Part one · Split the informal jobOne friendly face, four incompatible promises

A finance programme names an enthusiastic analyst its AI champion. Executives hear that she will mobilise the rollout. Data scientists ask her to translate model limitations. Colleagues assume she represents their concerns. Senior owners then ask her to approve the remaining risk because the change must be locally owned. Each audience thinks it delegated one reasonable task. Together they created an unpaid shadow institution.

The analyst has no authority over the business outcome, no formal representative mandate, no protected channel for dissent and no time removed from her ordinary work. She translates a provisional model result into plain language. A sponsor repeats it as a recommendation. Pilot volunteers welcome the tool, so the programme reports local readiness. Operations staff who handle exceptions were never asked. When trouble arrives, the same informal role can be blamed for weak advocacy, inaccurate translation, missing support and risk that a formal owner should have accepted.

“Champion” has become a catch-all title because four separate questions were never answered. Who creates motion around a bounded test? Who preserves meaning across specialist communities? What makes the process credible and contestable for a named local population? Who formally owns the decision and consequence? The first three are interfaces. The fourth is accountability. Combining them may be convenient, but convenience does not create mandate, capacity or legitimacy.

This paper's Three-Role Change Interface separates the jobs. A champion creates legitimate motion around a testable change. A translator carries a claim across knowledge and authority boundaries without hiding transformation or uncertainty. A local legitimacy anchor makes the process intelligible, influenceable and contestable for a named constituency. None becomes a business owner, worker representative, control function or independent assurer merely by helping the change travel.

The unit of performance is not the number of champions, meetings or positive messages. It is a consequential local claim that reaches competent authority without loss of meaning, coercion, source exposure or false representation, then returns as a decision whose reasoning and effect can be inspected. This changes the question from “How many advocates have we recruited?” to “Can a protected claim survive the journey and alter the decision?”

Part one · Split the informal jobMotion, meaning and legitimacy take different paths

The champion advocates a bounded opportunity, the translator protects the claim and the legitimacy anchor protects the process. The distinctions are practical because the evidence and conflicts for each job differ.

A champion frames an opportunity, finds resources, removes access barriers, convenes a test, sustains attention and escalates an orphaned decision. Advocacy is permitted, but its case, evidence, incentives and limits must be disclosed. Repeating a claim in ten forums does not strengthen its evidence. A champion who reports favourable participation must identify who participated and which population did not.

A translator identifies the boundary being crossed. At a syntactic boundary, groups use different labels or formats. At a semantic boundary, they attach different meanings to the same term. At a pragmatic boundary, the claim alters interests, workload or power, so translation may require transformation and negotiation rather than a glossary. The translator preserves uncertainty, exceptions, dissent and consequence. The role can certify fidelity, not the truth or legal acceptability of the underlying claim.

A local legitimacy anchor creates a credible process for a specific constituency, issue, place and period. The person or body explains what is changing, surfaces local history and context, opens protected voice, shows what local input can influence, discloses conflicts and returns decisions with challenge and remedy routes. Standing is renewable and bounded. A respected site manager may be credible with permanent staff yet lack legitimacy for contractors whose schedules will be scored by the tool.

Five formal roles remain visible beside the interface: the business outcome owner; product or service owner; data and technology owner; workforce, employee-representation or affected-party authority; and risk, control and independent assurance owner. One person can carry compatible hats in a small organisation, but each material action states the hat in use. A translator cannot approve what was translated. An anchor cannot accept risk for a population. Independent assurance cannot champion the rollout it must assess.

Part two · Make the role realA visible mandate prevents silent authority drift

A Role Mandate Card describes what one role may do for one named change object. It is visible to the constituency and expires. Without time, access, protection or a responsive owner, the role is decorative change labour.

The card names the role, change object, constituency, workflow, location and lifecycle stage. It lists permitted work and authority to convene, recommend, translate, challenge, escalate or pause. It also names decisions outside the role. Evidence and system access, confidentiality duties, formal owners, escalation routes, incentives, conflicts and vendor relationships are explicit. Time allocation, workload relief, budget, training, support, response service level, succession, rotation, withdrawal, expiry, effectiveness and harm complete the contract.

