The sustainability team has done everything right. The reporting AI is grounded on the approved factor library, every draft carries its provenance, and the pilot cut two weeks off the Scope 3 close. Then the controller in finance quietly tells her team not to rely on anything the model touched until it has been re-checked by hand, procurement declines to route supplier data through a tool that "is a sustainability thing," and legal asks for a memo before any AI-drafted narrative goes near the disclosure. Within a quarter the program that worked in the pilot is a sustainability side project that the rest of the building steps around. Nothing was wrong with the AI. What was missing was a coalition: a named champion in each function that co-owns the report, carrying the same contract into finance, procurement, IR, and legal, so the program belongs to the whole disclosure, not to one team.

Why Reporting AI Dies Inside One Team

Non-financial reporting is no longer a sustainability document produced by a sustainability team. Under CSRD and ISSB it is a co-owned corporate disclosure, and the functions that own its pieces sit outside the sustainability office. Finance owns the numbers, the controls, and increasingly the sign-off, because a sustainability statement filed alongside the financial statements is a finance-grade artifact. Procurement owns the supplier relationships that produce Scope 3 data, which averages around 75% of a company's total footprint across the 15 GHG Protocol categories. Investor relations owns the conversation with the analysts and investors who read the report and price the company on it. Legal owns the disclosure liability, including greenwashing exposure on every claim and target. The sustainability lead orchestrates, but she does not own most of the inputs, and she cannot compel the functions that do.

This is why a reporting AI program that lives only in sustainability dies. The pilot succeeds because the sustainability team controls its own corner. Then the program reaches the boundary of that corner and stops, because the data it needs, the controls it must satisfy, and the liability it must respect all live in functions that were never asked to co-own the tool. A model cannot cross an organizational boundary by itself. It needs a person on the other side who has already decided the program is theirs too, who will vouch for it inside their own function, and who will uphold the same discipline that makes the output assurable. That person is a champion, and a champion outside sustainability is not a nice-to-have. It is the difference between a pilot and a program.

The failure is rarely loud. No one vetoes the AI in a steering meeting. Instead finance keeps a shadow reconciliation, procurement routes supplier requests through its own untouched process, IR declines to cite an AI-assisted figure in an investor deck, and legal adds a review gate that no one staffs. Each function is being reasonable inside its own risk frame. Collectively they strand the program at the sustainability border. The strategist's job is not to overpower these functions but to convert one person in each of them into an owner, because a coalition of owners carries the program where authority alone cannot push it.

The regulatory backdrop makes this urgent rather than optional. Post-Omnibus, the companies left in CSRD scope (large undertakings with more than 1,000 employees and more than EUR 450M turnover under Directive (EU) 2026/470) are the largest ones, where a failed disclosure is a board-level event. In parallel, 73% of large global companies now obtain external assurance on at least some sustainability disclosures, up from 51% in 2019, and GHG emissions are the most-assured category. That combination means the report is simultaneously higher-stakes and more heavily tested than ever, and it is exactly the report whose inputs are spread across functions the sustainability team does not control. A tool that cannot follow the data into those functions is a tool that cannot survive the engagement.

A reporting AI program owned by one team is a pilot. A program with a champion in every function that co-owns the report is an institution.

What Makes a Champion, and How to Spot One in Each Function

A champion is not the most senior person in a function and not the most enthusiastic about AI. Seniority without credibility does not move the working level, and enthusiasm without judgment produces exactly the ungoverned shortcuts that fail assurance. The champion you want has a specific profile. They are credible inside their own function, so their word carries weight with the people who actually do the work. They already feel the pain the AI addresses, so they have a reason to care beyond a mandate. They respect the assurance discipline rather than seeing it as friction, so they will uphold traceability instead of quietly eroding it. And they have enough standing to change how their function operates, not just to have an opinion about it. Credibility, felt pain, assurance respect, and the standing to act: those are the four traits, and all four matter.

