Human involvement map · Workflow and analysis
A human on the line is only the beginning
What an energy dispute reveals about useful human help: access, authority and action that actually follows.
Public research edition · 14 November 2026 · Human by choice — a Policy Window advocacy programme
A person needs power to help
Imagine disputing an energy bill. You finally reach a person, explain the problem and receive a sympathetic response. But the adviser cannot see the earlier correspondence, change the account or tell you who will act. You have obtained human contact. You have not yet obtained useful help. This is an illustrative scenario, not a reported case.
Human by choice should judge a human option by what someone can do with it: explain an unusual situation, obtain a reasoned response, challenge an error and get an agreed remedy carried out. That requires authority, information and time as well as a person answering. It also leaves room for automation that removes routine work and makes useful human help easier to obtain.
Five stages of useful human help
HBC-W001 is a provisional analytical workflow for a household energy-service dispute. These are proposed contributions to test, not a statutory procedure, a reported customer journey or five legally reserved human functions. Cases may loop, skip stages or need urgent help. The role and skill labels are research aids without official occupational codes.
A protection might apply to a power, such as reviewing a disputed correction, rather than every task in a customer-service job. Conversely, a named person can become a decorative signatory if the system gives them neither discretion nor time. Test the practical allocation of work and authority before proposing a reserved function.
Stage 1 · HBC-W001-S1
Make the concern understood
A usable chance to explain context, correct a misunderstanding and ask for a person.
The test: Can someone who cannot use the default channel still reach useful help?
Tasks, roles, skills and alternatives — stage 1
- Activities
- Receive the concern; Recognise accessibility or urgent-support needs
- Possible roles
- Customer or chosen advocate; Customer-service adviser
- Skills to investigate
- Listening and clarification; Accessible communication
- Resources needed
- Accessible channels, trained staff and a route that does not depend on completing an unsuitable automated interaction.
- Alternative to test
- Optional automated intake, with a clear bypass and preservation of the customer’s account.
Stage 2 · HBC-W001-S2
Investigate the disputed issue
Examine contested facts and give reasons that the person can challenge.
The test: Can the reviewer obtain evidence and change an incorrect outcome, rather than merely endorse it?
Tasks, roles, skills and alternatives — stage 2
- Activities
- Inspect billing and case records; Check a proposed correction
- Possible roles
- Billing specialist; Case reviewer
- Skills to investigate
- Numeracy; Evidence assessment; Contextual judgment
- Resources needed
- Usable records, access to relevant teams, competence, review time and permission to question a system output.
- Alternative to test
- Automated checking of routine errors, with staffed review of contested or uncertain results.
Stage 3 · HBC-W001-S3
Own the response and escalation
Someone accountable for a next action, its owner and its timing.
The test: Does each transfer preserve responsibility, or require the customer to start again?
Tasks, roles, skills and alternatives — stage 3
- Activities
- Keep responsibility for the case; Escalate beyond delegated authority
- Possible roles
- Case owner; Authorised supervisor
- Skills to investigate
- Coordination; Explanation; Working within and challenging authority limits
- Resources needed
- A manageable caseload, continuity, decision rights and a supported escalation route.
- Alternative to test
- Automated routing and status updates, with an identifiable accountable owner when the case is contested.
Stage 4 · HBC-W001-S4
Obtain independent consideration
A route to contest an unresolved decision with appropriate independence and authority.
The test: Can people actually enter the route and challenge its reasoning? No human-only legal reservation is inferred from a reviewer’s title.
Tasks, roles, skills and alternatives — stage 4
- Activities
- Prepare an external challenge; Review evidence from both sides
- Possible roles
- Customer or advocate; Independent dispute reviewer
- Skills to investigate
- Evidence organisation; Fair assessment; Reason giving
- Resources needed
- Accessible evidence submission, clear eligibility, representation where needed and reviewer capacity.
- Alternative to test
- Assisted evidence preparation and triage, with accountable review of contested findings.
Stage 5 · HBC-W001-S5
Carry out and confirm the remedy
A responsible person who can investigate a disputed or failed implementation.
The test: Was the promised change completed and understandable, rather than the case simply marked closed?
Tasks, roles, skills and alternatives — stage 5
- Activities
- Apply an agreed correction; Check whether it worked
- Possible roles
- Remedy owner; Customer or advocate
- Skills to investigate
- Coordination; Verification; Accessible explanation
- Resources needed
- Access to the relevant operational team, follow-through time and a route to address non-delivery.
- Alternative to test
- Automated payment or account correction, with confirmation and staffed handling of errors.
Compare three published supplier procedures against these stages
What the energy evidence can tell us
The January 2026 Ofgem/Citizens Advice survey covered 3,911 domestic energy bill payers in Great Britain. Of the 1,473 who had contacted or tried to contact their supplier in the previous three months, 51% reported a difficulty and 67% said their query was fully resolved. The report gives these answers separately; it does not show which respondents reported both. [1]
The survey already asks about inability to speak to a human, but that question is routed only to phone and webchat users. Other contact difficulties cannot all be called failures to reach a person. The survey includes people who do not use digital services; it did not randomly assign people to human or AI support. [1]
A separate October 2026 report for Ofgem combined three survey rounds, using 4,262 responses with the required information. People’s perceptions of empathy, handling and resolution were linked to satisfaction. That does not show what caused satisfaction to change or whether replacing AI with a person would help. The reported contact channels do not reliably identify whether a person or a system handled the interaction. [2]
A warm response can accompany an unresolved problem. An automated correction can fix one. A human may be indispensable for a contested decision without needing to perform every administrative step. The evidence gives reasons to investigate the service arrangement; it does not settle how much work must be reserved to people.
