As-Built

Posted by George Stevenson from a linkedin exchange with Mark Austin
Mark – George great work as always, maybe the challenge lies in ‘practical completion’ in JCT as opposed to ‘completion’ in NEC and FIDIC ICheme in infrastructure projects the handover and completion provisions can be very extensive including as-built information. This included on one project not receiving a Take Over certificate because of 2 missing as-built drawings

Posted by George Stevenson from a linkedin exchange with Charlotte Brogan (ACIOB) Senior Design and Technical Manager at Fortem
Charlotte – Brilliant piece of work, George Stevenson.
Having looked over the CLC / BIMinAM consultation details, focusing phase one entirely on what asset teams actually need is exactly the right starting point.
From a design management and document control perspective, the “how” requires a complete process shift. A true “as-built” relies on taking the contractor’s real-time site updates, paired with the site manager’s verified red-line drawings, to accurately update the master design information. Crucially, this validation must happen progressively—one work package at a time—rather than a mad rush at the end of the project.
To achieve the data integrity this guidance could benefit from the following statement; “Principal Contractors must contractually ensure designers are scoped to physically visit and audit the site during RIBA Stages 5 and 6.” If designers aren’t progressively on site verifying package completions alongside site managers, we cannot provide clients with the absolute clarity and trusted evidence that the Golden Thread demands. Looking forward to seeing how the next round of workshops tackles delivery!

Posted by George Stevenson from a linkedin exchange with Ivan Young – Principal Designer (Building Regulations) | Author of the Gateway Bibles | Gateway 2 and Gateway 3, board-level strategy then the appointments underneath it | Competence assessed against the work
Ivan – I am not sure this is as unresolved as the discussion suggests. As built information already has a definition.
The regulations require plans showing that the work as built complies, and require the information to be handed to the person who will hold it. On higher risk buildings that is settled, and nobody needs a workshop to discover what it means.
They also answer your question. Nobody certifies what they did not verify, because the person who made the change records the change, before it is made, with the affected documents revised.
Do that and the as built is a by-product rather than a deliverable anyone attests to at the end.
And I would note where the resistance is coming from. The firms worried about the implications for them and their insurers are not worried about the definition. They are worried about being asked to evidence what they built.
George – Ivan Young – that is what people SHOULD be doing but it is clear that most are not doing so – and in the absence of those updated drawings and records, the client often gets a large number unverified drawings and schedules.
The industry clearly needs to change in light of regulations and potential enforcement but I feel we need to do it step by step and getting an industry-wide consensus of what “as-built” should include is one such step
Ivan – I disagree. Everybody understands what as built entails. If it isn’t provided to the regulator the application should be rejected. Understanding what as built means is not some higher level of understanding, it’s basic 1st year student content. Exploring new ways to counter the process serves nobody.
AS-BUILT INFORMATION, WHERE IT ACTUALLY SITS
SI 2023/909 reg 40(2)(c). The operative requirement. Such other plans as are necessary to show the work complied with all applicable requirements. A test, not a list. It never uses the words as built.
Reg 40(2)(d) to (f). Agreed documents, change control log, compliance declarations. reg 38(1) to (3). Handover to the relevant person by completion, filing structure retained, receipt acknowledged as sufficient.
38(5) defines it back to reg 40, so it carries no independent list.
Reg 31(2)(a) to (h). Properties of the information, not its content. Electronic, transferable, accurate, readable, available on request, secure, change logged, consistent.
Building Regulations 2010 reg 38. Fire safety information. Erection, extension and relevant change of use only. Not material alteration. Approved Document B vol 2 s19 and vol 1 s17. The only itemised list. Essential, plus additional for complex buildings. Guidance. BS 8644-1:2022 for structure.
We are once again attempting to change the new regime to suit old patterns and work methodology. we cannot cherry pick through the instruments to find the best fit, the documents should be read completely.
George – Thanks for the detail Ivan.
If everyone knows what should be delivered, why do so few projects delivery complete as-built drawings?
Historically, architects were commissioned to do them informed by contractors redlining where things had changed.
That no longer happens but it seems that many projects dont do what you and I agree they should.
You are much more experienced in this than me so will have seen many more projects on the ground.
Are you saying that all the detailed MEP drawings have been 100% updated with the actual installation details?