Permission verbs matter. “May convene a test” is different from “may approve release”. “May recommend a pause” differs from “may suspend use”. “May verify source meaning” does not mean “may validate model performance”. The card is not a miniature job description. It is an interface contract that blocks accidental transfers of accountability and tells the local population how to challenge the role.

The mandate is also a refusal device. When a senior owner asks a champion to sign residual risk, the card makes the transfer visible and sends it back. When a translator is told to remove caveats for an executive audience, the card requires the loss to be disclosed. When an anchor is asked to report that a silent room supports the change, the representation boundary defeats the claim.

Part two · Make the role realA translation needs a receipt, not a polished slide

Consequential claims change as they move. The Translation Receipt records the original, every material transformation and the returned decision so source and receiver can see whether they still mean the same thing.

The receipt starts with the source claim in the source community's language, source authority and protected treatment. It identifies the decision, workflow, population and consequence; target audience and decision sought; and whether the boundary is syntactic, semantic or pragmatic. Terms, assumptions and context that do not transfer directly remain visible. Evidence, uncertainty, dissent and excluded cases travel with the claim.

The translator lists compression and loss introduced, plus the boundary object, example, test or artefact used. Source verification asks whether the originating community recognises the translated statement. Receiver verification asks whether it is understood and actionable, not whether the receiver agrees. The formal owner, response deadline, returned decision, rationale, unresolved difference, expiry and retranslation trigger close the loop.

The Meaning-Loss Test asks eight questions. Which facts, populations, exceptions or consequences disappeared? Did an observation become a recommendation? Did local uncertainty become enterprise certainty? Did a professional or rights boundary become reluctance? Did technical language hide workload, job quality or distribution? Did business shorthand turn an outcome into a usage target? Did the translator add advocacy, authority or consensus? Can the source community recognise and safely dispute the returned statement?

Part three · Earn the local processStanding cannot compensate for a closed decision

The Local Legitimacy Test holds nine conditions separately. Personal credibility can help, but it cannot redeem predetermined outcomes, missing voice, conflict or absent remedy.

Relevant standing asks why a person or body is credible for this issue and constituency. Proximity requires direct understanding of task, history, affected population and consequence. Plural voice covers supporters, sceptics, quiet participants, minorities, contractors, shifts, locations, languages and accessibility. Influence requires evidence that input can change design, boundary, resources, timing or the decision itself.

Procedural fairness includes consistency, accuracy, ethicality, bias control, representation, correctability, explanation and respectful treatment. Independence and conflict discloses reporting lines, performance targets, vendor ties, career incentives and retaliation exposure. Accountable route reaches an owner who can act with a deadline and fallback. Return and remedy sends the decision, rationale, unresolved dissent, recourse and repair back. Renewability recognises that legitimacy weakens or expires as person, issue, constituency, evidence or consequence changes.

Do not sum the nine conditions. Ninety per cent participation cannot compensate for retaliation. Formal appointment cannot repair missing information. High standing cannot make a closed decision influenceable. The evidence panel should show disagreement among the conditions because that disagreement tells an owner what must change.

Part three · Earn the local processParticipation, representation and consent are different claims

The Representation and Consent Boundary prevents a helpful local role from speaking beyond its authority. It also protects disagreement after a fair process from being rewritten as programme failure.

A champion speaks for the case being made, not everyone reached. A translator speaks to translation fidelity, not the merits of the source claim. A legitimacy anchor speaks to process credibility for a named constituency, not universal approval. A manager speaks inside managerial authority and does not automatically represent workers, customers or a profession. A formal employee representative acts within applicable representative rights and processes. An affected party contributes situated evidence and does not accept enterprise risk.

Attendance, silence, access, pilot use and positive feedback are not consent unless a valid consent regime applies and its conditions are met. Participation can be useful without being voluntary agreement. Local influence can improve a decision while some people still oppose it. Honest reporting should say who was engaged, through which route, what could change, what changed, what dissent remains and who finally decided.

