State of the Industry

30 construction technology products mapped across Permission, Promise and Prevention

13 min read · Research checked 20 August 2026

Part II: 30 construction technology products and platforms, audited as examples

Introduction

Part I introduced the 3Ps of Data Control: Permission, Promise and Prevention. Now we put the framework to work, examining how 30 construction technology products handle customer data, what their suppliers are allowed to do with it, and how much control customers retain.

The findings reveal considerable variation, even within individual products. Depending on the feature, deployment model or contract, the same software may allow data reuse, restrict processing through contractual promises, or keep project information within a customer-controlled environment.

Why these products?

We selected 30 construction technology products and platforms to illustrate the range of ways project data is handled across the industry. The selection includes major cloud platforms, desktop applications, locally hosted systems and AI-enabled tools, with different approaches to data reuse, external processing and customer control.

Many of these products will be familiar to construction professionals. Examining their published terms, features and deployment options makes the framework tangible and shows how significantly data control can vary, even between products that appear to offer similar services. The selection is intended to illustrate the range of approaches in the market, rather than provide a representative sample or rank suppliers.

How to read this report

The profiles are based on first-party terms, privacy policies, security pages, product documentation and AI notices reviewed on 20 August 2026. This is an audit of publicly available documentation, rather than an independent technical assessment or legal opinion. Statements such as “no training”, “zero retention” or “customer-controlled” are treated as supplier claims unless independently verified.

Each assessment covers the construction software provider and the third-party services it selects or integrates into its product. External AI services independently selected and contracted by the customer are considered separately. The same product may appear in more than one position because its data practices can vary by feature, deployment model, jurisdiction or contract. The unit of analysis is the data flow, rather than the brand. Published terms provide a starting point, but enterprise agreements, order forms, data-processing agreements and amendments may establish different rights and obligations. Products and policies also change over time, so the findings reflect the information available at the review date. Before relying on any assessment, check the current terms and the specific configuration you intend to use.

The 30-product map

The map shows how the 3Ps of Data Control apply to 30 construction technology products and platforms. Read it from left to right: from Permission, where suppliers may reuse customer information beyond delivering the service, through Promise, where processing is restricted by contractual commitments, to Prevention, where customers can keep project data within an environment they control.

Some products appear in more than one position, reflecting differences between features, deployment options or contractual terms. The map is organised by data-control position, not by product quality, security or overall performance. Products sharing a position are not ranked against one another.

The useful unit of analysis is the data flow, not the company or even the product.

How far can your construction data travel

Swipe sideways to see the full map →

Product
PermissionPromisePrevention
Raw reuse
Raw reuse
Raw reuse to Derived reuse (ACC cloud); safeguards towards Restricted processing (reuse right remains) (feature safeguards)
Raw reuse to Derived reuse (platform / internal AI); safeguards towards Restricted processing (reuse right remains) (feature safeguards)
Derived reuse
Derived reuse
Derived reuse
Derived reuse
Derived reuse
Raw reuse; safeguards towards Restricted processing (reuse right remains) (training controls / opt-out)
optional: Raw reuse (AI training opt-in); Restricted processing (core hosted iTwin)
Raw reuse (US terms); Restricted processing (UK base, AI off)
optional: Raw reuse (AI training opt-in); Restricted processing to Customer-managed (hosted / customer-managed)
optional: Derived reuse (training opt-in); Restricted processing (default inference)
optional: Raw reuse (ACC-connected); Local & offline (local files)
optional: Raw reuse (Autodesk cloud); Customer-managed to Local & offline (file server / local)
optional: Raw reuse to Derived reuse (Studio / cloud + optional AI); Local & offline (Revu Offline)
optional: Raw reuse (AI Visualizer); Restricted processing to Local & offline (BIMcloud / local)
optional: Raw reuse (Takeoff Boost); Local & offline (manual local OST)
optional: Raw reuse (Takeoff Boost); Local & offline (local PlanSwift)
optional: Raw reuse (web / Connect / AI); Local & offline (Desktop local)
optional: Raw reuse (CheckPoint SaaS); Local & offline (Security+ offline)
optional: Raw reuse (cloud Model Sharing); Customer-managed to Local & offline (on-premises / local)
optional: Raw reuse (published cloud project); Local & offline (Do not publish)
Restricted processing
Restricted processing
optional: Derived reuse (RIB-managed cloud); Customer-managed to Local & offline (self-managed / local)
Restricted processing (Oracle-hosted EPPM); Customer-managed to Local & offline (customer-hosted / SQLite)
Local & offline
Local & offline
Documented positionOptional mode: an opt-in, add-on or cloud route you can avoidOne mode spanning positionsDifferent modes of the same productSafeguards that do not remove the reuse right
Hover over a point to see the mode it represents; select a product name to open its profile. Customer-selected external AI is assessed separately.

