For ENSEK · what an agent swarm did to one company
We took ENSEK apart.
Then rebuilt it as working AI.
One company in. We harvested every live vacancy, piled on the public record — Companies House, competitors, customers, the regulator, the macro timeline — reverse-engineered the entire operating model, and distilled every role into what AI builds, what AI runs, and what stays human. All driven by an agent swarm. Everything below is what it produced — live, sourced, dated, not slides.
01 · What the system did
Company in. Operating model out.
No one filled in a form. An agent swarm ran the whole pipeline against ENSEK from the outside:
① Harvested the vacancies
Every live role from the careers site — the spring-2026 hiring wave of 25 advertised vacancies, plus historical posts. The job specs are the seed.
② Piled on the public record
Companies House filings, competitors, customers, the regulator and the macro timeline — 55 sources fused into one indexed knowledge base.
③ Reverse-engineered the org
Right-to-left from what ENSEK delivers, back through the meter-to-cash chain — 8 departments, the real named people, the board-vs-exec split.
④ Distilled every role
Each job split three ways: AI builds the system · AI where needed · human remains — graded honestly, role by role.
02 · The research it piled on
Not just the job specs — the whole picture.
Companies House
Centrica-owned since Sept 2024; a 9-director statutory board distinct from the operating C-suite; FY24 turnover £50.7m; ~474 staff. Primary filings, not guesswork.
Competitors
Kraken (Octopus) is the apex threat — a $8.65bn spin-out across 70M+ accounts; Kaluza and Gentrack close behind. ENSEK's edge: Margin Truth + UK DCC depth.
Customers & market
British Gas is the flagship — and the risk: ENSEK's own accounts state the majority of revenue comes from a single customer. A contestable market of ~22 suppliers.
Regulator & macro
Born of the 2021 energy crisis and the supplier-failure wave; the MHHS half-hourly-settlement cutover (2 Jul 2027) forces every supplier to re-platform — ENSEK's tailwind.
03 · What it produced — explore it
Real, sourced, dated, and running.
The 25 ai66 workspaces →
24 vacancy ai66s + 1 cross-domain proof = 25 workspaces — we call each an ai66. Every one decomposed and run live over a synthetic ENSEK-shaped dataset. Edit the SQL yourself.
The org chart →
Board vs operating exec, 8 departments, the 25 roles with residual headcount, and ~37 real named people.
IGNIS — the system itself →
We didn't stop at analysis. We built the operating core, AI-operated, with a live margin-truth dashboard.
The predicted relaunch →
The strategy + brand reveal we forecast from the signals — confidence-ranked, falsifier-flagged.
The live analyst →
Watch the agent answer in English, run real SQL, and write the decision-ready read.
Ask the Captain →
Ask anything about ENSEK — data, roles, competitors, the org. Grounded in the research, never made up.
The headline
25 advertised roles → ~19.5 you'd actually hire.
A ~22% headcount delta on this hiring wave — not a wipeout. The function doesn't go to zero; it changes shape. What remains is the irreducible human core: regulated sign-off, named relationships, accountability. We name every one.