Posted by George Stevenson from a linkedin exchange with Marc Bradfield FCIOB – Director – Nucleus Compliance Group Ltd
Marc – your challenge point above “the suggestion that as-built information and handover information should be treated as entirely separate deliverables.”
I could not AGREE more. 
Teams often become hyper focussed on ticking (often regulatory) boxes, with little understanding and fail to consider what is actually needed for a safe and successful building.
George – Thanks Marc. Your background and experience makes your contribution particularly valuable to us.

Posted by George Stevenson from a linkedin exchange with Gavin Edwards – Experienced Architectural and Construction Specialist at GE Consulting
Gavin – Designers shouldn’t produce “As Built” information
George – Yet that is often what clients are given in their handover information
Gavin – it would just be the last construction issue by the consultant – you’d never go back to site and update your drawing to match what you found had been constructed – as built information is prepared by the contractor as his construction progresses
George – if only that really happened. I agree that it’s not the responsibility of the designer – but that is often what we get. As can be seen from the survey we did last week.
If you know of contractors who actually ensure all the drawings are truly “as-built” please ask them to contact me.
There must be a unicorn out there somewhere
Gavin – there’s meat here for a significant debate but if you call something As Built and your Services say thats what you’ll provide then the Client has every right to expect just that – I always delete As Built because I know it is not possible for a designer to provide that – I’m surprised that there’s not been more comments on the topic – a good subject though so 10/10 on your post…
George – Thanks Gavin. I think the important thing is for clients to understand the process of procuring buildings and services will need to change if they really need all of the drawings to reflect what was built.
I am not convinced that is necessary for ALL record drawings. We should prioritise what is really important and what is nice to have. Also have a status of Verified or Unverified.
Gavin – I agree totally but a cold beer needed

This comment is posted by george stevenson but originated from David Shepherd from Houses of Parliament project.
David – Thanks for this thoughtful post. The distinction between record information and verification of installed reality is important.
A designer’s final construction issue information may be entirely correct as a record of design intent and approved changes, without constituting evidence that the asset was installed exactly as shown.
What stands out for me is the reference to “the level of verification that has been applied”, but that naturally raises a further question: “when was that verification performed, and what evidence supports it?”
Many asset attributes can only be captured during a limited evidential window (e.g., installation and commissioning).
If identity, location, commissioning status or compliance evidence are not captured while directly observable, the industry often has little choice but to recover, infer or reconstruct the information later.
Surely, the next step is not just distinguishing between design information and installed reality, but also between information that is observed, derived and inferred.
The missing step is to define evidence capture during the evidential window as a contractual deliverable, not merely as handover documentation.

As part of the draft consultation, I would offer the following observations:

1.  Lack of a concise definition, or robust framework for definition, of the ‘physical’ meaning of as-built, 
2. Blurring of the ‘as-built’ information, handover, and O&M information
3. Scope drift into validation, inspection and test plans, BIM, and asset management 
4. Text that could easily mislead readers, either by its contents, or the context in which it has been placed
5. Significant repetition of content and language 

 
The ancillary issues that the guidance touches on are important, but should not be included in this note as they are already extensively covered in other guidance.
 Similarly, many elements of the draft could be split out into a separate ‘handover and O&M definition’ document – this would allow the as-built definition to be concisely and cleanly defined.
 Is there room for the working group to take time to take stock of the current comments and restructure the document?

—–As a Safety Officer, I believe that health and safety information should be an essential part of the handover documentation. It should include:

– As-built drawings showing safety-critical systems and installations.
– Fire protection and emergency arrangements.
– Risk assessments and safe operating procedures.
– Inspection, testing, and maintenance records.
– Relevant safety information for future maintenance and operations.

Providing accurate and complete safety information at handover will help building owners and operators manage risks effectively throughout the building’s lifecycle.

Thank you for the opportunity to contribute.

—-Suggested omissions include;

• Design specifications / performance specifications
• Design change logs.
• Inspection and Test Plans, An ITR often has no context without the corresponding inspection and test plan which sets out the tolerances to be met!

Without the above the building operator maintainer can not start understand what has changed from the design intent and the work arounds required to keep the occupants safe.

Note that the approach by CLC and the above comments relates to BR requirements and not CDM2015 requirements.

The CDM 2015 requirement is suggested to be for an As-Built ‘access and working space drawing’ showing vehicle and pedestrian routes around the site and structures, access to below ground services, access to services inside and outside the structure, locations for access equipment, cranes and emergency response vehicles. The drawings need to show all installed plant, equipment fittings and fixtures and the access space around them and warnings where work equipment is required to access / reach a working space. These drawings are also suggested to identify locations where services are isolated.