The wrong champion is worse than none. An enthusiast with no assurance instinct becomes the person who pastes supplier spend into an open chatbot and calls the result data, creating the exact finding the program was meant to prevent. A senior sponsor with no working credibility endorses the tool in a memo that the working level ignores. The champion who lasts is the practitioner other practitioners trust, who wants the problem solved and wants it solved in a way that survives an assurer. You are recruiting a specific person, not filling a seat.

The Finance Champion

In finance, look for the controller or the assurance-readiness lead who already treats the sustainability numbers as auditable figures. This person feels the pain of reconciling non-financial data that arrives late and unsupported, and they instinctively ask where a number came from. That instinct is the asset: a finance champion who insists on an audit trail is not resisting the AI, they are demanding the exact thing that makes the AI assurable. Recruit the person who says "show me the basis" as a reflex, because they will hold the coalition to the standard that earns the assurer.

The Procurement Champion

In procurement, look for the category manager or supplier-data lead who is already drowning in Scope 3 supplier requests and knows that 79% of reporters cite supplier-data availability as a top barrier because they live it. This person feels the pain daily: chasing suppliers for primary data, receiving nothing usable, and being blamed for a gap they did not create. A reporting AI that triages supplier responses, flags implausible figures, and distinguishes primary from secondary data is relief to them, not an intrusion. The right procurement champion sees the tool as leverage over an impossible workload, and they understand that a labeled estimate is honest while a laundered one is a liability.

The IR and Legal Champions

In investor relations, look for the person who fields the hard questions from analysts and ESG raters and who has been caught out by a number they could not substantiate on a call. They want figures they can defend live, which is precisely what a traceable, assurance-ready file provides. In legal, look for the disclosure counsel who already owns greenwashing and misstatement risk and who is more comfortable with a documented, controlled process than with an opaque one. The legal champion is often mistaken for the program's opponent. In fact they are its natural ally, because a program built on labeled estimates, human sign-off, and a reconstructable file is a program that reduces the liability they are paid to worry about. Recruit the counsel who would rather engineer the control than write the disclaimer.

The Operating Rhythm of the Coalition

A coalition that meets when there is a crisis is not a coalition; it is a fire drill. What carries reporting AI across functions is a rhythm, a predictable cadence and a standing forum where the work of the report flows between the functions that co-own it. Without a rhythm, the champions revert to their day jobs and the program fragments back into a sustainability project. With one, the coalition becomes the place where cross-functional reporting decisions actually get made.

The spine of the rhythm is a standing forum, often best framed as a working group under the existing disclosure or ESG governance committee rather than a new body competing with it. It meets on a cadence tied to the reporting calendar: more frequently in the run-up to the close and assurance, less frequently in the quiet quarters, but never disbanding, because a coalition that disbands has to be rebuilt from scratch every cycle. The membership is the champions, one per function, plus the sustainability lead who orchestrates and a standing line to the assurer so surprises are surfaced early rather than discovered late.

The purpose of the forum is not status reporting; it is decision-making and hand-off. Work on a non-financial report flows across functions in a chain, and the coalition exists to keep that chain traceable at every hand-off. Procurement's champion owns the supplier-data intake and the labeling of primary versus secondary at the source. The sustainability team runs the AI-assisted estimation and inventory build. Finance's champion owns the controls, the reconciliation, and assurance-readiness. IR's champion pressure-tests how the figures will read to an investor. Legal's champion checks the claims and targets for liability. The forum is where each hand-off is confirmed, where a figure that changed is traced, and where a disagreement is resolved before it reaches the assurer as a gap.

A useful discipline is to give the forum a short standing agenda that never changes: what changed since last time and why, what is at risk for the upcoming close, what an assurer would question in the current file, and what each function needs from the others before the next hand-off. The agenda is boring on purpose. Its job is to make the traceable behavior habitual, so that upholding the contract is the routine of the coalition rather than an act of heroism at deadline.