Each point represents a documented data-control position. Where a product has more than one point, the connector shows distinct features, deployments or contract choices within the same product; it does not imply that every position in between is available.

The 30 product profiles

Each profile explains the product’s position on the map, based on its published terms, data practices and deployment options.

Where customers can influence how their data is handled, we include a Practical reading: specific choices that may change their level of control, such as negotiating contractual terms, opting out of data reuse, disabling certain features or choosing a different deployment model.

OpenSpacePermission (raw-content reuse)

OpenSpace’s terms allow it to use customer project data to improve its services. They also allow it to retain and use derived data for AI development and commercial purposes.

OpenSpace has strong security controls, including SOC 2 and FedRAMP, but those controls protect the data; they do not remove the permission to reuse it.

Sources: terms, technology

Togal.AIPermission (raw-content reuse)

Togal.AI’s terms allow uploaded construction content to be used to develop its machine-learning models. The permission applies to the uploaded content itself, not only to anonymised or derived data.

Its security controls govern how that data is handled, but the standard terms still allow the reuse.

Sources: terms

Autodesk Construction CloudPermission, with stronger Promise controls

Autodesk’s General Terms allow Autodesk to use customer content to improve its offerings. Autodesk also publishes detailed security controls and feature-level AI transparency information.

Those controls do not remove the underlying permission in the standard terms. A signed enterprise agreement can override the General Terms, so the customer’s actual position may depend on the contract it has negotiated.
(General Terms §§17.12 and 18, AI Transparency Cards)

Practical reading: If secondary use is not required, the restriction needs to appear in the controlling enterprise agreement rather than relying only on individual AI settings.

Procore platform and AIPermission, with stronger Promise controls

Procore’s public agreement allows Customer Data to be used to provide and improve its services. Its AI documentation also allows some inputs and outputs to be used to improve internal models.

Procore publishes stronger controls for particular AI features, including restrictions on external providers and project-level AI administration. Those controls narrow some uses, but they do not remove the broader improvement right in the base agreement.

Sources: SSA §11.11, AI FAQ, AI administration, AI Transparency

Practical reading: If the customer wants a Promise position, the restriction should be written into the controlling Order or amendment. Turning AI off is not the same as removing the contractual permission.

BuildotsPermission (anonymous or derived reuse)

Buildots limits use of raw project data to providing the service, but it retains rights over anonymised visuals, metadata and analytics for development and research.

The key question is what “anonymous” means. Its definition focuses on whether a person can be identified, which does not necessarily remove information about the project, sequence or assembly.

Sources: US terms, Intelligence Lab

DaluxPermission (derived reuse)

Dalux preserves customer ownership of uploaded material, but its terms allow it to create and use derivatives for improvement, testing and operation.

Its DPA and security controls protect data custody. They do not remove the separate permission to use derived information.

Sources: General Terms, DPA

Document CrunchPermission (derived and usage data)

Document Crunch’s base agreement allows it to create non-identifying algorithm, trend and usage data from customer information.

Its Data Use Amendment narrows those rights by removing some algorithm and trend uses, but it still permits broad non-identifying Usage Data. The amendment reduces the scope of Permission without removing it completely.

Sources: Customer Agreement, Data Use Amendment

FieldwirePermission (anonymised reuse)

Fieldwire’s terms allow Customer Data to be reused in anonymised form for product development, tailoring and market research, including after the contract ends.

The definition of anonymity focuses on whether an individual employee can be identified. It does not necessarily mean that the project, site, assembly or method can no longer be identified.

Sources: US agreement, European agreement

qbiqPermission (anonymised and derived reuse)

qbiq requires uploaded plans to be anonymised and grants itself rights to use that material for further development. It also owns anonymous analytics and machine-learning data created from the service.

The difficult part is the definition of anonymity: a spatial plan may remain technically or commercially revealing even after obvious project identifiers have been removed.

Sources: Terms

Practical reading: Define what “anonymous” must mean at project and assembly level before relying on it as the condition for upload.

BuildCruxPermission, with training controls that strengthen the Promise

BuildCrux’s terms allow raw customer data to be used to provide and improve the service. That keeps the base position in Permission.

It also provides more specific protections around model training: first-party training is de-identified by default, administrators can stop new training uses, and external AI providers are required not to train on submitted data.

(BuildCrux terms)

Practical reading: Use the training opt-out, but if raw-data improvement is also unacceptable, that restriction needs to be written into the contract.

Bentley iTwin platform and AIPermission (training opt-in) / Promise (core hosted iTwin)

Bentley’s core hosted iTwin service restricts customer data to providing the service and applies confidentiality protections, which puts the normal hosted service in Promise.

A separate AI agreement allows customers to opt in to sharing Subscriber Data for Bentley model training. That moves the specific training use into Permission.

