Human involvement map · Document comparison
Who owns the next action?
Three energy complaint procedures, five stages of useful help—and the difference between a published promise and proof.
Procedures checked 14 November 2026 · Reporting checked 15 November 2026 · Human by choice — a Policy Window advocacy programme
Purposive review of three domestic energy complaint-procedure document sets inspected 14 November 2026, supplemented by selected Q2 2026 performance tables and reporting guidance inspected 15 November. Descriptions and research questions, not supplier rankings, current legal advice or independently verified performance.
Read a service promise as a promise
A complaint policy can tell you who should help and what should happen next. It cannot, on its own, show what happened to a disputed bill. This desk review compares three published supplier procedures against HBC’s five-stage workflow. It finds useful commitments to investigate, preserve responsibility and seek a remedy; the next question is how to test their delivery.
British Gas, Octopus Energy and EDF were selected before coding for a bounded document review, not as a representative sample or a supplier ranking. The comparison concerns household energy-service complaints relevant to a billing dispute, not all customer support or other installed products. Each cell records a relevant published statement. None is a rating of whether the supplier delivers the whole stage.
Published wording, delivery unverified. Each card describes part of a procedure. A relevant statement does not show that every question for that stage is answered.
Stage 1 · HBC-W001-S1
Make the concern understood
Can someone explain the problem through a usable channel?
British Gas
Offers chat, telephone, online and post; written or online complaints have a two-working-day response aim.
PDF p. 2, contact section. Source BG
Octopus Energy
Offers email, telephone and post; asks about the impact and preferred resolution.
Step One, contact and background information. Source OE
EDF
Offers email, telephone and post; the booklet offers alternative formats or languages on request.
Web Step 1; booklet p. 1, footer. Source EDF
What remains to test
A listed channel does not establish usable access for someone facing language, disability or digital barriers. None of these entries identifies who actually answered a particular contact.
Delivery test: Test completion of an explanation through each relevant channel, with needed adjustments; record failed attempts and repeat explanations.
Stage 2 · HBC-W001-S2
Investigate the disputed issue
Who can examine the evidence and change an incorrect outcome?
British Gas
Promises investigation and an explanation of the next action; advisers cover billing and metering.
PDF pp. 2–3, investigation and adviser descriptions. Source BG
Octopus Energy
Describes trained specialists with management-level decision powers, including refunds and rebilling.
Step One, specialist training and powers. Source OE
EDF
Describes investigation, updates and agreement on a solution; specialists can arrange reassessment.
Web introduction and Step 2; booklet Step 2. Source EDF
What remains to test
A promise to investigate is different from a specified decision power. This review does not establish staff permissions, access to evidence, caseloads or time available in real cases.
Delivery test: For a contested bill, establish what the reviewer can inspect, correct or escalate, and whether a reasoned challenge can change the result.
Stage 3 · HBC-W001-S3
Own the response and escalation
Does responsibility survive a transfer or absence?
British Gas
Offers escalation to a complaints manager, with direct contact within two working days.
PDF p. 3, still-unhappy section. Source BG
Octopus Energy
Makes the specialist raising the complaint responsible throughout; further specialist review is available.
Step One, ownership and unresolved-complaint sections. Source OE
EDF
At escalation, email reconnects the same specialist; callers may reach a knowledgeable teammate. The webpage stresses team continuity; the booklet names the first specialist.
Web Step 2; booklet Step 2. Source EDF
What remains to test
A response deadline, a named owner and a continuing team are different commitments. A named owner may be unavailable; a team may lose the thread. Their relative effectiveness is untested here.
Delivery test: Track responsibility, missed actions and repeated explanations across transfers, staff absence and escalation. Test a supported team handover alongside named ownership.
Stage 4 · HBC-W001-S4
Obtain independent consideration
Can an unresolved case enter a usable external route?
British Gas
Describes free, independent Ombudsman consideration for unresolved complaints.
PDF p. 3, Ombudsman section. Source BG
Octopus Energy
Describes independent Ombudsman review and a decision binding on the supplier, not the customer.
Step Two. Source OE
EDF
Describes Ombudsman referral after eight weeks or a deadlock letter.
Web Step 3; booklet Step 3. Source EDF
What remains to test
The documents direct readers towards external review; they do not measure how many eligible people successfully enter it. The Ombudsman’s own guidance describes eight weeks or earlier deadlock and additional support. [EO]
Delivery test: Check whether the trigger and next step are intelligible, evidence transfers are usable and people needing help can enter. Consult current official guidance when acting.