• My initial feel is that the product safety aspect is very light touch and needs reinforced, and equally I have asked if our (SAFed) committee can consider preparing something (if we don’t have a document already) for lifts similar to the fire door documents produced by NAHFO.

The next tranche of comments have been received directly by Anthony Taylor and are transcribed/copied across to this record for the purposes of ‘The Record’ being in one place and to facilitate analysis.

I am an ARB-registered architect who has switched to producing information-management software for the AEC sector, so I should declare a commercial position. My comment is made in my professional capacity, with the safety of the public in mind.
Two kinds of information in the handover pack
The draft rightly anchors trustworthy information in people: how it was collated, verified and validated, and by whom (section 7), and competent, independent sign-off (9.5). The provenance, traceability and signatures it asks for are valuable because a competent person stands behind them. My reason for writing is not to challenge that, nor to debate the meaning of the term “as-built”, but to point out that the pack holds two distinct kinds of information, and that emerging AI workflows threaten the competence behind one of them in a way the record itself cannot surface.
The first kind is a record of the building: drawings and models of the final state, the location of concealed items, attribute data. For these, as an oversimplification, “verification confirms the information matches what was built” works as a test, and automation is little threat where human sign-off is required, because the act of verifying is a distinct act grounded in the physical world.
The second kind is derived from documents, not from the building: key design assumptions for future modification, derogations accepted by the client, residual risk registers, outstanding issues and assumptions. For these there is no building to check against, and the requirements can say a register must exist but cannot say what must be in it; what belongs in it can only be found by reading the project record. The draft asks for traceability (9.2) and for registers to name their sources (9.6), but it does not ask anyone to read a compiled register back against the record it was drawn from, so nothing in the project’s own verification confirms that such a register is complete.
Where AI compilation will land, and what it does to the reviewer
This is important because it is the content people will want to use AI to compile, because it is the thousands-of-documents part of the pack, and as a result it is where the behavioural problems will be the hardest to guard against. The people compiling it are designers, contractors and information managers whose job is producing and passing on this information, for whom checking feels like overhead, even if necessary and critical overhead.
The AI promise of compiling with a tool and “fixing the remaining few per cent” is the path of least resistance and would be seen as welcome by many. In this context, fixing the remainder means reviewing the tool’s output rather than reading the sources, and vigilance in reviewing output is well documented to fall as the tool appears more capable, and, critically, to fall without the reviewer being aware of it. Once a certain accuracy threshold is reached, reviewer confidence increases, automation complacency increases, and review skills can atrophy as a result.
Worse, the omissions are the hardest to spot, because items the tool left out give the reviewer nothing to look at, and over time the problem compounds. The ability to notice that a register is incomplete comes from having worked through material of that kind, and if the compiling is delegated, that familiarity stops being built. The record would still carry its signature, provenance and traceability, but the competence that gives those attributes their value would have thinned, and nothing in the record would show it. Professionals aware of this phenomenon are less likely to fall for these failures, but current industry discourse is far too quiet on this and it remains a growing risk area.
Deterministic and probabilistic compilation
Whether this matters depends on how the compilation is done. A deterministic process can be tested, its errors corrected through its logic and its output reconciled against a known rule. A probabilistic one, as most current AI systems are, makes a fresh judgement on every item, so confirming that each item was compiled correctly becomes the human’s task, which is the monitoring task described above. AI run alongside an existing review reduces risk; AI run in place of the first read, with a human fixing the difference, is the direction of travel much of the industry discussion assumes and can, counterintuitively, substantially increase the human scope of a review task and make proportionate review a much muddier concept.
Three suggestions

1. 9.5 should say what verification means for content whose completeness can only be established from the project record, since matching the building does not apply. A competent, independent person should check the compiled register against the documents it was drawn from and record the evidence chain, starting from the documents and confirming that each relevant item reached the register. Checking the other way, from the register back to the documents, only shows that what is there is right; it does not show that nothing is missing.
2. Where a tool did the compiling, the record should say so under section 7’s first question, and say whether the tool makes a judgement on each item or applies a repeatable rule. If it makes a judgement, the compiled content should carry the draft’s existing “unverified” marking (8.3, 9.6) until someone has done the check above in full detail and reconciled that autonomous judgement with their own.
3. Phase two should consider which compilation methods are acceptable, and how the people signing off keep the competence to notice what a tool has left out, or tackle convincing-looking errors.