Sources: Enterprise 365 terms, AI terms

Practical reading: Treat the training contribution as a separate decision rather than part of ordinary platform use.

PlanRadarPermission (US terms) / Promise (UK base with AI off)

PlanRadar’s public US terms allow Customer Data to be used for internal research, development and service improvement, which creates a Permission position.

Its UK base platform, when AI is disabled, does not contain the same reuse right and is supported by confidentiality and processor terms, placing that route in Promise. AI-enabled UK use needs separate review because PlanRadar says it may retain prompts and responses to improve user experience.

Sources: US terms, UK terms, AI privacy

Practical reading: Specify the governing regional terms and AI state in the order instead of procuring the PlanRadar name in the abstract.

Bentley ProjectWisePermission (AI training opt-in) / Promise (hosted) / Prevention (customer-managed)

Bentley-hosted ProjectWise operates under service-only and confidentiality restrictions, which places the hosted route in Promise.

ProjectWise can also run on customer-managed infrastructure, creating a Prevention route. Separately, customers can opt in under an AI agreement to share data for Bentley model training, which creates a Permission route.

Sources: Enterprise 365 terms, AI terms, hosting

Practical reading: Treat the training addendum as a separate, deliberate information-rights decision rather than routine feature enablement.

Trunk ToolsPermission (training opt-in) / Promise (default inference)

Trunk Tools says its default AI inference uses external model providers under zero-retention and no-training restrictions. That places normal inference in Promise.

Customers can separately opt in, in writing, to contribute de-identified data to develop Trunk’s own construction models. That specific choice moves the data into Permission.

Sources: data security FAQ

Practical reading: Leave training consent off unless the organisation has confirmed both its authority to contribute the data and the effectiveness of the promised de-identification.

Autodesk NavisworksPermission (ACC-connected) / Prevention (local files)

Navisworks Manage and Simulate can work with coordination files stored locally or on customer-controlled storage, which creates a Prevention route.

When those workflows are connected to BIM 360 or Autodesk Construction Cloud, the uploaded content becomes subject to Autodesk’s broader content-improvement terms, moving that route into Permission.

Sources: local workflow, ACC integration, General Terms

Autodesk RevitPermission (Autodesk cloud) / Prevention (file server or local)

Revit can keep RVT and worksharing data on local or customer-managed storage, which allows a Prevention position.

When the project uses Autodesk-hosted collaboration or cloud AI features, the applicable Autodesk terms allow broader use of customer content for improvement. Those routes therefore move into Permission.

Sources: General Terms, collaboration methods, Optimize Annotations card

Bluebeam Revu and StudioPermission (Studio, cloud and optional AI) / Prevention (Revu Offline)

Bluebeam Revu Offline is designed to operate without communicating with Bluebeam’s cloud services, creating a clear Prevention route.

Studio, cloud services and some AI-enabled features operate under terms that allow broader use of customer content for improvement, analytics and AI-related purposes, which creates Permission.

Sources: General Terms, addendum, Revu Offline

Practical reading: If structural prevention is required, procure the documented offline product rather than assuming that ordinary Revu use is equivalent.

Graphisoft Archicad and BIMcloudPermission (AI Visualizer) / Promise or Prevention (BIMcloud and local modes)

Archicad files and customer-hosted BIMcloud deployments can remain inside customer-controlled infrastructure, creating Prevention.

Graphisoft also offers hosted services and AI features with different terms. AI Visualizer allows prompts to be stored and used for service improvement and development, creating Permission, while the newer AI Assistant publishes stronger restrictions on project-content training.

The result depends on which Graphisoft service and AI feature is actually being used.

Sources: Archicad EULA, BIMcloud EULA, AI Assistant

On-Screen Takeoff and Takeoff BoostPermission (Takeoff Boost) / Prevention (manual local OST)

Manual On-Screen Takeoff can keep plans and its database on customer-controlled storage, creating a Prevention route. Takeoff Boost uploads plans under ConstructConnect terms that grant broad rights over Project Material and treat uploaded Subscriber Data as non-confidential. That cloud AI route sits in Permission.

Sources: Boost workflow, local database, MSA

Practical reading: Do not enable Takeoff Boost for restricted project information simply because the underlying desktop product can operate locally.

PlanSwift and Takeoff BoostPermission (Takeoff Boost) / Prevention (local PlanSwift)

Standard PlanSwift can keep project folders on a local computer or customer network, which supports Prevention. Takeoff Boost uploads plans under ConstructConnect’s broader cloud terms, moving that workflow into Permission.

Sources: data storage, Boost, MSA

Practical reading: Keep Takeoff Boost out of the deployment where the organisation does not have authority to submit the drawings under those terms.

SketchUpPermission (web, Connect and AI) / Prevention (Desktop local)

SketchUp Desktop can save models locally and operate offline, creating a Prevention route.