The Local Legitimacy Coverage Map tests workflow and exception path; role, profession, grade and layer; location, shift, channel and working arrangement; employment, supplier and contractor status; language, disability and technology access; model, tool, data and vendor; customer, citizen, worker and other affected populations; job, income, monitoring, workload, safety, privacy and rights consequence; lifecycle stage; and support, challenge, appeal, remedy and decision-return routes.

Coverage exists only when a material claim can travel safely, retain meaning, reach authority and receive a visible response. One volunteer in a departmental cell proves none of those conditions. Coverage can be narrower than universal presence, but exclusions must be explicit and proportionate to consequence.

Part four · Fund and renew the interfaceInvisible labour predicts brittle legitimacy

The Change-Role Burden and Risk Ledger exposes time, emotional labour, retaliation exposure, conflict and dependency before the organisation confuses personal sacrifice with a functioning interface.

Record paid and unpaid time; translation, convening, emotional and conflict labour; workload relief and backfill; retaliation, status and professional exposure; pressure to manufacture positivity; dependence on one trusted person; constituencies claimed; distance from authority; unresolved signals and latency; conflicts among advocacy, representation, control and assurance; vendor or sponsor capture; attribution and career effect; succession, rotation and safe withdrawal.

The ledger triggers funding, separation, additional coverage, protected escalation, rotation or termination. It must never reward informal overtime, message volume or positive sentiment. Concentration matters even when hours look manageable. One bilingual translator may carry every interface among data science, legal, operations and employee representatives. Her departure, illness or conflict then becomes a single point of meaning failure.

The Role Renewal and Exit Rule reopens the mandate when the change object, constituency, evidence, organisation, workflow, model, vendor, law, workforce consequence or risk changes materially. It also triggers after serious disagreement, confidentiality breach, retaliation report, repeated meaning loss, false representation, overload or unresolved owner response. Expiry removes stale directories and reassigns open receipts. An honorary role that never ends is not continuity; it is unreviewed authority.

The Three-Role Evidence Panel keeps seven views apart: mandate; translation fidelity; local legitimacy; burden; decision reach; representation integrity; and outcome plus harm. Do not create a champion score or influencer ranking. A visible advocate can have poor translation fidelity. A quiet translator may avert a serious failure without changing usage volume.

Part four · Fund and renew the interfaceTen cases that break the catch-all champion

Each case maps the three interfaces and formal owners, tests meaning and legitimacy, defines the representation boundary, records burden and ends with renewal or exit.

Part five · Exercise the designChampion, Translator and Legitimacy Designer

This instrument routes a design problem; it does not score people or predict compliance. Start with a named scenario, inspect every field and change the selections to test a local case.

Use the output to draft or repair Mandate Cards, a Translation Receipt, the representation boundary and decision route. It cannot establish whether a person is trusted, whether consent is legally valid or whether a model is fit. Those findings require local evidence and competent authority.

ConclusionLet the three roles meet without pretending to be one

AI adoption becomes credible when motion, meaning and local legitimacy meet accountable authority without impersonating it. The work depends on trusted people, but it should not consume their reputation, overtime or vulnerability as free infrastructure.

A champion can create momentum and still say that the evidence is incomplete. A translator can make a claim actionable and still refuse to endorse it. A legitimacy anchor can certify a fair, influenceable process while reporting persistent disagreement. These are signs of role integrity, not weakness.

The operating design is concrete. Name the change object and constituency. Write visible permissions and exclusions. Fund time, access, protection and succession. Record consequential translations from source through loss to returned decision. Test standing, proximity, voice, influence, fairness, conflict, authority, remedy and renewal separately. Draw a hard line among advocacy, fidelity, process legitimacy, formal representation, consent and accountable decision rights. Rebalance burden before one indispensable person becomes the system.

The final test is not whether a champion persuaded the room. Ask whether a local claim retained its meaning, reached somebody able to decide, returned with reasons and changed the condition for the people concerned. If those things occurred, the interface earned its place. If the organisation obtained only applause, it created advocacy without legitimacy.