In manufacturing, there's a concept called "lights out manufacturing". It means the factory runs around the clock with no one there. No lighting needed. The robots see in the dark.
That concept is now moving into e-commerce. And it can transform how you run your administrative processes.
Your processes cost more than they should
Think about what happens every day in your organization. Supplier invoices are matched against purchase orders. Product information is rewritten, adapted and published across channels. The procurement team compares prices, negotiates terms and updates agreements. Returns are processed, credited and recorded. A/B tests are set up, analyzed and reported.
Every process requires people. Every person costs salary, office space, management and onboarding. And every time someone leaves, you start over. You know what it costs, but you've accepted it as a fact of life.
It no longer is.
What has changed
AI agents can now perform administrative tasks at a quality that matches or exceeds manual work. Not as an experiment. Not as a demo. In production.
In software development, there are already companies where three people run what previously required teams of thirty. Not by working harder, but by letting AI agents do the execution while people steer and ensure quality.
The same principle applies to your processes. The difference is that the cost of executing a task with AI drops every month, while your salary costs rise.
Five processes you can outsource today
Supplier invoices. The invoice arrives. The agent matches it against purchase orders and contracts, verifies amounts and VAT, assigns accounts and books it. If everything checks out, the invoice goes straight into your accounting system. If something deviates, it's flagged for review. You pay per correctly handled invoice, not per hour.
Procurement and supplier monitoring. The agent monitors contract prices, compares against market rates, flags deviations and prepares briefs ahead of negotiations. Instead of a buyer spending half their time gathering data, they get ready-made decision support every morning.
Product information. You have hundreds or thousands of items that need descriptions, attributes and images adapted for each channel. The agent fetches data from suppliers, enriches, translates and publishes. A process that currently takes weeks can run in hours, with consistent quality.
A/B testing and optimization. The agent sets up tests, monitors results, analyzes statistical significance and delivers recommendations. No one waits for the "analytics person" to have time anymore. Tests run continuously and you get reports with clear action items.
Returns and claims. A customer returns an item. The agent receives the case, assesses whether it meets the return policy, creates a return label, issues credit and updates inventory. A case that currently requires three manual steps and twelve minutes takes seconds.
How it works in practice
We start by mapping your processes. Not all at once, but those that consume the most time relative to the value they create.
Then we build a digital twin of the process. A copy of your flow where the agent can work without touching real data. We test until quality is proven.
Only then do we connect to production. The agent works in your real system. Every transaction is automatically verified before it's executed. You get SLA on quality and throughput.
What you don't need to do: recruit, train, manage or replace staff for these processes.
What you save
Savings vary by process, but the pattern is clear.
Direct cost reduction. A process that currently requires five full-time employees can often be handled at a fraction of the cost. You pay for completed transactions, not for occupied desks.
Eliminated staff turnover. Every time an administrator leaves, you lose three to six months in onboarding. With Lights Out, that risk disappears entirely.
Scalability without hiring. Black Friday, seasonal fluctuations, expansion to new markets. You scale up and down without hiring or laying off.
Consistent quality. The agent doesn't make mistakes from fatigue, stress or carelessness. Every transaction follows exactly the same process, every time.
Freed-up time for what creates value. Your skilled employees stop entering data and start focusing on strategy, negotiations and customer relationships.
Why not do it yourself?
You can. But it's harder than it looks.
Automating a process with AI requires specifying every step, every exception and every quality check with a precision that most organizations lack. It requires systems competence closer to engineering than administration.
On top of that, your systems are rarely ready. Fifteen years of customizations have created flows that no one has fully documented. Before an agent can take over, someone needs to map what actually happens today. That work is thorough and time-consuming.
We've done it before. We have the tools, the methodology and the experience.
The shift
This isn't about making administration slightly cheaper. It's about making it error-free, infinitely scalable and available around the clock.
Input is your processes as they look today. Output is structured, verified data in your own systems. With SLA.
The question isn't whether your processes can be automated. The question is how long you can afford to wait.
Want to know what Lights Out can save you? Get in touch with Spinout.
Sources
StrongDM Dark Factory model (3 engineers, zero manual code review)
Token Economics: Jevons Paradox applied to intelligence
Lights Out Manufacturing as a concept applied to the service sector
Nate Herk, Dark Factories: The AI Business Model That Changes Everything