Insights
Thinking from the WAJD team
Papers on the decisions that are genuinely hard: how much of your supply chain you can actually see, whether your resilience plan would survive being tested, and what a digital twin has to do to be worth trusting. Written for the person who has to make the call, not for a search engine.
Cyber-physical defence
Knowing the asset is vulnerable is not the same as knowing what happens.
Industrial security answers whether an attacker can reach a controller. It rarely answers what the controller would then do to the process. A proposed architecture for closing that gap with small models that run inside the plant.
AI as industrial infrastructure
The model is rarely the thing holding you.
AI now depends on organisations, robots, accelerators, substations and public consent. Six layers, and the one that is binding is almost never the one being bought.
Autotwins
Twins that build themselves, and act.
A twin nobody maintains is a twin nobody trusts. What it takes to make one assemble itself, prove its own accuracy, and carry agents that can act inside guardrails.
Autotwins · 12 min read
Autotwins
Most digital twins die of maintenance. The three layers of a twin that assembles itself from your systems of record, measures its own fidelity, and carries agents that act within limits you set.
Read the paperAgentic AI · 11 min read · New
The backbone, the boost and the reins
What an agentic backbone is actually for, what 90 days genuinely buys a manufacturer, and the corridor question answered honestly: if agent fleets start reasoning for themselves, what could go wrong, and do we have a stop button?
Read the paperAgentic AI · 4 min teaser · full paper by email
Page not found
When an agent stands between people and records, history becomes whatever the agent serves. Read the opening argument, answer one question, and get the full paper, with the witness log, the factory gate and the code of practice, by email.
Read the paperAgentic AI · 9 min read · PDF
The rules are the asset
An agent accumulates institutional memory, and almost nobody backs it up or can say who may change the rules it works inside. Writing ground rules that survive a refactor, encoding them as tests, and putting recovery and audit around all of it.
Read the paperAgentic AI · 9 min read
Agentic AI in operations
From dashboards to decisions: software that perceives, reasons and acts on the operation, safely, with guardrails and people in control.
Read the paperYour topic
Want our view on something?
If there is a decision you are weighing, we will write the paper we wish we had read.
Suggest a topicSupplier risk and resilience
The exposure that stops you is rarely on the register.
Finding the concentration hiding below tier one, keeping supplier risk live rather than annual, and testing resilience hard enough that the test can fail.
Risk & resilience · 9 min read · PDF
The 95 risks and the supply chain
The National Risk Register 2026 gives you 95 fully specified scenarios. Almost none land on you directly: they arrive through a supplier. How to hold all 95 against a supplier book, and answer the only question that matters.
Read the paperRisk & resilience · 10 min read
Two suppliers, one factory
Why dual sourcing at tier one is often an illusion, how sub-tier convergence is actually discovered, and why ranking suppliers by spend hides your worst exposures.
Read the paperRisk & resilience · 10 min read
Severe but plausible
A continuity plan that has never failed a test has never been tested. Impact tolerances that bite, dependency maps that stay true, and scenarios designed to break something.
Read the paperRisk & resilience · 8 min read
Continuous third-party risk
Why point-in-time supplier checks fail, and what continuous, AI-assisted monitoring looks like across cyber, financial and ESG signals.
Read the paperManufacturing · 8 min read
Engineering manufacturing resilience
Treating resilience as something to engineer: early warning from equipment, supply and quality signals that keeps the line running.
Read the paperManufacturing · M&A · 9 min read
The deal is won or lost at integration
83% of failed deals point to integration. Why technology and security leaders belong in the deal from the start, and how one operational backbone turns diligence and day one into engineering.
Read the paperData foundations
None of it works without the backbone.
Twins, agents and live risk monitoring all rest on the same thing: data that arrives reliably, means what it says, and can be observed when it does not.
Data & AI · 12 min read
Building the digital backbone
Moving to a Databricks-centred lakehouse, joined by middleware and observed with Grafana, leading on the data migration path and agentic AI.
Recognise your problem in one of these?
Tell us where you are stuck. We will tell you what we would do about it, whether or not you engage us.
Start a conversation