The 'feed the cats' irreducible core
The point isn't full automation.
It's knowing exactly where a human must stay.
A system that automates everything and is accountable for nothing is a liability — especially in regulated energy retail. This design is built around its hand-offs. Here is every control point, and the reason it's a person's job.
Three reasons a human stays in the loop
Governance · Accountability · Edge cases.
Governance
Someone must own the canonical metric definitions and sign off what becomes org-wide truth. The AI proposes and enforces consistency; a person is accountable for the standard.
Accountability
When a bill is corrected, a refund issued, or an account chased, ENSEK answers to the customer and to Ofgem. A named human makes that call — the AI only detects and quantifies.
Edge cases
Vulnerability, affordability, ambiguous exec asks, experimental validity — the long tail where context and judgement beat pattern-matching. Routed to a person by design.
Every control point in the system
Where the AI stops and a person owns it.
| AI does | → Hands to a human for | Because |
|---|---|---|
| Computes the canonical revenue / consumption series | Sign-off on the metric definition | One org-wide source of truth needs a named owner. |
| Flags 100+ billing discrepancies with £ variance & probable cause | Approving each customer-facing correction or refund | Regulatory accuracy + customer accountability. |
| Builds the persistently-estimated-read cohort | Deciding the customer treatment | Vulnerability & affordability rules apply. |
| Reconstructs the debt book & early-arrears list | All collections / dunning actions | Consumer-protection & FCA-adjacent obligations — never automate the action. |
| Sets up & computes A/B tests | The validity judgement & ship/no-ship call | Experimental rigour and commercial consequence. |
| Drafts the decision-ready report | The interpretation that leaves the building | The analyst's name is on the recommendation. |
| (nothing) | Framing ambiguous exec problems & coaching the team | Org context, politics, mentorship — fully human. |
On the live analyst, every result already carries its control point inline — the hand-off is a first-class part of the output, not an afterthought.
The residual org
What ENSEK would actually still hire.
The recurring-production work — write the SQL, refresh the dashboard, sample-audit bills — is absorbed by the agentic backend. What remains is accountability and judgement. Hire for that.
| Role | Heads | Mandate |
|---|---|---|
| Analytics lead / owner | 1 | Metric governance, sign-off, accountable for what ships. The human core. |
| Billing / assurance analyst | 1–2 | Adjudicates flagged discrepancies, owns corrections & regulator-facing accuracy. |
| Data engineer | 1 | Owns pipelines & tariff/reference data. Garbage in still means garbage out. |
| Senior / principal analyst | 1 | Problem framing, experiment validity, exec partnership, coaching. |
The honest read: the open Senior Data Analyst seat is best filled by someone who runs this system and owns its hand-offs — not someone who races it on query throughput. That's the role we built for.