AI Built for enterprise document intelligence

Make every legacy document searchable, structured, and ready for action.

DocRocket is an enterprise-grade AI portal that reads, classifies and indexes your full back-catalogue of documentation — scanned, handwritten, or born-digital — and pushes clean, structured outputs into your operational systems.

62
Document types pre-loaded
17
Engineering disciplines
4
Role-based access tiers
100%
Audited & encrypted
The challenge

Decades of documentation. Locked in PDFs, scans, and filing cabinets.

Legacy documentation is the single biggest blocker to a modern DMS migration. The information is there — it's just not readable by software. DocRocket fixes that.

The solution

A standing operational portal — not a one-time migration script.

DocRocket combines AI-driven OCR, handwriting recognition and document classification into a single enterprise web portal. Configure once, then ingest documents on an ongoing basis — forever.

AI-driven OCR & HWR

Read native PDFs, scanned images, and handwritten notes. Every output carries a confidence score so you know what's reliable and what needs a human eye.

Smart classification

Documents are auto-classified against your taxonomy (62 types × 17 disciplines pre-seeded). Customise types, attributes and extraction hints with no code.

Structured extraction

Every classified document yields a structured record: reference number, dates, asset IDs, certificate numbers and any custom field you define.

Reviewer workflow

Anything below confidence threshold drops into a dedicated review queue. Approve, correct, re-classify or reject — with SLA tracking and full audit log.

Live REST connector

Push approved records straight into your DMS via REST API. OAuth 2.0, API key or bearer token. Field mapping is configured in the UI — no integration code required on your side.

Enterprise-grade security

AES-256 at rest, TLS 1.3 in transit, role-based permissions, full audit log, GDPR-compliant retention, and admin-controlled decommission mode.

The processing pipeline

Seven automated stages, fully observable, all asynchronous.

Every document moves through the same pipeline. The system handles batches of thousands without blocking the UI — and a live dashboard shows queue depth, throughput and per-stage status in real time.

1

Pre-processing

File detection, image normalisation, page splitting

2

OCR

Full optical character recognition across all scans

3

HWR

Handwriting detection & recognition with confidence score

4

Classification

AI maps to your taxonomy using example docs as context

5

Extraction

Structured fields pulled per document type

6

Scoring

Confidence and completeness metrics calculated

7

Routing

Processed / Needs Review / Failed status assignment

Built-in taxonomy

62 document types × 17 engineering disciplines, pre-loaded on day one.

Out of the box, DocRocket understands the language of engineering operations — from Test & Commissioning reports to Hydrographic surveys to Statutory Inspections. Every type is fully editable.

Disciplines

17 total
AArchitecture
KBuilding Management (BMS) & Control Systems
CCivil Engineering
EElectrical Engineering — LV Systems
BEnvironment, Health & Safety
YFire Fighting & Alarm Systems
GGIS & Land Surveys
VHeating, Ventilation & Cooling (HVAC)
HHydrographic
DMarine Engineering
MMechanical Engineering
SStructural Engineering

Document types

62 total
CERCertificationSafety acceptance · ISO · Declaration of Conformity
INSInspectionStatutory · Principal · Routine · Safety inspections
TACTest & CommissioningFAT · SAT · Load test · Commissioning reports
RAMRisk Assessment & Method StatementRAMS, COSHH, HAZOP
OMMOperations & Maintenance ManualO&M documentation
GEAGeneral Arrangement DrawingGA Plan, Location, Layout
SCESchematic DrawingElectrical / Mechanical schematics
PERWork PermitExcavation · Lifting · Hot works · Confined spaces
REGRegisterRisk register · Drawing register
CALCalculationLighting · Structural calculations
Reference numbering

Every document gets a permanent ID — automatically.

The reference number is generated at classification time and becomes the document's identity throughout the system, in all exports, and in every push to your DMS. The pattern is human-readable, sequential per discipline × type, and zero-padded.

S-CAL-0042 Structural Engineering · Calculation · document #42
E-SCE-0017 Electrical Engineering · Schematic · document #17
B-ASS-0103 Environment H&S · Assessment · document #103
Setup space

You stay in control of the taxonomy — no developer required.

Every document type, attribute, extraction hint and AI behaviour is configured through a clean admin UI. Adding a new type or tweaking a field is a 60-second job.

  • Define document types with 3-character codes, descriptions and example content
  • Add dynamic attributes per type — text, date, number, boolean, dropdown
  • Provide extraction hints in plain English (e.g. "Usually found in top-right header")
  • Upload 1–5 example documents per type to teach the AI
  • Toggle auto-suggest behaviour for unclassified documents
docrocket.app / setup / types / new
Document type
Test & Commissioning
Code
TAC
Auto-suggest
● ENABLED
Attributes (4)
Test datedate · req
Equipment IDtext · req
Engineertext · opt
Pass / Failbool · req
3 example docs uploaded · AI context active
Document table

Every classified document, side-by-side with the original.

The master document table is filterable, sortable and inline-editable. Click any row to open a side-by-side detail view — original on the left, structured fields on the right. Make corrections, add notes, re-classify, or push to your DMS — all in one place.