Web use, Trimble Connect and online AI features place submitted content under terms that allow customer data to be used to improve Trimble and affiliate offerings. Those routes sit in Permission.

The position therefore depends on where the model is stored and which online features are used.

Sources: Trimble terms, saving models, offline use

Solibri Desktop and CheckPointPermission (CheckPoint SaaS) / Prevention (Security+ offline)

Solibri Security+ can perform model checking locally without an internet connection, creating Prevention.

CheckPoint is a SaaS service operating under terms that grant broad rights over Statistical Information derived from Customer Data. That route sits in Permission.

The local and SaaS products therefore create very different data-control positions.

Sources: Security+, CheckPoint terms

Tekla Structures and Model SharingPermission (cloud Model Sharing) / Prevention (on-premises or local)

Tekla Structures can operate with local or customer-hosted model storage, which supports Prevention.

Tekla Model Sharing uses Trimble’s cloud services. The applicable online terms allow customer content to be processed into derivatives for service improvement and development of other products, moving the cloud route into Permission.

Sources: Tekla Online terms, on-premises sharing, sharing guide

Trimble ConnectPermission (published cloud project) / Prevention (Do not publish)

Trimble Connect allows a genuinely local “Do not publish” project that remains on the user’s machine, creating a Prevention route.

Published cloud projects are governed by terms that allow model-training and product-improvement uses, which puts those workflows in Permission. Ordinary Offline Mode does not create the same result because it later synchronises.

Sources: Connect terms, local project, offline mode

Practical reading: Use an unpublished “Do not publish” project if the requirement is to keep project data local.

Oracle AconexPromise

Aconex’s core terms limit use of project content to providing the ordered service. Oracle also publishes regional hosting and organisational security controls around that custody.

That places the core Aconex service in Promise. Separately purchased Oracle analytics, AI or integrations create additional data flows that need to be assessed on their own.

Sources: cloud agreement, Aconex security

ReviztoPromise

Revizto’s EULA does not grant a general right to use project data for training or service improvement. Its security documentation also says staff do not access project information unless the customer makes it available for support. Project information still sits in Revizto-controlled cloud systems, so this remains Promise rather than Prevention.

Sources: EULA, security and sovereignty

RIB CostXPermission (RIB-managed cloud) / Prevention (self-managed or local)

Standalone and customer-managed CostX deployments can keep estimating data on customer-controlled infrastructure, creating Prevention.

RIB-managed CostX Cloud retains rights over de-identified or aggregated data derived from Customer Data. That hosted route therefore sits in Permission.

Sources: deployment options, cloud brochure, legal resources

Practical reading: Choose customer-managed or standalone deployment if ownership of derived estimating intelligence is unacceptable.

Oracle Primavera P6Promise (Oracle-hosted EPPM) / Prevention (customer-hosted or SQLite)

Oracle-hosted Primavera P6 EPPM operates as a cloud service under Oracle custody, placing that route in Promise.

P6 can also be customer-hosted, and P6 Professional supports a local single-user database. Those deployments can move the same product into Prevention.

Additional cloud, gateway or AI integrations create separate data flows and need to be assessed separately.

Sources: standalone installation, cloud/on-premises integration

Purple HammerPrevention

Purple Hammer says its normal project workflow is local, with AI disabled by default. It also states that project content does not enter Purple Hammer’s servers or train its models.

External AI, if selected by the customer, is a separate relationship rather than part of Purple Hammer’s own data path.

Sources: legal material

Practical reading: Because the architecture is self-published, the claimed boundary should be checked against the behaviour of the deployed application.

TestFit MCPPrevention

TestFit says its desktop engine does not send project information to the assistant, receive prompts, store conversations or train on the design.

That keeps the TestFit side of the workflow in Prevention. If the customer connects a separate AI assistant, that provider creates a separate data flow and may sit anywhere within the 3Ps.

Sources: MCP, data explanation

Practical reading: Assess the selected AI provider separately; TestFit’s local architecture does not determine what that external provider does with the information it receives.

What the 30 products reveal

cross the 30 products reviewed, one finding stands out: the level of control customers have over their project data depends on the specific feature, deployment model and contract, rather than the software brand alone.

The same product can offer very different levels of control. Project information may remain on a local device, enter a cloud platform, pass through external AI services or become subject to broader reuse rights. Each route carries different implications for who can access the information and what they are permitted to do with it.

Understanding these differences is essential when choosing construction technology. Part III turns the framework into a practical procurement checklist, helping buyers ask the right questions before committing their project data to a new system.

About the research

The profiles are based on first-party terms, privacy policies, security pages, product documentation and AI notices reviewed on 20 August 2026. They are a public-document audit, not a penetration test or legal opinion.

Let’s talk!

Questions about the research, or about how your product handles customer data? We’d like to hear from you.