ARCHArchitecture & intelligence

Engineering intelligence across systems.

A domain-aware architecture for connected decision-making.

DVA's architectural approach combines governed data integration, construction-domain modelling, knowledge relationships, deterministic rules and AI-assisted reasoning. It is designed to work with existing systems of record rather than require an organisation to replace its established technology investments.

  1. Authorised sources
  2. Governed integration
  3. Construction ontology
  4. AI & specialist reasoning
  5. Decision intelligence
  6. Human-approved action

01Construction ontology

Connecting the built environment through ontology.

A drawing revision, an RFI, a procurement order and a construction activity may be related, but those relationships are rarely visible in one place.

A shared model of the project

A construction ontology provides a shared way to represent projects, spaces, work packages, assets, documents, activities, organisations and their relationships.

  • Projects
  • Spaces
  • Work packages
  • Assets
  • Documents
  • Activities
  • Organisations
Evidence kept, assumptions labelled

DVA uses this approach to establish meaningful connections between information held in different systems, while preserving source evidence and distinguishing verified relationships from assumptions.

See an illustrative relationship graph

02Layered architecture

Seven layers between your systems and a decision.

Select a layer to see what it does and how one fictional scenario passes through it: a chilled-water installation where a tender, a drawing revision, an RFI and a purchase order may or may not affect each other.

  1. Your systems: email, Excel, DWG, PDFs, CDE, BIM, RFIs, ERP, takeoff, schedules, field apps.
  2. Governed connectors: APIs where approved, file imports, credentials, permissions.
  3. Canonical records & evidence: project, object IDs, source, time, revision, lineage.
  4. Construction ontology: project, work package, location, BIM, RFI, tender, labour, purchase order.
  5. Reasoning: identity checks, exceptions, capacity rules, specialist results, AI synthesis.
  6. Decision & action: evidence inspector, unknowns, reviewer permission, approve / reject / request.
  7. Client workspace: executive overview, project view, workflow queues, reports, alerts.
A layer over your systems, not a replacement

DVA can work as a federated intelligence layer across several authoritative systems, rather than one giant database that replaces them. Whether to build a central repository or data lake is your decision, made after a review of rights, data locality and security.

Built for enterprise adaptability

DVA's intelligence architecture can be designed around existing enterprise data and AI platforms, selected specialist applications or a combination of approved technologies. The result is a client-specific approach to connecting information, investigating operational constraints and supporting decisions within the organisation's governance requirements.

Enterprise platforms are architectural options for your environment, not DVA partnerships.

03Governance

Your systems remain your systems.

Source provenance, permissions, project boundaries, uncertainty and human approval are designed into every proposed workflow, not added at the end.

  • Source provenance

    Every record keeps its source system, record ID, time and revision. You can always open the original.

  • Permissions

    Only sources you have authorised, seen only by users allowed to see them. A source without granted rights is quarantined, not used.

  • Project boundaries

    Records stay inside the projects and companies they belong to.

  • Uncertainty shown

    Candidate, ambiguous and unknown links are labelled. Where there is no evidence, no link is drawn.

  • Human approval

    No autonomous approvals, bids, purchases or professional judgements. A named reviewer approves, rejects or asks for more evidence.

Five things we never collapse into one green "verified"

A record being real doesn't make a link real. A link doesn't prove a consequence. A suspected consequence isn't a calculated impact. And none of it is a decision until someone authorised makes it.

Five separate states, with an example from the fictional demo scenario
StateWhat it meansIn the demo scenario Fictional
1Verified record identityThe record is what it says it is: source, ID and revision confirmed.PO-1007 is confirmed in the procurement system.
2Verified relationshipExplicit source evidence links two records.The schedule's material requirement for ACT-MECH-20 cites PO-1007.
3Suspected consequenceSomething might be affected. It's flagged for review, not asserted.RFI-021 and PO-1007 both concern Level 3 chilled-water items, but neither cites the other. Candidate only.
4Specialist-calculated impactA specialist tool or qualified person has calculated the effect.Not yet: quantity, schedule and cost impacts of drawing M-204 Rev C wait for the estimator and planner.
5Authorised decisionA named person with the right permission has decided.Pending: a manager chooses between the bid and staffing options.

Every decision ends at a person

  1. EvidenceSource records, revisions, reasons for each link
  2. UnknownsWhat is missing or unconfirmed, stated plainly
  3. ReviewerThe person with permission for this decision
  4. DecisionApproveRejectRequest more evidence

04Integrations

Connect what you have, by the route you approve.

Specialist vendors and the systems you own keep their domain authority. We connect them only where your contracts, licences and the vendor's terms allow it.