  • System-generated reference numbers (e.g. C-GEA-0001)
  • AI-extracted summaries (max 20 words, title + purpose + source)
  • Asset ID linking — one document can reference 60+ assets
  • Classification confidence & completeness scoring per row
  • Status badges: Processed · Needs Review · Failed
docrocket.app / documents
Reference
Document
Type
Confidence
Status
C-GEA-0042
Berth 4 GA PlanKing's Yard · Phase 2
GEA
98%
D-TAC-0017
FAT Report — Capstan 4Marine Engineering
TAC
94%
B-ASS-0103
COSHH Assessment (handwritten)HWR confidence below threshold
ASS
62%
S-CAL-0089
Crane Foundation CalcsStructural Engineering
CAL
91%
Y-CER-0231
Fire Alarm Compliance CertIssued by Acme Fire Ltd
CER
96%
E-SCE-0058
LV Distribution SchematicMissing required field: panel ID
SCE
88%
Role-based access

Four roles. Every action logged. Permissions enforced everywhere.

DocRocket is designed for multi-user enterprise teams. Every view, edit, export and API push is recorded with user account, timestamp, action type and document reference.

Admin
Full system access — manage users & roles, configure document types, edit AI prompt, control decommission mode.
Document Manager
Upload documents, manage batches, configure document types, export data and run REST pushes.
Reviewer
Work the review queue — approve, correct, re-classify or mark as failed. Add reviewer notes.
Read-Only
View processed documents and reports only. No upload, no edit, no export.
AI configuration

The classification prompt is yours to tune — versioned and rollback-safe.

Unlike a black-box AI tool, DocRocket exposes the actual classification prompt in the admin UI. Edit it to match your terminology, tune for your industry, or roll back to any previous version with one click.

  • Edit the live AI prompt in plain English
  • Every change is versioned with timestamp and user
  • One-click rollback to any prior version
  • Adjust confidence threshold per deployment (default 75%)
  • Summary style is configurable — 20-word default, no filler language
docrocket.app / settings / ai-config
// classification.prompt · v4 · edited 12 May 2026
You are an AI assistant specialised in
analysing, classifying, and summarising
technical and engineering documents.

Your task is to extract and return
structured information in a valid JSON array.

// fields per document:
- document_name
- summary // max 20 words
- document_type_code // from {doc_types}
- discipline_code // from {disciplines}
- first_identified_date

● Livev4 of 7Rollback ↺
Output

Excel, CSV, or live REST push to your DMS.

DocRocket isn't a silo — it's a feeder. Every record can flow out as a structured export or stream live into your downstream systems.

Excel / CSV export

Full table or filtered subset. Column selection, separate sheets per discipline, summary tabs and saved presets. File path column included for DMS import mapping.

XLSX CSV Saved presets

REST API connector

Map extracted attributes to your DMS schema directly in the UI. Push on approval, scheduled nightly, or manual bulk. Every push logged with success/failure and retry capability.

OAuth 2.0 API Key Bearer Token
Security & compliance

Enterprise-grade by default. Not bolted on.

Every byte encrypted, every action audited, every retention policy configurable. DocRocket is built to pass security review, not work around it.

AES-256 at rest

All processed documents and metadata encrypted on storage. Keys rotated on schedule.

TLS 1.3 in transit

Every connection — UI, API, export — protected end-to-end with modern TLS.

Full audit log

Every view, edit, export and API push recorded with user, timestamp and document reference.

Configurable retention

Retention policies set per document type to match GDPR, sector regulation or internal policy.

RBAC throughout

Four roles enforce who can upload, review, export, configure or decommission. No bypass paths.

Decommission mode

One-click admin action purges all data and removes infrastructure config when the project ends.

Delivery plan

From kickoff to live portal in five phases.

A predictable, phase-gated programme designed to land working software in your hands as fast as possible — and to be confident, observable and safe at every step.

Phase 1Week 1

Discovery & configuration workshop

We work directly with your team to confirm document types, attributes, asset ID structures, DMS target schema and integration end-points. Output: configured taxonomy + signed-off field mappings.

Workshop Taxonomy lock DMS schema
Phase 2Weeks 2–4

Build & configure portal

Core platform deployed. Document types and attributes seeded. AI classification prompt tuned to your terminology. Demo Mode wired up with synthetic samples so the portal looks fully operational from day one.

Setup space AI prompt tune Demo Mode
Phase 3Weeks 5–6

Pilot batch & review tuning

Process a representative batch of your real documents. Tune confidence thresholds, extraction hints and example documents based on observed results. Reviewer workflow tested end-to-end.

Real-document pilot Threshold tuning Reviewer training
Phase 4Weeks 7–8

DMS integration & production launch

REST connector configured against your live DMS. Field mappings, authentication and push schedules validated. Go-live for the full document migration with monitoring and on-call support.

REST connector Go-live Monitoring
Phase 5Ongoing

Operate, support & iterate

DocRocket becomes a standing operational tool. New document types added as your needs evolve. Quarterly reviews of confidence trends, throughput and accuracy. Decommission mode available on project completion.

SLA support Quarterly review Decommission ready
Engagement & investment

Productised platform. Configured for you.

DocRocket is delivered as a productised platform with a fixed setup engagement and an ongoing operational fee. You get a configured portal, not a custom build — which means faster delivery and a more predictable cost.

  • Fixed-price setup covering discovery, configuration, pilot tuning and DMS integration.
  • Monthly operational fee covering hosting, AI processing, support and ongoing tuning.
  • Per-volume pricing on document throughput above the included monthly allowance.
  • No lock-in — decommission mode purges all data and config on request.
  • Tailored quote issued after the discovery call — based on your real volume, DMS, and SLA requirements.

Ready to put DocRocket in front of your documents?

Book a 30-minute discovery call. We'll walk through your taxonomy, your DMS target, and your real document mix — and come back with a tailored engagement plan within five working days.

Book a discovery call Back to top