ILLUMINATED PRACTICE
Decision-Making, Human Agency and the Future of Regulated Professional Practice
A Whitepaper and Founding Constitution — mob1.co.uk
Preamble: Why This Exists
Something is breaking in the professions that govern the quality of people’s homes and lives. The knowledge required to navigate housing law, disability adaptations, enforcement obligations and the intersecting duties of care has grown beyond what any individual can hold in their head. Experienced practitioners retire and take decades of tacit knowledge with them. Junior officers inherit complexity without context. The legislation that should protect vulnerable people too often fails them — not through malice, but through the quiet erosion of institutional wisdom.
At the same time, the tools available to these professionals have barely changed. Checklists. Spreadsheets. Guidance documents that assume the reader already knows what they need to know. Software that captures outputs but cannot illuminate reasoning.
This document describes a different approach. It is both a practical vision — a technical and organisational architecture for decision-support tools in regulated housing practice — and a statement of the values that must underpin any system that claims to help humans make better decisions.
We are not building automation. We are building illumination. There is a profound difference, and everything that follows depends on holding that distinction.
This is our founding constitution. It describes what we are building, why we are building it this way, and what we will not compromise in the process. It is the document we return to when decisions are uncertain and the thing we ask collaborators, commissioners and users to read before we begin.
I. The Problem We Are Solving
Complexity Without Support
Housing law in England is a web of interlocking obligations. The Housing Grants, Construction and Regeneration Act 1996 imposes mandatory duties on local housing authorities when certain conditions are met. The Care Act 2014 imposes assessment and duty-to-meet-needs obligations on social care authorities. The Housing Act 2004 creates enforcement powers and duties around hazard assessment. The Equality Act 2010 requires public bodies to have due regard to the needs of disabled people. Planning law, building regulations, landlord and tenant legislation — all of these intersect, sometimes harmoniously, sometimes in tension.
A housing officer conducting a Disabled Facilities Grant assessment is, in that moment, navigating all of it simultaneously. They are determining whether a standard is met, whether a referral is required, whether the clock is running, whether evidence is adequate, and whether the decision they are about to record will withstand tribunal scrutiny. They are doing this for dozens of cases at once, often without dedicated legal support, often in an organisation where the person who knew this work deeply left two years ago.
The cost of getting it wrong falls on the most vulnerable people in the system. Delayed adaptations mean people stuck in homes that no longer accommodate their bodies. Wrongly refused grants mean people who cannot access the legal entitlements that exist to protect them. Poorly recorded decisions mean challenges that consume institutional resource and rarely produce better outcomes for anyone.
The Limits of Existing Tools
Current software in this space is largely designed around process management: tracking applications, recording outcomes, generating correspondence. These are useful functions. They are not sufficient. What they do not provide is decision support — the lighting of the path that connects an officer’s observations to a lawful, well-reasoned, auditable outcome.
What is missing is a system that knows the law, knows the guidance, knows the case law, knows what evidence is required and why, knows what implied professional steps a competent practitioner would take even where the statute is silent — and makes all of that available to the decision-maker at the moment they need it, in a form they can engage with rather than merely execute.
The Opportunity Within the Problem
Heavily regulated environments are not merely challenging. They are, properly understood, a market. Every requirement that complexity places on participation — every form, every threshold, every intersecting obligation — currently costs money. Lawyers. Consultants. Compliance officers. Training. Insurance. These are taxes on doing the right thing.
A system that makes the cost of navigating complexity approach zero — that distributes professional expertise to every tier of an organisation and to every case rather than only the ones complex enough to escalate — is not just useful. It is transformative. The organisations that build such systems, and the professionals who use them, will operate in a different category from those who do not.
Europe’s regulatory architecture, often cited as a competitive disadvantage in the race to deploy AI, is precisely the environment in which this kind of work has the greatest value. Complexity is the market. We are building the tools to navigate it.
II. What We Are Building
A Decision-Making IDE
The central product of this enterprise is what we call a Decision-Making Integrated Development Environment — a Decision IDE. The term is borrowed deliberately from software development. A code IDE does not write programs for the developer. It illuminates the environment: it surfaces errors before they compile, it suggests completions, it shows the developer the implications of choices before they are committed. The developer remains the author. The IDE makes authorship better.
