PILLAR · 19 May 2026
The new equation: build your own, don't rent SaaS
E-commerce has grown up on a decade of SaaS. That model was rational when it took shape. Then it flipped. A story about why custom software is now both faster and cheaper, and how lights-out.ai makes it possible with minimum technical debt.
Read article →
AGENT OPS · 10 June 2026
Steer or dispatch: the two modes of working with agents
A debate between two coding tools has named something most teams have been struggling to articulate. The question is not which AI is smarter. The question is whether you are steering, or dispatching.
Read article →
E-COM OPS · 22 May 2026
Your inventory numbers don't match. Again.
Four systems, four inventory counts, zero agreement. The real cost of manual reconciliation and what an operational layer above your systems looks like.
Read article →
E-COM OPS · 22 May 2026
14 SaaS tools. Nobody sees the full picture.
You spend 200-400k SEK monthly on SaaS. Each tool works. The gap between them doesn't. How an operational layer gives your team one shared view.
Read article →
E-COM OPS · 22 May 2026
Wrong market, wrong price, wrong description
Your PIM handles the standard case. The problem is exceptions. How automated quality scoring and AI-generated content cut copywriter time by 80%.
Read article →
E-COM OPS · 22 May 2026
Your VC money is leaking through workarounds
Operational debt grows silently in scaling companies. Every workaround has four cost layers, and most companies only see the first one.
Read article →
E-COM OPS · 22 May 2026
Returns, complaints, and the hidden margin killer
A return takes a customer 90 seconds. Inside your company, it triggers a chain across five systems and three people. The cost per return is 3-5x higher than it needs to be.
Read article →
E-COM OPS · 22 May 2026
Nobody wants another dashboard
Dashboards tell you what happened. Command centers tell you what to do next. Why your team has dashboard fatigue and what replaces it.
Read article →
ROADMAP (FICTITIOUS) · Oct 2026
100 % Swedish: we closed the last back door
A vision we are building towards. Our platform on Swedish infrastructure, Swedish AI inference and a security-classified agent runtime. No data leaves the country.
Read article →
21 May 2026
What does your back office actually cost?
Your CFO knows revenue per order. But what does each invoice, each product update and each return cost in admin? Here is the maths.
Read article →
17 May 2026
Who is liable when the agent gets it wrong?
An AI chatbot promised a customer a discount that didn't exist. The airline had to pay. Who pays when it happens to you?
Read article →
14 May 2026
Stop piloting. Start operating.
78% of enterprises are running AI agent pilots. 14% have reached production. The problem is not the technology.
Read article →
7 May 2026
Dark Factory: from shop floor to back office
FANUC's factory in Japan has been running without lights since 2001. Now the same logic is moving into e-commerce back offices.
Read article →
3 May 2026
Why 95% of AI pilots die
Of $684 billion invested in AI during 2025, $547 billion produced zero measurable results. The pattern in what the 5% who succeed do differently.
Read article →
30 Apr 2026
GDPR, customer data and AI agents: the questions e-commerce leaders must ask
Before e-commerce companies let AI into customer data and operations flows, clear answers on personal data, sub-processors, logging and accountability are needed.
Read article →
23 Apr 2026
Entry, Standard, Enterprise: three steps from proof of value to managed automation
How the three tiers of lights-out.ai work: Entry for first value, Standard for SaaS replacement and Enterprise for agent-driven operations.
Read article →
12 Apr 2026
From PIM chaos to a product data agent in six weeks
A concrete example of how an e-commerce company can go from manual product data management to a controlled agent-supported workspace in 4 to 8 weeks.
Read article →
5 Apr 2026
AI agent or plain automation? How to choose the right one
Not everything needs to be an AI agent. A practical guide to when e-commerce companies should use deterministic workflows, copilots or autonomous agents.
Read article →
29 Mar 2026
The command centre: why e-commerce needs an operational layer on top of SaaS
E-commerce companies do not always need more systems. Often they need an operational layer that connects SaaS, data and decisions.
Read article →
22 Mar 2026
When should you replace SaaS, and when should you leave it alone?
Not all SaaS should be replaced. Here is a practical model for when e-commerce companies should build their own, keep standard tools or add an operational layer on top.
Read article →
15 Mar 2026
Excel is your most expensive system
Most e-commerce companies have a hidden system landscape in Excel. Here is why reports, exports and copy-paste often cost more than the SaaS licences.
Read article →
19 Mar 2026
Meet Ember and Umbra: the agents in the platform
Two agents that come with the platform and are activated in your own lights-out.ai instance. They work around the clock with clear boundaries and full traceability.
Read article →
25 Feb 2026
Intent engineering: the missing layer in your AI strategy
The third discipline of enterprise AI. Why business intent is what actually determines whether AI delivers value or just shifts cost.
Read article →
24 Feb 2025
Why 95 percent of AI projects fail
Morningside AI has implemented AI for some of the world's biggest brands. Their conclusion: companies do it in the wrong order. Here is what the five percent who succeed do differently.
Read article →
12 Feb 2025
lights-out.ai for e-commerce: five areas where the platform creates value
Five areas in a typical e-commerce setup where SafeZone, GuardRails, and FastTrack combined with the three levels create concrete value.
Read article →
9 Feb 2025
AI archaeology: the bottleneck behind failed automation
How to break brownfield lock-in and build executable specifications for reliable agent operations.
Read article →
2 Feb 2025
From outsourcing to intelligence arbitrage
Why value moves from wages to tokens and what that means when the company builds its own on the platform.
Read article →