BDO: One model behind every sustainability applicability decision

BDO’s audit practice runs every sustainability applicability decision through Rainbird. One knowledge graph replaced multiple regulatory decision trees, giving consistent answers at scale with a full evidence trail behind every one.

Rainbird releases around 5,200 chargeable hours a year for BDO through Sustainability Compass, worth roughly £1.2m. Instead of multiple decision trees, auditors work from one deterministic Rainbird model that tells them which sustainability rules apply to a given client, and shows its reasoning for each answer.

The problem

Audit is about 60% of BDO’s business, delivered by thousands of auditors. As sustainability and ESG regulation grew, those auditors had to work out which climate-related reporting obligations applied to each audited entity. It doesn’t apply to most audits, but for the ones it does, the judgement has to be consistent and defensible, especially given such obligations apply to usually larger clients or those under closer scrutiny, 

Sustainability regulation isn’t part of standard auditor training. A small specialist methodology team carried the knowledge, which was documented across several decision trees and long documents, one per regulation. Auditors read the guidance, followed each tree, and eventually reached an answer by hand.

That created friction, not least because the same questions appeared across different trees, so people answered them again and again. Manual interpretation meant outcomes could differ between audits and the methodology team had little sight of how often each regulation was triggered, or whether the logic was being applied consistently.

Why Rainbird

BDO Labs saw a chance to fold the separate policy logic into one model, and built it on Rainbird.

The reason comes down to the kind of decision this is. There’s a right answer, set by the regulation and the entity’s characteristics, and it has to hold up under inspection. A probabilistic model can’t provide a defensible outcome. Rainbird’s reasoning is deterministic: the same inputs always produce the same outcome, and every outcome comes with a proof tree showing the logical chain of reasoning and how the decision was reached.

For an audit firm, that proof tree is of key value. When a regulator or an internal quality review asks how a judgement was made, the answer is already documented, decision by decision, as a byproduct of Rainbird’s architecture

What BDO built

Sustainability Compass is a Rainbird-powered tool that tells an auditor which sustainability and climate regulations apply to a given client. The auditor answers a dynamic and intelligently structured set of questions and gets a clear outcome that guides the rest of the audit.

Behind it sits a single Rainbird knowledge graph that is an encoding of the logic from multiple regulations architected as computable policy. It is a holistic architecture that ensures questions are asked once and applied across regulations. Interaction is optimised so auditors reach an answer in the fewest questions possible. The model can pull data from any external and internal sources, including Companies House, so available data is ingested automatically reducing human interaction.

BDO built their own front-end on Rainbird’s APIs, to fit how their teams work, and every decision is stored centrally. BDO Labs ran the build, with the ESG methodology team building and supporting the knowledge graph. Compass went live in December 2024.

The outcome

BDO estimates Compass releases around 5,200 chargeable hours a year, worth roughly £1.2m. That time goes back into audit work that delivers optimal value to clients and earns fee income, which lifts both capacity and margin across the practice.

The consistency gain matters as much as the time saved. The one model approach means the same inputs always give the same regulatory outcome, so applicability decisions are standardised across thousands of audits instead of suffering from the noise and variability that is an innate limitation of human operation.  

The shift shows up across the whole process. Where auditors once read several documents and worked through multiple decision trees by hand, they now have one Rainbird-powered interaction with inputs pre-filled where the data exists. Where manual interpretation could produce different outcomes, sophisticated deterministic logic now reaches the same result for the same inputs. Where reasoning sat in people’s heads and written guidance, there’s now a standardised model and evidence trees behind every decision. And where the methodology team had little visibility, decisions are stored and reportable centrally, so they can see trigger rates and how audit logic is applied.

An unplanned benefit

The oversight piece wasn’t the original goal, but it’s turned out to be one of the most useful outcomes. The methodology team can now see which regulations come up most often and check the logic is being applied the same way across the firm, without reviewing audits one by one.

The wider point

The audit story sits on top of a more general one. Any regulated decision that has to be governed by regulation policy or procedure to ensure that it is consistent, explainable and defensible hits the same wall with probabilistic AI. 

Rainbird’s deterministic, knowledge-graph approach enables any AI agent to call Rainbird as a tool-call for precise and deterministic reasoning. This is a powerful architectural approach and is applicable whether decisions sit in audit, tax, banking, insurance or healthcare.