A Decision IDE for regulated housing practice does the same for the professional decision-maker. It does not make decisions. It makes the structure of good decision-making visible, navigable and auditable.
The professional remains the author of every decision. The IDE makes authorship better — and makes the quality of reasoning visible for the first time.
Three Architectural Layers
Compliance Scaffolding
The floor of the system. Every statutory obligation, power, prohibition and condition relevant to the case at hand is identified, classified by its normative force, and surfaced to the decision-maker. Mandatory duties are non-negotiable — the system will not allow the workflow to proceed past a mandatory gate without the required conditions being met and recorded. Powers are flagged as available. Standards — the places where statute requires professional judgement rather than rule-following — are identified as such, with the factors that case law and guidance tell us must be considered.
This layer knows the difference between “shall” and “may”. It knows that “necessary and appropriate” requires an occupational therapist’s assessment and that “reasonable and practicable” requires the housing authority’s own recorded reasoning. It knows that the 6-month determination clock starts on receipt of a valid application and that failure to determine in time creates a right of appeal. It makes the invisible architecture of law visible.
Contextual Knowledge Surfacing
The second layer illuminates the space the law leaves underdetermined. Between the explicit obligations and the professional decision lies a vast territory of implied steps, good practice guidance, professional standards, ombudsman findings and case law that defines what competent practice looks like. A competent practitioner knows to record the date of every referral, to chase responses within a defined window, to document every alternative solution considered and why it was rejected, to ensure that refusal reasons address each statutory factor specifically. They know this because they learned it from experienced colleagues, from difficult cases, from professional development that took years.
The contextual knowledge layer formalises this expertise and makes it available at the point where it is needed. Not as a reference document to be searched, but as active intelligence surfaced by the case context. When the system detects a disabled occupant with Care Act eligible needs, it automatically surfaces the intersection with Care Act duties. When it detects a proposed adaptation in an older property, it flags the building regulations considerations. The expertise travels with the case.
Metacognitive Transparency
The third layer is the most radical and the most important. It makes the reasoning of the system itself visible. Not just what the decision is, but why: which provision fired, what edge in the legislative graph activated it, what weight was assigned and on what basis, what the system is uncertain about and why.
This transparency serves three functions. It enables the professional to engage with the quality of the system’s reasoning rather than merely its outputs — which is the difference between genuine oversight and rubber-stamping. It creates an audit trail of reasoning, not just action, which is what regulators, tribunals and ombudsmen actually need to assess the quality of a decision. And it means that when the system is wrong — when the law has changed, when a case falls outside the patterns the library anticipated — the professional can see where the reasoning breaks down and intervene accordingly.
The Legislative Node Library
Underneath the IDE is a structured library of legislative knowledge. Every statutory provision relevant to housing practice is parsed into a node: classified by its normative force, linked to the provisions it depends on and the provisions it activates, enriched with the implied professional steps that competent practice requires, connected by typed relationships that carry the meaning of the entire network.
When the Care Act imposes a duty that modifies what “necessary and appropriate” means in a Disabled Facilities Grant assessment, that relationship is an explicit, versioned edge in the graph. When an ombudsman finding establishes that failure to chase an occupational therapy referral constitutes maladministration, that implied step is embedded as a mandatory node. The library is not a document. It is a living, queryable map of how housing law actually works.
As new legislation is enacted, new cases decided, new guidance issued, the library grows. The system that serves a housing officer in Coventry today will know about a tribunal decision made last month. The expertise is institutional, not individual — and it does not retire.
III. The Human in the System
The Transition We Are Navigating
We are honest about the fact that we are building in a period of profound transition. The role of the human professional in decision-making systems will change significantly over the next decade, and pretending otherwise would be a disservice to the people who use our tools and to the people those tools serve.
The transition moves through several phases, and each phase has a different answer to the question: what is the human for?
Phase One — Now
The human professional is the decision-maker and carries full legal accountability. The IDE illuminates the path; the officer walks it and records their reasoning. Their value is judgement, professional accountability and the ability to exercise discretion in ways the system cannot anticipate. This is where we are building.