The likely objection
The likely objection is that the source documents were already signed off by the designer or contractor. That sign-off vouches for the documents. It does not vouch for whether the register reflects them as a competent person reading the originals would have understood them. Nor are the designer and contractor immune: they may meet the same automation complacency when producing those documents, so errors missed at their sign-off pass through to every later stage. This is a conceptual problem for the lifecycle of information on a project, not only for its finalisation at handover, which is why addressing it is vital. Prompt and document injection, where text inside a source document steers or corrupts the tool’s output, is a separate and more complex facet of the same problem and not addressed here.
Further research
The phenomenon of AI omissions for professional reviewers is set out in a preprint, The Omission Gap (SSRN, 2026), and I am continuing research at the intersection of professional competence, document integrity and AI systems. I would be more than happy to discuss this further if it would be beneficial to the consultation.
You can message info@vbrief.ai to reach me.
James Wright-Zhang

There needs to be a consideration of the production of appropriate design and design check calculations, to be provided as part of the as-constructed information. Below are some pointers.
NR standards include provision of suitable as-built drawings and design and check calculations to allow future use.
A common-sense approach is needed, with some suppliers offering meaningful packs which include key outputs of calculations where they are critical to safe future use, maintenance, alteration, or demolition,

• Information on design loadings, 
• Key calculations demonstrating structural strength and stability,
• Utilisation factors,
• Ground investigation information,
• Construction sequence,
• Fire engineering design
• Departures from standard design requirements

Some suppliers produce reams of wholly inadequate outputs which are meaningless to anyone looking at them in future, hence are useless. 
A test is “ Will someone need this calculation (or its outcome) to avoid risk in the future”. Guidance on what a calculation pack within a H&S file should contain is welcomed, but here is a quick summary suggestion.
“The CDM Health & Safety File should not replicate design calculations but must clearly identify and communicate all safety-critical design assumptions, limits, and constraints derived from those calculations that are necessary to ensure the safe operation, maintenance, alteration, or demolition of the asset”

Network rail have clear expectations for H&S files (including as constructed drawings) in standards NR/L2/OHS/0047, and NR/L2/INF/02202. Suggest a paragraph is added referencing other such documentation or acknowledging such standards exist e.g. add thus
“This document compliments other documents. For instance, Network rail have clear expectations for H&S files (including as-constructed drawings) in standards NR/L2/OHS/0047, and NR/L2/INF/02202″

Our principal argument set out in the file note we produced for delineating and defining scope of For Record Drawings is that consultant Record Information and physical verification of the completed works are different and discrete evidential functions, and should remain so. The note defines consultant information as the final coordinated design intent, updated for formally notified and approved changes, while expressly excluding continuous supervision, comprehensive dimensional verification and certification of every installed element.
 
What the guidance seeks to establish is a somehow all-encompassing unified record where distinction between functions seems blurred and therefore either renders responsibilities null or infers total responsibility for all aspects on a specified agent (client / PC / PD / other dutyholders), depending on how you view it, which flies in the face of the regulations key tenets of sequential responsibilities discharged by competent actors towards a holistic record – Golden Thread.
 
What it seems to legitimately be trying to achieve is trustworthy, usable lifecycle information. 
 
Its difficulty is that Sections 6–9 and Appendix 1 repeatedly moves between:

• an overarching composite handover record;
• individual drawings, models and data;
• verification of documentary requirements; and
• verification of the physical works.

Terms such as “authoritative record”, “verified and validated record information of what has actually been built”, “As-Built Drawing”, and “As Constructed Record Document” therefore risk attaching the assurance status of the whole package to each consultant-produced record drawing.
Recommended consultation position: 
Retain a composite, evidence-led objective, but expressly define consultant’s For Record Drawings / Record Information as one constituent source, explicitly stating that an individual designer’s record information is not certification of the physical installed condition unless a separately scoped, competent verification service has been commissioned and performed. This may result in the light of responsibility for this information falls on the PC or client, but we don’t feel that it should as it is not necessary or supported by the regulatory framework.
What is the legislative position:

• Regulation 31 requires approved and revised documents, change records and compliance evidence to be maintained through the Golden Thread.
• Regulation 38 transfers the completion application, accompanying documents and relevant operational information to the relevant person.
• Regulation 40 requires the latest agreed documents, plans necessary to show compliance, the completed change-control log and PD/PC duty-fulfilment declarations.