Route 1

API, where approved

A direct connection, used when both you and the vendor permit it and the API rights are confirmed.

Route 2

File import or export

Registers, reports and spreadsheets exported from the system that owns them, on an agreed schedule.

Route 3

Authorised everyday sources

Inboxes, shared folders and site records you authorise, for teams starting from email and Excel.

  • Source of truth stays put. Each application remains authoritative for its own records.
  • Read first. Anything written back to your systems goes through an approved automation pathway, with a person's sign-off.
  • Credentials per source. Access is granted, scoped and reviewable for each system.

Systems we aim to connect

  • Email & documents

    Tender inquiries, PDFs, Excel, shared file storage

  • Drawings & CDE

    Drawing registers, revisions, RFIs, submittals

  • BIM models

    Element identities, zones, model revisions

  • Takeoff & estimating

    Accepted quantities, BOQ items, labour-hour assumptions

  • Scheduling

    Activities, dependencies, milestones, lookaheads

  • ERP & procurement

    Purchase orders, deliveries, supplier confirmations

  • Field & QA/QC

    Inspections, progress records, defects, site photos

  • Workforce

    Crews, skills, assignments, capacity

These are technology categories, not partners or live integrations. How a technology moves from "under evaluation" to "implemented" or "partner".

05Engagement model

Begin with the challenge. Engineer the solution.

Every engagement starts the same way: an introductory discovery, a scoped diagnostic, then a measured pilot. Behind those three steps is a five-stage ladder, climbed one workflow at a time. You can stop after any stage with something useful in hand.

  1. 01

    Introductory discovery

    A 30-minute call about one decision your team finds hard to make today, and the systems and records behind it.

    Before stage 01

  2. 02

    Scoped diagnostic

    One workflow examined end to end, ending in a workflow map, a systems and data inventory, a measurable baseline and a pilot proposal.

    Stages 01 Discover · 02 Design

  3. 03

    Measured pilot

    One bounded workflow on agreed connectors, with exception detection and decision briefs, measured against the baseline.

    Stages 03 Connect · 04 Intelligence

After a pilot, stage 05 Enable & Improve covers training, operating procedures and further workflows. Scope, timing and fees are agreed in writing before each step. Results depend on your systems, data access and pilot validation, and aren't guaranteed.

Stage 01 · Discover

You receive

  • A map of the workflow as it really runs
  • A diagnosis of pain points and digital maturity
  • The target decision, and who owns it
  • An inventory of systems and data
  • A measurable baseline

What DVA does

  • Interviews the people who make and feed the decision
  • Maps the process and any ITT/IDD obligations
  • Defines data rights and feasibility

Next step

Start with a problem-led conversation, not a product pitch.

Book a Discovery Call
Stage 02 · Design

You receive

  • A bounded solution blueprint
  • An integration architecture
  • Technology evaluation and specialist-tool recommendations
  • Outcome metrics agreed up front

What DVA does

  • Decides what to build, buy or integrate
  • Identifies which system is authoritative for each record
  • Sets security requirements and approval gates

Next step

Agree what a pilot covers and how it will be judged.

Discuss Pilot Scope
Stage 03 · Connect

You receive

  • Agreed connectors or approved import paths
  • Canonical records across systems
  • Evidence lineage for every record
  • Working dashboards

What DVA does

  • Configures software and connectors
  • Reconciles IDs, revisions and permissions
  • Brings in historical data where it's useful and permitted

Next step

See which of your systems connect, and by which route.

Request Integration Plan
Stage 04 · Intelligence

You receive

  • Exception detection
  • Relationships across domains: tender, drawings, RFIs, procurement, schedule, crews
  • Decision briefs
  • Approved automation pathways

What DVA does

  • Applies deterministic checks and rules
  • Uses results from your specialist tools
  • Adds governed AI synthesis, with human approval on every decision

Next step

See the decisions this stage is built around.

See Use Cases
Stage 05 · Enable & Improve

You receive

  • Training for the people who use it
  • Measurement against the baseline
  • Operating procedures and support
  • Monitored improvements

What DVA does

  • Validates outcomes against what was agreed
  • Teaches your team to reuse the capability on other projects

Next step

Plan how the capability rolls out beyond the pilot.

Discuss Implementation
Path A · Start with what you have

Begin from authorised email, Excel, DWG/PDF, file storage and simple site records. Structured data and specialist tools come in only where ROI and tender requirements justify them.

Path B · Connect the mature stack

Integrate the BIM/CDE, estimating, scheduling, ERP, procurement and field tools you already license into one governed management view. Each application stays the source of truth.

See the architecture on a worked scenario.

Follow one fictional chilled-water scenario from source records to a decision a named person approves.