Stage 5 · HBC-W001-S5
Carry out and confirm the remedy
Was the promised change actually completed?
British Gas
Says it will explain the fix and close the complaint once the customer is happy.
PDF p. 2, Step 3. Source BG
Octopus Energy
Promises an explanation of actions and next steps; describes remedies following Ombudsman findings.
Opening commitments; Step Two. Source OE
EDF
Describes possible Ombudsman remedies and their binding effect on the supplier.
Web Step 3; booklet Step 3. Source EDF
What remains to test
A closure description is not an independently checked completed correction. No common case-level dataset was used to verify delivery, failed remedies or reopening. The Ombudsman guidance specifies implementation within 28 days after acceptance and follow-up if remedies remain outstanding. [EO]
Delivery test: Distinguish supplier agreement, accepted external decision, implementation due date, completed correction and reopening. Include outstanding cases when reporting results.
What the labels mean
“Explicit commitment” means the selected text directly describes that action, channel or responsibility. It includes qualified aims, such as aiming to respond within a stated time: it does not mean a guaranteed result, contractual enforceability or an exclusively human performer. “General aim” would describe a broad aspiration without the specified action. “Not located” would mean the selected documents did not answer that question, not that the practice is absent. “Not applicable” would mark a question outside that actor’s role. No totals or comparative quality scores are calculated.
The five stages are HBC analytical categories, linked to workflow HBC-W001. We inspected official complaint webpages and linked domestic complaint leaflets, retaining their files, access dates and locators. We did not submit complaints, telephone advisers, log in, inspect staff systems or audit customer outcomes. Retrieval dates do not establish when a policy changed. The table’s delivery questions are HBC proposals. We have not contacted the organisations named here.
Start with existing duties and practice
Energy complaints already have statutory standards. The displayed revised 2008 Regulations require procedures describing investigation steps, likely timescales, internal review and remedies (regulation 3), and reasonably required resources for efficient, timely handling (regulation 7). This is a bounded reading of those provisions, not a supplier compliance assessment or a complete licence and legal-currentness audit. [LAW]
Public-sector remedy principles also say staff should know which remedies they can offer, explain decisions and keep reasons on record. They are a useful cross-sector benchmark, not rules made binding on energy suppliers by this comparison. [PHSO]
HBC should build on that work. The specific AI question is how systems allocate usable human authority, preserve capability and pass responsibility through to completion. These documents do not answer whether workers can challenge automated outputs, who benefits from released time, or how either affects service access. Those questions require further evidence, not a new label for familiar complaint-handling principles.
What the published numbers can—and cannot—show
The next check followed the three suppliers’ public performance pages. The selected tables all reported April–June 2026 when inspected on 15 November. This is a snapshot of those pages, not a claim that no later report exists elsewhere. The figures below are published counts, not independently audited results or a ranking.
Ofgem’s November 2025 guidance says quarterly resolved totals include cases received earlier. “Resolved” entails no outstanding provider action and satisfaction of the complainant; sending an Ombudsman signposting letter does not count. Later contact showing it is unresolved requires a repeat or reopened status. These are substantive reporting rules, not just an administrative closure label. [OG]
That explains why resolved totals can exceed received totals. Dividing the two would not show the share of that quarter’s new complaints successfully resolved. Neither count alone gives the number still waiting, their age or the experiences of people who never managed to complain.
Ofgem also specifies resolution-speed reporting and channel breakdowns. Channel is different from whether a person or automated system did the work. Measures per customer account use gas and electricity accounts: one dual-fuel customer has two. Counts of complaints are not counts of unique people. [OG]
Consumer Scotland’s December 2025 comparison already examines the accessibility, comprehensiveness and transparency of complaints reporting across energy, communications, finance and water. HBC need not recreate that broad reporting survey. Its narrower question is what usable human responsibility adds when people and AI share service work. [CS]
The selected totals do not link a case’s outcome to its human or automated handlers, their authority, time, handovers or working conditions. This bounded review does not establish whether such evidence exists elsewhere. It cannot determine which ownership model works better, whether AI caused an improvement, or whether workers shared any gains.