Phase Two — Near Term
As the system matures, the decision logic for well-defined mandatory paths becomes increasingly reliable. The human’s irreplaceable contribution shifts toward observation: the embodied presence in the property, the camera, the eye that notices what the checklist did not anticipate. Observations are guardrailed — selected from controlled vocabularies that the IDE defines — which ensures that what enters the system is structured and evidentially useful. This is not deskilling if the system is designed right. A guardrailed observation vocabulary that shows the professional why each observation matters is a continuous education in what to look for.
Phase Three — Medium Term
The professional who works alongside the IDE over time is being educated by it. Every time the system surfaces “this condition triggers this obligation because of this provision — here is why,” the practitioner who reads that is learning housing law in context, not in a classroom. The IDE becomes a continuous professional development engine as a side effect of its primary function. These professionals are not executing instructions. They are building the kind of deep, contextualised expertise that allows them to identify when the system is wrong, when a case falls outside its patterns, when a human call is required that no algorithm could make.
Phase Four — Longer Term
Robotic and automated sensing will eventually be capable of performing many of the observational tasks currently requiring human presence. The IDE’s ontology of what constitutes a hazard, what counts as an evidentially adequate observation, what a photograph must capture to satisfy a legal standard — all of this can, in principle, be translated into sensor requirements and autonomous behaviour. We name this phase not to celebrate it but to design for it: the question of what human value is preserved must be answered before the technology arrives, not after.
What We Will Not Compromise
Across all of these phases, we hold to three non-negotiable commitments about human involvement.
Genuine oversight, not legal cover
A system that presents conclusions without showing working does not enable human oversight. It creates the appearance of oversight while achieving its practical elimination: the human carries liability without exercising genuine judgement. Every component we build is designed to require real engagement with the quality of reasoning. The professional must be able to see where the system’s logic comes from, challenge it where they disagree, and record that disagreement with their own reasoning. If the human’s role is to ratify outputs, we have failed.
Education, not deskilling
The guardrailed observation interface that presents a vocabulary without explaining why any item matters produces a human sensor, not a professional. The same interface, designed to surface the significance of each observation in the legislative context, produces someone who understands their work more deeply each time they use the tool. We will always choose the second design. The test is simple: does this interaction make the professional more capable of working without the tool, or less?
Agency at the table
The professionals who use these tools accumulate, through their practice, something that no institution currently captures systematically: empirical evidence of how law operates in the real world. Which provisions consistently produce absurd outcomes. Where legislation failed to anticipate the actual condition of housing stock. Where the intersection of duties creates impossible conflicts for the officer trying to serve all of them. This evidence base is the foundation for genuine participation in shaping law and policy — not advocacy based on opinion, but argument based on data. We design our systems to make that evidence base legible and to ensure that the people who generate it have a voice in what is done with it.
IV. Complexity as Opportunity
The Infrastructure Question
Europe does not control the data centres, the satellite networks or the foundation model training pipelines on which AI infrastructure depends. This is a genuine vulnerability — not existential, because the economic logic of serving a wealthy, sophisticated market of 450 million people is too powerful to abandon, but real in the negotiating leverage it cedes.
The more significant dependency, however, is epistemic. When the models that power professional decision support are trained on data that does not reflect the specific legislative context, professional standards and case law of English housing practice, the values embedded in those models — what counts as a good decision, what evidence is adequate, what professional conduct looks like — are someone else’s. The Decision IDE answers this directly: the scaffold of law, guidance and professional knowledge that constrains and informs the AI layer is defined by the domain experts who build the library, not by whoever trained the underlying model.
Regulation as Design Specification
The AI Act’s requirements for transparency, human oversight and explainability in high-stakes decisions are not merely compliance burdens. They are a forcing function toward better-designed systems. A product built to meet the highest standards for explainability in automated decision-making — one that can show its reasoning, demonstrate human oversight, and provide an auditable trail — is a product that works in every regulated market. A product optimised for speed in a low-regulation environment often cannot enter regulated markets without being rebuilt from the ground up.
