecting it can consume a large share of the budget, especially when it is scattered across files and disconnected systems.
• Development and integration: building the solution and wiring it into your existing systems (store, accounting, CRM) so it works inside your workflow, not beside it.
• Running and subscriptions: the cost of models, APIs, or servers — a recurring monthly cost that scales with usage.
• Adoption and training: the best solution fails if your team does not use it. Staff training and change management are a real line item, not a luxury.
• Maintenance and iteration: models need monitoring and updating as your data and market shift. Budget for a full year, not just launch.

Where Does the Return Come From?

Return comes in two forms: tangible return that is easy to measure in riyals, and intangible return that shows over time. Anchor your decision on the first and treat the second as a bonus.

• Time and cost savings: automating repetitive tasks cuts work hours and reduces the need for extra hiring as demand grows.
• Fewer errors: human error in invoices, orders, or inventory costs real money; reducing it is a direct return.
• More sales: faster responses, sharper recommendations, and follow-up that never sleeps all lift conversion.
• Faster, better decisions: real-time data analysis surfaces opportunities and risks before your competitor sees them.

A Simple Formula for Return

You do not need a complex financial model to start. Two indicators are enough:

1. Return on investment: ROI = (annual net benefit ÷ total cost) × 100. If a solution saves SAR 120,000 a year and costs SAR 60,000, the return is 200% in the first year.
2. Payback period: how many months the project needs to cover its cost from its savings. The shorter it is, the lower the risk.

The practical rule: start with a project whose payback is under a year and whose return metric can be measured before and after.

An Illustrative Example

Imagine a store receiving hundreds of messages a day and staffing three customer-service agents. (The numbers here are illustrative, to keep the idea simple.)

• The cost: a smart assistant that answers repetitive questions and routes complex ones to a human, built once within Origami's mid-tier project range (SAR 30,000–80,000), plus a modest monthly running cost.
• The return: automating 60% of messages frees a full agent for higher-value work and speeds up replies, lifting order completion.
• The outcome: if the time saved and the added sales equal an annual saving above the cost, the project pays for itself within months, and beyond that it is net return.

Common Mistakes in Estimating Return

• Ignoring maintenance cost: counting launch only and forgetting a full year of running and iteration.
• No baseline: without measuring the current state (time, cost, errors) you cannot prove any improvement.
• Over-ambition: trying to solve everything at once instead of one project with a clear return.
• Chasing hype: adopting a tool because it is trending, not because it solves a genuinely costly problem.

How to Start With the Least Risk

The safest path is to begin with a single, clear, measurable use case: pick a painful, repetitive process, measure its current state, launch a limited first version, then scale based on results rather than assumptions. That turns risk from a big bet into a calculated step.

Origami's Role

At Origami we are a technology company that builds AI solutions and connects them to your real systems — but we always start with the financial question: what is the problem, what does it cost you today, and how do we measure the improvement? We help you define the metrics and the payback period before building, so you know your return before you commit to it. Our projects range from simple solutions (SAR 15,000–30,000) to mid-tier (SAR 30,000–80,000) to complex (SAR 80,000 and above) depending on scope.

Successful AI is not the most technically clever; it is the clearest in return.

Sources

• Saudi Data and AI Authority (SDAIA) — National Strategy for Data and AI: sdaia.gov.sa
• Saudi Vision 2030 — Digital Economy: vision2030.gov.sa