A proportionate next test would link responsibility and decision powers to a defined set of complaints, then follow unresolved and reopened cases as well as completed remedies. Include failed access attempts, adjustments, customer preferences, staff workload and costs. Compare like cases and account for other service changes. Use aggregated reporting and privacy safeguards; publishing individual complaint histories is unnecessary. This is a research design, not a new reporting duty or a claim of supplier cooperation.
What this changes for HBC
The useful research question is more specific than whether a supplier offers a person. It is whether someone can understand the concern, use relevant information, reach authority and keep responsibility until the remedy is complete. These documents are starting points for that inquiry. They do not establish a legal requirement that each stage be performed by a human or show whether AI helped or harmed a case.
A useful HBC contribution would test where human decision powers and responsibility sit when people and automated systems share this work. Compare a continuing team with reliable handovers against a named owner with cover, using comparable routing, staff powers and resources. If these also differ, evaluate the whole service package. Automated routing may support either ownership model. A nominal human signature can add delay without useful authority.
Suppliers could publish clearer operational explanations; Ofgem and consumer bodies are possible audiences for research on consistent reporting. This is a proposed direction for HBC, not an adopted regulatory rule or an institutional partnership. Before a formal policy demand, establish which requirements already apply, what reporting exists elsewhere, the competent institution’s powers, implementation costs and a workable adoption route.
Keep the three aims separate. A clearer route may strengthen customer agency without improving pay. More protected staff time may improve access while increasing prices or workloads. Automation may free time without sharing the gain. A later evaluation should define an eligible cohort and follow-up period, include abandoned and unresolved cases, test barriers, and account for staffing, training, systems and repair work. No benefit or saving is estimated here.
The reporting check narrows the next step: aggregate totals provide context, while testing human responsibility requires more specific evidence. Before seeking access, establish what further public evidence or a small, justified study could provide. A better-performing and less costly arrangement may weaken the case for a particular design; choice or a fair challenge may still have value beyond average satisfaction. Ask: who owns the next action, what can they change, when should it happen, and how will you know it is complete?
Sources and editions
- Source BG
British Gas — complaint leaflet
PDF pp. 2–4; printed code ends 04/25. Linked by the official useful-information page on 14 November 2026.
Four-page leaflet downloaded through its official browser link after automated retrieval was denied.
- Source OE
Octopus Energy — complaint procedure
Step One; unresolved complaints; Step Two. Live page inspected 14 November 2026; no update date established.
Energy-supply procedure; separately linked installation procedures excluded.
- Source EDF
EDF — complaint page and booklet
Web introduction, Steps 1–3; booklet p. 1 (R63.08/25.AW40.V6). Inspected 14 November 2026.
Read together; neither document establishes delivered continuity.
- Source EO
Energy Ombudsman — process guidance
Steps 2, 5 and 6. Inspected 14 November 2026. Published process, not a complete legal-currentness or outcome audit.
Common context for referral and remedy implementation; no supplier-specific outcome inferred.
- Source LAW
Energy complaint regulations — existing duties
Displayed revised text: regulation 3(3)(d), (e), (h); regulation 7(1). Inspected 14 November 2026.
The legislation site warns that revised texts may not be fully up to date. No human-only reservation is inferred.
- Source PHSO
Public-sector remedy principles — staff powers and reasons
Being open and accountable; inspected 14 November 2026.
A public-body benchmark, not a supplier compliance standard or an endorsement of HBC.
- Source BGD
British Gas — public complaint counts
Q2 2026 row; inspected 15 November 2026.
Supplier publication, not independently verified delivery.
- Source OED
Octopus Energy — public complaint counts
Q2 2026 row; inspected 15 November 2026.
Supplier publication, not independently verified delivery.
- Source EDD
EDF — public complaint counts
Q2 2026 table and footnote; inspected 15 November 2026.
Supplier publication, not independently verified delivery.
- Source OG
Ofgem — Guidance on submitting domestic customer complaints data
4 November 2025 edition; §§5.4–5.15 and 6.1, pp. 18–21.
Crown copyright. Inspected 15 November 2026; adherence unverified.
- Source CS
Consumer Scotland — Consumer experience: customer complaints
15 December 2025; Methodology and analytical framework.
A cross-sector publication benchmark; inspected 15 November 2026. No affiliation or endorsement.
HBC research and advocacy are distinct from neutral Policy Window reference material. Shared publication is not independent endorsement. No original customer or worker consultation, supplier partnership, compliance finding or policy uptake is claimed.