We are building to the highest standard from the start. Not because we are required to, but because the systems that are trusted in consequential settings are the ones that will matter. The race to ship the fastest model will be decided by a small number of very large players. The race to build AI that can be trusted in the places where decisions about people’s homes, health and care are made — that race has barely started.
Geographic Irrelevance
One consequence of this architecture is that the value it creates is not tied to geography. The legislative node library for English housing law can be populated from anywhere. A housing charity in New Zealand, a local authority in Germany working with equivalent legislation, an NGO in any jurisdiction where housing adaptations for disabled people are regulated by law — all face the same underlying problem: expertise that is expensive, unevenly distributed and constantly at risk of retiring.
The system we are building for English housing practice is a template for every regulated domain in every jurisdiction. The parser, the node schema, the workflow compiler, the IDE architecture — these are domain-agnostic. What changes is the library. And the library, once the method of building it is established, can be populated by domain experts anywhere.
V. Our Commitments — The Constitution
The following commitments are not aspirations. They are the conditions under which this enterprise operates. We hold ourselves to them in every design decision, every commercial relationship and every deployment.
1. We illuminate, we do not replace Every tool we build preserves the professional as the author of every consequential decision. We design for genuine engagement with reasoning, not the appearance of oversight. Where our tools reduce the need for human presence, we account for what human value is lost and design to preserve it in another form.
2. We make our reasoning transparent No component of our systems operates as a black box in a context where a human professional is accountable for the output. Every weight, every edge, every implied step has a recorded source and rationale. The system can explain itself, and professionals are trained to ask it to.
3. We educate through use Our tools are designed so that the professional who uses them consistently becomes more capable, not more dependent. The test of every interface decision is whether it builds understanding or merely transfers execution. We choose understanding.
4. We do not accept less than best practice Our workflow architecture holds best practice as the default and requires explicit, recorded justification for any departure. We do not build systems oriented toward minimum viable compliance. The floor is lawfulness. The standard is excellence. The system asks why when a professional chooses less.
5. We build the evidence base for better law The practice data that flows through our systems is — with appropriate governance and anonymisation — evidence of how law operates in the real world. We commit to making that evidence available to inform law reform, to support the professionals who generated it in having a voice in policy, and to never allowing it to be used against the interests of the people the law is designed to protect.
6. We work with the regulated, not around them Regulation in housing, disability and care exists because these are domains where power imbalances can cause serious harm to vulnerable people. We do not build tools designed to find the minimum compliance path. We build tools that help professionals meet the spirit of the law, not only its letter. Where the law is inadequate, we say so and contribute to changing it.
7. We are honest about the transition We name the phases of automation honestly. We do not pretend that the long-term trajectory of this technology leaves professional roles unchanged. We design for the version of human involvement that is genuinely valuable at each phase, and we refuse to design systems that preserve the appearance of human agency while eliminating its substance.
The north star is not efficiency. It is a world in which every person who needs a housing adaptation, an enforcement action or a care assessment receives one — made by a professional who understands their situation, supported by a system that holds the full weight of the law, guided by the accumulated wisdom of the best practice in the field. Geography irrelevant. Expertise redistributed. Quality floor raised for everyone.
A Note on This Document
This whitepaper was written in 2025 as the founding document of an enterprise that does not yet have a name larger than its domain. It begins with housing adaptations in England because that is where the need is immediate, the legislation is intricate, and the consequences of poor decision-making fall hardest on people who can least afford to absorb them.
It will be followed by tools, by content, by a community of professionals who use the system and contribute to its growth. The legislative node library will expand to cover every piece of housing legislation in England. The decision IDE will move from concept to prototype to deployed tool. The evidence base will accumulate.
What will not change is the constitution in Section V. If we are building in ten years and a decision feels uncertain, we return to those seven commitments. If a commercial opportunity requires us to compromise any of them, we decline the opportunity.
We believe that the systems most worth building are the ones that raise the floor for everyone — that take the knowledge currently held by the best and make it available to all. In regulated housing practice, that means the person in the property in the most difficult circumstances gets the same quality of decision as the person whose case landed on the most experienced officer’s desk.
That is what we are building. This is why.
mob1.co.uk — Housing Adaptations | Professional Knowledge Tools | Human-Agent Systems