One further point on the rhythm: the forum is also where the coalition maps a single fact base to the several frameworks the company reports under, whether ESRS datapoints under CSRD, IFRS S1 and S2 under ISSB, or a voluntary framework for a rater. The functions do not each build their own version of the truth. Procurement's labeled supplier data, finance's reconciled numbers, and the sustainability team's inventory feed one traceable fact base that every framework draws from. When the coalition holds that discipline, a figure questioned by an assurer under one framework reconciles cleanly to the same evidence under another, and the report stops being a set of parallel documents that quietly disagree with each other.

Turning Resistance Into Buy-In, Function by Function

Every function that co-owns the report has a rational objection to reporting AI, and the objection is not obstruction. It is the function protecting the thing it is accountable for. The strategist who treats the objection as an obstacle loses; the one who treats it as the doorway to buy-in wins, because each objection, taken seriously, points straight at the control that turns the skeptic into a champion.

Finance: Audit-Trail Skepticism

Finance's objection is that it cannot put its name on numbers whose origin it cannot see. This is the correct instinct, not resistance, and the answer is not to reassure finance but to hand it exactly what it is asking for. Show the controller that the program captures, for every AI-assisted figure, what the model produced, what the human did with it, who signed off, and what evidence the number traces to, all at the moment of creation. Finance's demand for an audit trail is satisfied by the very design that makes the output assurable. The objection dissolves into ownership the moment the controller realizes the AI program produces a cleaner trail than the manual process it replaced, because the manual process rarely recorded the reasoning at all.

Procurement: "Not My Job"

Procurement's objection is that sustainability reporting is not its mandate and that supplier data is already the hardest part of its week. Answering this with an appeal to corporate duty fails. The answer is to reframe the AI as leverage on procurement's own pain. The same tool that helps the report also triages the supplier questionnaires procurement already dreads, flags the responses that will not survive scrutiny, and turns a chaotic intake into a managed one. Procurement stops seeing a favor it is doing for sustainability and starts seeing a tool that makes its own job survivable. The reframing is honest: since Scope 3 supplier data is procurement's turf, a tool that governs that data well is genuinely procurement's tool, and the champion who understands that becomes the coalition's most valuable member because they hold the largest and riskiest part of the footprint.

Legal: Liability Caution

Legal's objection is that AI introduces new disclosure liability: a hallucinated figure, a softened negative impact, a target the company never set, any of which becomes a greenwashing or misstatement exposure with legal's name on the risk. Again the instinct is right, and again the answer is the control, not the reassurance. Show legal that the program forbids the model from committing a number, labels every estimate as an estimate with its method and uncertainty, grounds narrative in the underlying facts, and produces a file that can defend any claim on demand. A controlled AI process is easier to defend than an ad hoc human one, because it is documented by construction. Legal's caution, properly channeled, makes it the guardian of the labeling discipline, which is one of the strongest defenses the program has.

IR: Substantiation Pressure

IR's objection is subtler: it does not distrust the AI so much as fear being handed a number it cannot substantiate when an analyst pushes back on a call. The answer meets that fear directly. A traceable, assurance-ready figure is one IR can defend live, with the basis one click away rather than a scramble after the call. IR converts from cautious consumer to active champion when it realizes the coalition gives it the thing it wants most: figures that survive the hardest question in the room.

The Champion as Carrier of the Contract

The single most important thing a champion does is carry the program's core contract into their own function and make it the default there. That contract is the sentence the whole program is built on: AI assists, the human decides, the file proves it. The sustainability team cannot enforce that sentence inside finance, procurement, IR, or legal. Only a champion embedded in each function can, because they translate the contract into the language and workflow of their own function and hold their colleagues to it.

In practice this means the finance champion insists that no AI-assisted figure reaches the controls without a captured decision and a source. The procurement champion insists that supplier data is labeled primary or secondary, and estimated or measured, at the point of intake, so nothing is laundered downstream. The IR champion refuses to cite a figure that cannot be reconstructed. The legal champion insists every claim traces to evidence. Four functions, one contract, each enforced from the inside by someone the function trusts. This is how a slogan becomes a culture: not by a policy the sustainability team publishes, but by a champion in each function who makes the traceable way the only way, so that doing it right is the routine rather than the exception.