Follow the whole clock
Energy Ombudsman’s Q2 2026 report records 20,084 accepted domestic gas-and-electricity supply cases. A separate performance panel gives a 32.9-day average from investigation to conclusion, without a domestic-only denominator. It is not the customer’s whole complaint journey. Domestic signposting is 53%: supplier-reported correct referral among cases, not all customers’ access to redress. [4]
Start the investigation earlier. When did the person first seek help? When was the concern recognised as a complaint? How many times did they repeat it? When did someone with useful authority take responsibility? When was the agreed correction actually visible on the account? Averages for one stage cannot answer all of these questions.
A fall in complaints can mean fewer problems, but it can also reflect people giving up or using another route. More accepted cases can reflect more problems, better access to redress or a change in coverage. Ask what changed before treating either direction as success or failure. We have not linked the survey respondents to Ombudsman cases or supplier staffing records.
Build on the work already under way
Citizens Advice rates supplier service using complaints, ease of contact, billing and metering, and customer commitments. Its contact measures include supplier-reported waits and responses through other channels. A rating does not establish whether an adviser could change a decision or whether one customer’s promised correction was carried out. [3]
The government’s June 2026 response already proposes faster escalation and decisions, automatic onboarding work and stronger remedy enforcement. Some changes require legislation; others are for Ofgem and the Ombudsman to progress. These proposals are not proof that the changes have taken effect. [5]
The Ofgem and Ombudsman guidance checked for this article still describes an eight-week supplier stage or earlier deadlock referral; the government response recommends six weeks. These are different publication types and stages. Consult the current official process when seeking help; this article does not certify the complete current legal or implementation position. Ofgem supervises the sector but does not resolve individual supplier disputes. [6]
These efforts address important parts of a complaint journey. The question here is what a person must be able to do, with which resources, for an unresolved problem and an agreed correction to receive useful attention.
What HBC should ask institutions to test
Start with a desk audit of one journey: a disputed household energy bill. Compare published complaint policies, contact measures, referral rules and remedy reporting against these five stages. Publish a short gap table showing which questions can be answered now, which cannot, and what evidence a later study would need. This first deliverable requires no private case files or field trial. Supplier records and published work by Ofgem, consumer bodies and the Ombudsman could inform later research; no partnership or access is claimed.
For a later comparison, define who qualifies and the follow-up period. Compare a person available on request from the outset with support requiring an initial automated attempt before escalation. Preserve an accessible bypass and urgent help in both. Hold adviser powers and resources stable to isolate request timing; otherwise evaluate the whole service design. Self-selected users cannot establish a causal effect by themselves.
Ask four practical questions: who owns the next action, what can they change, when should it happen, and how will the customer know it is complete? Follow all eligible cases, including unresolved, abandoned and still-open cases when the study ends. Record repeat contacts and transfers; separate quick queries from complex disputes. Include disability, language and digital barriers. Any later case-level evaluation needs justified data access and appropriate consent arrangements; this publication collects no case information.
Count the full cohort’s costs, including unresolved cases: staff time, training, interpreters, systems, checks and repairs. Report completed useful outcomes alongside outstanding problems. Cost per answered call can hide an unresolved case. No staffing ratio or saving is estimated here.
Keep choice, income and access separate. Human contact does not prove higher pay or better access; automation can release time without sharing the gains. A choice or a fair chance to contest a decision may deserve protection even without improving average satisfaction. That justification must be argued openly, alongside costs and exclusion.
The case for a particular service design weakens if another arrangement offers equally usable choice, fair decisions and reliable remedies at lower cost. It strengthens if meaningful human authority corrects harms consistently left unresolved elsewhere. A nominal human role without power needs redesign. Judge the next advocacy demand by these tests, not a count of human hand-offs.
Sources and limits
- Source 1
Ofgem / Citizens Advice / BMG — January 2026 survey
Findings pp. 7, 42–45; technical report pp. 8, 21, 26, 76 (E12 routing). Fieldwork 5 January–3 February 2026. Weighted quota survey using several collection methods; reported aggregates, no HBC microdata analysis.
- Source 2
BMG / Ofgem — satisfaction driver analysis, October 2026
Report pp. 9–14, 30–36. Observational logistic models; HBC has not reproduced them or inferred a human-versus-AI treatment effect.
- Source 3
Citizens Advice — how the ratings are calculated
Category and contact measures. Live method inspected 14 November 2026; HBC has not replicated or validated supplier ratings.
- Source 4
Energy Ombudsman — Q2 2026 disputes data
Q2 report pp. 2–4, 10, 16. Excludes disputes resolved before its process. Case counts, supplier-reported signposting and the general performance panel are distinct. Reporting scope expanded in 2026.
- Source 5
DESNZ — energy-redress consultation response
Updated 17 June 2026: Executive Summary; Sections 1–3; Next Steps. Policy response, not proof that every reform has commenced.
- Source 6
Ofgem — complaint routes and institutional roles
What Ofgem can do; escalation. Official guidance inspected 14 November 2026; check current requirements when acting.
HBC analysis and advocacy are distinct from neutral Policy Window reference material. Shared publication is not independent endorsement. No original customer or worker consultation, partnership or policy uptake is claimed.