The Anti-Consulting Manifesto.
PH—020
Twenty volumes. Thirty installs. Seven verticals. Here is what we have concluded. Enterprise AI is a software problem dressed as a consulting problem. The right answer is software. And the industry that got it wrong has five years to catch up — or watch a new kind of studio eat its lunch.
I What the Twenty Volumes Have Argued
Over the last twenty days we have published nineteen volumes of research and field reports, plus the volume that started the series. A short summary, in one paragraph, of the claim each made:
Vol 01: enterprises are not failing at AI. They are failing at AI delivery. Vol 02: the readiness score measures distance, not grade. Vols 03, 06, 09, 12, 15: four field reports of what happened when we installed on real customers. Vol 04 and 13: the per-industry ontology is the translator layer no generalist model contains. Vol 05: retrieval alone hits a ceiling; the agent starts where retrieval stops. Vol 07: the base model is not the moat. Four layers above it are. Vol 08: agents must be inspectable to be operable. Vol 10: context engineering is the practice, not prompt engineering. Vol 11: build-hire-automate is a timeline problem, not a cost problem. Vol 14: retraining without regression is how the moat widens quarterly. Vol 16: compliance is a recommender filter, not a gate. Vol 17: invisibility for mechanics, never for responsibility. Vol 18: the base is swappable because we designed it to be. Vol 19: twelve questions that separate software you can keep from software that leaves.
Underneath each argument is the same claim: this is what enterprise AI looks like when it is treated as software engineering rather than as consulting.
II The Consulting Legacy (and Why It’s Over)
The consulting industry shaped how enterprises buy expertise for a hundred years. The artifact was a deck. The price was time-and-materials. The deliverable ended on the last day of the engagement. Whether the advice worked, whether it was implemented, whether the organization changed as a result — none of these were the consultant’s problem after the invoice.
This structure worked for a while. It worked because expertise was the scarce resource. The consultant knew something the organization did not, communicated it, and left. The transfer was the work.
Expertise, as a scarce resource, has collapsed. Not because experts disappeared. Because the capability an expert used to embody is now in software. The 2026 corpus contains the commercialization wisdom that fifty years of McKinsey engagements produced, compressed and shipping. An ontology encoded by three subject-matter experts and ten installs is better than a generalist consultant’s advice at month three of a project. This is not a flex. It is an architectural fact.
Consulting, as structured today, is an expensive way to rent the wrong artifact.
III The New Kind of Studio
A studio that installs software, rather than delivers strategy, operates differently in five ways.
First, the deliverable is software. The engagement ends when the software is running, not when the deck is presented.
Second, the engagement compounds. Every install teaches the methodology. Consulting casebooks catch up after the fact; software methodology catches up during the work.
Third, the pricing is transparent. Install price is fixed. Monthly license is predictable. No change-orders. No scope creep. No billable hours.
Fourth, the buyer owns the asset. Perpetual license. Runs in their tenancy. Reversible in days, not months.
Fifth, the studio publishes. Not marketing content. Research. What we learned this install, at this customer, on this ontology. We publish it so the next install is sharper and so buyers can see the work before they sign.
IV What This Requires of Buyers
The shift is not free for buyers. It requires three changes in how they procure.
One, procure software, not advice. If the artifact you are about to sign for is a deck, you are buying the wrong thing for this decade.
Two, read the methodology before the sales deck. Any studio worth installing publishes what it knows. Read the research. If the research is thin, the methodology is thin.
Three, staff an operator, not a program manager. The new artifact requires a human who runs it, not a human who coordinates its delivery. Different hire. Different org chart.
Buyers who make these three moves stop buying pilots and start buying installs. Their AI programs stop presenting and start running. This is not a rhetorical shift. It is a procurement discipline.
V What This Requires of Us
The shift is not free for us either. It requires us to ship software that actually works, on a cadence that actually compounds, with research that actually teaches. Every promise we make is visible in our install base and in our publication history. Neither is something a marketing team can fabricate.
We publish these twenty volumes because the commitment is visible now. Any buyer considering Terrein can read the nineteen volumes before this one and see whether the methodology is real. They can look at the sector catalog and see twelve named clients. They can read the field reports and see what held, what broke, what we’d do differently.
The discipline is public. That is the point.
VI Coda · The Next Twenty Volumes
We will publish one volume per Wednesday, continuing from the close of this series, for the next twenty weeks. The titles are in the research index. Healthcare commercialization will be a three-volume arc. Two more field reports from installs in progress will ship before the end of Q2. The 2026.Q3 brain update will ship in July.
What we will not do: publish filler to meet a cadence. If an install is still in flight, we wait until it is done. If a methodology is still being argued internally, we argue it until there is a stance. What goes out must be earned. Twenty volumes in twenty days set the register. Every week after that continues it.
Enterprise AI is a software problem. You cannot consult your way through it. You cannot pilot your way through it. You cannot workshop your way through it.
You install it.