INTEL / AI LEXICON

What we learn, in public.

Running notes from building and operating AI systems in production. Updated weekly. Filter by topic, industry or format.

INTEL · UPDATED WEEKLY · 21 ENTRIES
TOPIC
INDUSTRY
FORMAT
METHOD·FINANCE

Transaction monitoring: rules engine, agent, and audit trail

Rules alone produce too much noise. Pure ML alone gives no explanation. How to combine them with an agent that keeps the auditor's shoulders relaxed.

UPDATED 2026-04-299 min
AI IN OPERATIONS·INDUSTRY

Sensor data to work order: a pragmatic pipeline

Industrial sensors produce data every day. The data is rarely the problem. The flow from signal to work order with clear ownership and a deadline is.

UPDATED 2026-04-298 min
DOCUMENTS·CONSTRUCTION

The RFI flow that answers in time

The biggest losses on construction projects aren't materials. They're RFIs that take 14 days to answer when they should take two. An operational flow for contractors.

UPDATED 2026-04-297 min
METHOD·PUBLIC SECTOR

Public-sector casework with AI under transparency rules

Public administration has to be auditable. What survives a freedom-of-information request, what doesn't, and how an agent-driven case flow holds both at the same time.

UPDATED 2026-04-299 min
METHOD

What is an operating AI platform? A pragmatic definition

The term gets used a lot, defined rarely. The difference between an operating AI platform, a chatbot, and a SaaS system, in operator terms.

UPDATED 2026-04-295 min
TOOL REVIEW

Observatory: what we track and why

A tool review of Observatory on an agent fleet in production. Which signals we actually read every day, which we ignore, and how we set thresholds that don't create alert noise.

UPDATED 2026-04-297 min
METHOD

How to measure AI ROI without guessing

A concrete framework with formulas, metrics and pitfalls to make AI ROI measurable in practice. Not just prettier presentations.

UPDATED 2026-03-296 min
DOCUMENTS·REAL ESTATE

Lease contracts agents can actually read

A portfolio of 800 leases is a small Excel hell with high operational risk. How to structure them once, so agents can answer the questions that matter.

UPDATED 2026-04-297 min
AI IN OPERATIONS·FINANCE

KYC in 2026: what agents handle, what stays human

KYC has become an operating flow, not a checklist. What agents can take, what humans must own, and how to move the line without losing control.

UPDATED 2026-04-298 min
DOCUMENTS·ENERGY

Incident reporting that doesn't cost a week

Each incident at a grid operator triggers three reports, five information requests and two weeks of backlog. How to structure the flow so the facts are there when someone asks.

UPDATED 2026-04-295 min
AI IN OPERATIONS·LOGISTICS

Exception handling without phoning anyone

Delays and damage are normal flow in logistics. The cost is in how they're handled. An operational flow that lets agents catch and structure exceptions before they escalate.

UPDATED 2026-04-297 min
METHOD

DPA, DPIA and data residency: pragmatic Nordic approach

Privacy becomes an obstacle only if you treat it as an afterthought. A pragmatic walkthrough of DPA, DPIA and data residency for AI projects in 2026.

UPDATED 2026-04-298 min
DOCUMENTS·LEGAL

Contract review with agents: what still needs human eyes

Agents read contracts faster than lawyers. They don't do the job alone. How to build a review flow that holds up under pressure.

UPDATED 2026-04-298 min
DOCUMENTS·HEALTH

Clinical notes as structured data, without replacing the EHR

Your EHR isn't being replaced this year. How to extract structured data from clinical notes anyway, and let it flow into quality registries, reporting and downstream care.

UPDATED 2026-04-298 min
METHOD

Build vs buy for AI infrastructure: 5 questions

Most companies should neither build everything nor buy everything. Five questions that help you land the mix that actually fits your operation.

UPDATED 2026-04-295 min
TOOL REVIEW

Bridge in practice: typifying 50,000 documents

A tool review of Bridge against a realistic document pile. What worked out of the box, what required tuning, and what still needs human judgement.

UPDATED 2026-04-297 min
DOCUMENTS

From chaos to system: document handling with AI in practice

Document flow is the bottleneck nobody owns but everyone feels. Concrete workflows for invoices, contracts and email, without major rebuilds.

UPDATED 2026-04-197 min
METHOD

Command Center: steer the company from one surface

An operations dashboard that puts speed into work every day. The control plane over your agents and systems, on mobile and desktop.

UPDATED 2026-04-195 min
DOCUMENTS·REAL ESTATE

Automation in property management: time and cost saved with AI

Only 14 percent of Norwegian property operators have invested in AI. Five processes you can automate today, and what to expect to save.

UPDATED 2026-04-199 min
AI IN OPERATIONS

AI agents around the clock: how they actually work

AI agents aren't smarter chatbots. They're working flows that actually do the work. Concrete examples from sales, customer service and operations.

UPDATED 2026-04-198 min
AI IN OPERATIONS

Five signs your operation is ready for AI

Don't wait until everything is perfect. The question is whether you're ready enough to start in a controlled way. Five clear signs and a decision framework.

UPDATED 2026-04-195 min