Assurance is the reason this matters and the test it must pass. The point of a champion in every function is that every function's data reaches the assurer already traceable, because it was made traceable at the source by someone who owns that source. An assurer who samples a Scope 3 figure and finds procurement labeled it at intake, finance reconciled it under control, and the file reconstructs the whole chain files that figure in the well-controlled column, and the coalition has done its job. The coalition connects to governance by living under the disclosure committee, and it connects to assurance by making traceability the shared standard of every function that touches the report.

A Worked Example: The Same Scope 3 Category, Coalition Versus No Coalition

Consider a company building its purchased-goods-and-services emissions, the largest slice of a Scope 3 inventory that is itself around three-quarters of the footprint. Watch the same category move through the organization two ways.

Without a coalition, the sustainability team runs a capable reporting AI in isolation. Procurement forwards a spreadsheet of supplier spend with no labels, treating the request as a sustainability chore. The AI produces spend-based estimates, correctly grounded and labeled inside the sustainability workflow, but finance has never seen the control design, so it keeps a parallel manual reconciliation it trusts more and quietly disregards the AI output. IR, unsure of the basis, avoids the figure in its investor materials. Legal, never shown the labeling discipline, adds a review gate that stalls the narrative. When the assurer arrives, he finds a sustainability tool the rest of the company does not stand behind, a finance function running a shadow process, and supplier data that was never labeled at the source. He raises his assessed risk for the whole category, expands testing, and the engagement scoped as limited assurance becomes a scramble. The AI was fine. The absence of a coalition made it unassurable at the organizational level.

With a coalition, the same category moves differently. Procurement's champion has already made labeling primary versus secondary the default at supplier intake, so the data arrives typed and honest. The sustainability team runs the grounded estimation. Finance's champion co-designed the controls, so the reconciliation runs inside the program rather than beside it, and the controller stands behind the numbers. IR's champion has pressure-tested how the figures read and is ready to defend them on a call. Legal's champion confirmed every claim traces to evidence and every estimate is labeled. When the assurer samples a supplier, the whole chain reconstructs on demand: labeled intake, grounded estimate, human decision, reconciliation, source. He files it in the well-controlled column and the scope holds. Same category, same underlying AI, opposite outcome. The only difference was a champion in every function carrying the same contract, so the report was co-owned rather than stranded at the sustainability border.

Key Takeaways

• Non-financial reporting is co-owned: finance owns the numbers and sign-off, procurement owns Scope 3 supplier data (around 75% of the footprint), IR faces investors, and legal owns disclosure liability. A reporting AI that lives only in sustainability dies at the first organizational boundary.
• A champion is a specific person, not a seat: credible inside their function, already feeling the pain the AI addresses, respectful of assurance discipline, and with the standing to change how the function operates. The wrong champion (enthusiasm without assurance instinct) is worse than none.
• Spot the champion in each function: the controller who reflexively asks "show me the basis," the category manager drowning in supplier data (79% cite supplier-data availability as a barrier), the IR lead caught out by an unsubstantiated number, and the disclosure counsel who prefers a documented control to a disclaimer.
• The coalition runs on a rhythm: a standing forum under the disclosure committee, on a cadence tied to the reporting calendar, that never disbands. Its job is decision-making and traceable hand-offs, not status reporting.
• Every function's objection is the doorway to buy-in, not obstruction. Finance's audit-trail skepticism, procurement's "not my job," legal's liability caution, and IR's substantiation fear each point straight at the control that turns the skeptic into an owner.
• The champion's core job is to carry the contract (AI assists, the human decides, the file proves it) into their own function and make the traceable way the only way, translated into that function's workflow and enforced from the inside.
• Every function's champion must uphold traceability at the source: procurement labels primary versus secondary at intake, finance reconciles under control, legal confirms claims trace to evidence, so data reaches the assurer already assurable.
• The coalition connects to governance by living under the disclosure committee and to assurance by making traceability the shared standard of every function that touches the report. A program with a champion in every co-owning function is an institution, not a pilot.