Your forecast isn't the bottleneck anymore. Your decision loop is.
For retail / FMCG / logistics executives — stop buying prediction and start owning the decision loop.
Read articleField notes / Valnative
Decisions worth
operating.
Evidence-bounded essays on decision intelligence, durable execution, and the operating systems around applied AI.
RSS feed20 long reads · 4 notes
Operstead
Enterprises have agents. Almost none have outcomes — and a smarter model won't fix it.
For executives sponsoring agent programmes — stop buying capability demos and start buying production survival.
Read article
You already have more agents than you can see. Scaling them is an operating-model problem, not a capability one.
For CIOs, CTOs and COOs sponsoring or inheriting agent programmes — count what you already have, then govern it.
Read article
Cost per task is not cost per outcome
For CFOs and CIOs — price the outcome, not the attempt.
Read article
The triangle did not disappear. It moved.
Why rare people are expensive, stated properly The usual explanation is that talent is scarce.
Read article
The score belongs to the pairing, not the model
The two numbers that sit side by side The cleanest result comes from Claw-SWE-Bench, published in June: 350 real GitHub issue-resolution tasks across eight programming languages and 43 repositories, with nine models swept against five variants of the same agent framework.
Read articleSHEPORD
Don't ask an LLM to make the decision. Formalize it.
For retail leaders and their data/AI owners — stop trying to make the model decide, and start formalizing the decision.
Read article
83% see the potential. 3% feel ready. Your AI ROI is capped by readiness, not the model.
For retail leaders and their data owners — stop measuring AI by adoption and spend, and start measuring it by readiness-to-decide.
Read articleReliable enterprise AI is a closed loop, not a shopping list
For leaders scaling AI — stop scoring maturity by tool count and start scoring it by loop closure.
Read article
Human-in-the-loop is a perishable control
For COO / risk leaders — your strongest AI control may be decaying without any signal reaching you.
Read article
The retail AI prize is on the shelf, not at the storefront.
For retail chain and FMCG supplier executives: (1) retail's AI divide is not a technology-access gap, most of the same tools are available to slow and fast movers alike; it is an investment-priority gap between where attention goes (customer-facing layers still 'too early to tell' on value) and wher
Read article
Copilot Is Not a Stage of Maturity. It's a Wiring Problem.
For COO / CDO leaders in retail and CPG: (1) the copilot plateau BCG describes (67% still approving every AI step, only 9% in autopilot, from a 39-executive survey) is not a sign of caution paying off; it is a sign autonomy was assigned to the wrong noun.
Read article
The agent that won wasn't the one that knew more. It was the one that asked.
For retail and FMCG leaders — stop trying to fix decision quality by giving the model more knowledge, and start specifying what each decision is not allowed to be made without.
Read article
Groupthink used to need a group
For COO and decision owners — AI does not remove group bias from decisions, it reproduces the worst dynamic of a meeting more cheaply and less visibly.
Read article
The pyramid did not disappear. We stopped seeing its base.
Which pyramid we are talking about Two structures used to sit almost exactly on top of each other, and it matters that they have come apart.
Read article
The remainder is harder: what automation leaves behind, and what moving people costs
What actually happened at one company Ingka Group, which operates most IKEA stores, has a customer service bot named after the Billy bookcase.
Read article
The review desk: why checking costs what it costs
The diagnosis, stated fairly The argument runs like this.
Read article
The order was right. The shelf was empty.
A counting result that should change what you fund In a European grocery chain, researchers ran a field quasi-experiment.
Read article
Approving what you cannot restate
What the experiment actually did In October 2025, Ivan Aslanov, Patricio Felmer and Ernesto Guerra ran a study with 102 participants, split into three groups.
Read articleSHEPORD + Operstead
A working glossary for agent operations
The words that decide whether automation is safe to run are absent from the vendor conversation.
Read articleField notes
-
Your organisation has always run on pass^k
For COO/CIO — stop evaluating agents on demos (pass@1) and price supervision like you already do for people: score the chain not the step, put checks where error cost × observed inconsistency is highest, relax supervision as pass^k accumulates, keep blind spot-checks after autonomy.
-
Your agent didn't get dumber. Its window did.
For COO/CIO — when an agent "gets worse", the cheap and usually correct first move is window inspection, not model replacement: capability gains do not buy consistency (Princeton 2026), window discipline is a property of the harness you run and fully under your control; require that the team can rep
-
An agent's context is assembled, not accumulated
For COO/CIO — ask whether your agent stack has a POLICY for three flows (what enters the window, what leaves it, what gets remembered and in which kind of memory) or whether all three happen by accident; the four memory kinds need different handling, and dumping everything into chat history is the d
-
The thirty seconds before the prompt
I had a very good week recently.