The BSA-2022 enables regulations covering certificates, information creation and retention, and inspection/testing – which taken together suggests ‘an authoritative record’, however it does not itself appoint the architect (as PD, or another design dutyholder) as universal verifier of the completed asset, which clearly is beyond their independent design scope or expertise.
 
There is nothing in the legislation that determines that any one dutyholder has an obligation to prove that contractors installed every element in accordance with their design.
 
Equally, as Pricipal Designer, the GW3 declaration is not to be worded as an architect’s certification of dimensional probity, workmanship, concealed construction or correctness and appropriateness of all installed products.
 
Likewise, the Principal Contractor is not require to re-survey every trade installation; the legislation determines that the PC is responsible for supporting a managed evidence chain using competent contractors, specialists, inspections and tests, in conjunction with technical and testing standards and approvals for the materials and product specified and/or controlled through the orchestrated change protocols (and records).
 
The BSR’s fundemental remit is to inspect the completed work and issue the completion certificate only if, after taking all reasonable steps, it is satisfied that the work complies, prescribed information is complete and accurate, and required information has been handed over. This does not constitute a requirement to insist on reconstituting all contributors inputs, across the entire and lengthy process (Stage 4-6, over sometimes.many years), into an authoritative ‘As-built’ production. Their final assessment is not replaced by a designer’s record drawing; even their Completion Certificate is evidence, but not conclusive evidence, of compliance.
 
The fundementals underpinned in regulation…..
 
There is no provision requiring the architect to certify that its drawings physically match every installed condition. 
 
This conclusion is interpretive, but follows directly from the allocation in the regulations: 

1. Designers are responsible for compliant design and sufficient design information; 
2. the PD manages design compliance; 
3. the PC manages and evidences construction compliance; 
4. the client owns the completion application and information arrangements; 
5. and BSR makes the regulatory determination after inspection post-completion.

Composite traceability is the essence of the golden thread evidence rather than Composing reality….
 
This in essence gathers together the approved design, authorised changes, construction records, installed systems, tests, non-conformances and the accepted baseline. This is substantially the evidential suite proposed in our previous file note outlining ‘For Record’ suites of information from relevant dutyholders.
 
A reasonable analogy (perhaps not a brilliant one) would be that of an auditor (BSR) looking at a company’s financial report at the end of the year (post-completion) – they aren’t required to observe each and every receipt for each and every purchase, they are required to evaluate and enquire in order to satisfy themselves, taking all reasonable steps, that the book represents the actuality. The responsibility rests with the company (Dutyholders) that their report is accurate. But its the process across the year that’s important (The Golden Thread) not some retrospectively produced overarching composite record that attempts to bind every element.

Many thanks for forwarding, George. Admittedly I am time poor but a few comments / observations.
Have The Higher-Risk Buildings (Keeping and Provision of Information etc.) (England) Regulations 2024 been considered? I cannot find it referenced in the draft. Whilst I accept the work deliberately zooms out beyond HRBs, for completeness I think it should be referenced (particularly Schedule 1).
On that subject, I welcome the decision to zoom out (which we touched on in the original business case discussions) . I have thought for a long time that the ‘golden thread concept’ should be applied proportionately to any scheme. The draft is right to frame the question around what information is proportionately required for the safe and successful operation of an asset, rather than around the discharge of individual regulatory obligations.
As-built information has been a point of discussion for many years and a source of frustration in practice. People have genuinely read “final construction issue” and “record drawings” as meaning as-built.
Over the last nine plus years, where property owners have had cause to carry out building safety investigations, the record has been either missing or unlocatable and when found it more often than not accurately reflected what was actually built. That is before dealing with the discipline needed to keep information current through the occupation phase and the ‘changes’ a building undergoes. Removing the ambiguity in the term is a necessary first step.
The HOW will be difficult! A significant element of the HOW is producing an output with sufficient detail for those instigating the work. But before that, the WHY has to be sold to those holding the purse strings, meaning asset owners and operators. (which is why it is so great to see many who fall into that category in the list of supporters / contributors!)
We flagged in 2024 that OPEX requirements are not cascaded down the supply chain, because the end user does not fully understand them, cannot communicate them, or is focused on immediate CAPEX.
Marc Bradfield
NCG

Why are the emojis in the comments so massive? It makes it difficult to read the comments.

There is a lot of information handed over at the end of a project – not all of it what the industry would define as ‘as-built’. The document recognises this to some degree in the content, but retains the title -as-built’ which I think is a bit confusing as the section on ‘as-built’ is actually quite small and does not deal with the practical difficulties experienced in its definition and preparation.