Dorcas

The Problem

Every organization runs on approvals. Purchase orders need sign-off. Expense reports need review. Contracts need legal clearance. Insurance claims need adjudication. Hiring decisions need headcount approval.

Yet the tools built for this are fundamentally broken:

Rigid and siloed. SAP Ariba handles procurement. Concur handles expenses. DocuSign CLM handles contracts. Guidewire handles insurance claims. Each tool is a walled garden — separate vendor, separate implementation, separate data, separate team. Adding a new approval type means buying another tool.

Not intelligent. Legacy approval tools are glorified routing engines. They send requests to pre-defined approvers based on static rules. They don't understand the content of what they're routing. A $50 office supply purchase follows the same approval chain as a $500,000 capital expenditure.

Expensive and slow to implement. Enterprise approval tools like SAP Ariba or Coupa take 12-18 months to implement and cost hundreds of thousands in licensing and consulting. Small and mid-sized businesses are priced out entirely.

The cost is staggering. Inefficient approval processes consume 26% of employee productive time and cost organizations up to 30% of annual revenue. 62% of procurement leaders cite approval delays as their number one operational bottleneck.

---

What Dorcas Does

Dorcas is an AI-native platform for creating any type of approval workflow. Instead of buying separate tools for procurement, expenses, contracts, HR, and compliance, organizations use one platform that handles all of them.

The key architectural insight: 29 composable primitives replace 300+ rigid tools.

Where traditional platforms build a separate tool for every use case, Dorcas provides a set of powerful, reusable building blocks that AI assembles into domain-specific solutions. Adding a new approval type — say, marketing creative reviews or construction change orders — is configuration, not engineering.

---

Core Capabilities

Intelligent Auto-Approval

Dorcas doesn't just route approvals — it makes them. The AI decision engine analyzes each request against policies, historical patterns, and risk signals to auto-approve routine items with full confidence scoring and reasoning.

• A $200 expense report with a valid receipt and correct GL code? Auto-approved in seconds with a 0.95 confidence score.
• A $50,000 purchase order from a new vendor with no prior history? Escalated to procurement with risk flags and recommended review points.

In practice, 60-75% of procurement approvals and 50-75% of healthcare prior authorizations can be fully automated. The remaining items get routed to the right person with AI-generated context, reducing human review time by 80%.

Universal Document Intelligence

One extraction engine handles every document type that flows through approval workflows:

• Receipts: Merchant, amount, date, category, tax, tip
• Invoices: Line items, quantities, unit prices, payment terms, GL codes
• Contracts: Clauses, renewal dates, termination conditions, risk factors
• Medical records: Diagnosis codes, procedure codes, clinical notes
• Identity documents: Names, addresses, ID numbers, biometric data

No need for 6 separate parsing tools. The same

ai.extract

primitive handles them all — learning from corrections and improving over time.

Policy-as-Configuration

Compliance rules, approval matrices, delegation policies, and escalation thresholds are defined as YAML configuration — not hardcoded logic. This means:

• New regulation? Add a policy file. No code changes.
• New approval matrix? Update the configuration. Deploy in minutes.
• New compliance requirement? SOX, HIPAA, GDPR, FTC — each is a config file that can be added, updated, or versioned independently.

Compliance isn't bolted on after the fact. Every action in Dorcas writes an immutable audit entry by default. The audit trail is woven into the architecture, not layered on top.

Predictive Routing

Instead of routing approvals based on rigid org charts, Dorcas routes based on:

• Workload: Who has capacity right now?
• Expertise: Who has approved similar requests before?
• Availability: Who's on PTO or out of office?
• Urgency: What's the SLA, and who can meet it?

This eliminates the common problem of approvals sitting in someone's inbox for days because they're overloaded or traveling.

Natural Language Interface

Users can interact with Dorcas conversationally:

• "Text me a receipt photo" — Dorcas receives it, extracts the data, creates an expense entry, and submits for approval.
• "Show me all rejected POs over $10K this quarter" — Instant query across approval history.
• "What's the status of the Acme contract?" — Real-time status with predicted completion date.

---

Use Cases

Procurement (Purchase Orders)

The full procure-to-pay cycle automated end-to-end:

Requisition — Employee submits a purchase request via form, chat, or email. AI pre-fills vendor details, validates budget availability, and assigns GL codes automatically.

Approval routing — Dorcas evaluates the request against approval matrices: amount thresholds, department budgets, vendor risk profiles, historical patterns. Routine orders are auto-approved. Complex orders are routed to the right approver with AI-generated context.

Three-way matching — When the invoice arrives, Dorcas matches it against the purchase order and goods receipt automatically. Price discrepancies, quantity mismatches, and duplicate invoices are flagged instantly.

Payment — Approved invoices trigger payment processing across ACH, wire, or card rails. Budget commitments are released and spend analytics updated in real-time.

Impact: Reduce procurement cycle time from 2 weeks to 2 days. Auto-approve 60-70% of routine purchases. Eliminate manual three-way matching errors.

Finance (Expense Management)

From receipt capture to reimbursement in minutes, not weeks:

Capture — Employee photographs a receipt or forwards an email. AI extracts merchant, amount, date, category, tax, and tip automatically.

Pre-submission compliance — Before the employee even submits, Dorcas checks against expense policy: Is this category allowed? Is the amount within limits? Is the receipt valid? Are there duplicates? Issues are flagged before submission, not after review.

Intelligent approval — The AI decision engine evaluates each expense against policy compliance, historical patterns, and fraud signals. In-policy expenses under configurable thresholds are auto-approved with full audit trail.

Reimbursement — Approved expenses trigger automatic reimbursement via integrated payment rails. Corporate card transactions are reconciled automatically.

Impact: Reduce per-expense processing time from 20+ minutes to under 3 minutes. Catch policy violations before submission, not after. Eliminate receipt chasing.

Legal (Contract Approvals)

From contract request to signature with AI-powered risk intelligence:

Intake — Business teams submit contract requests with requirements. Dorcas selects the appropriate template and pre-fills terms based on deal parameters.

AI clause analysis — Every contract is analyzed for risk: non-standard clauses, unfavorable terms, missing protections, liability exposure. Each clause gets a risk score with explanation.

Smart routing — High-risk contracts route to senior legal. Standard contracts with low risk scores route to junior reviewers or auto-approve. Finance reviews payment terms. Business reviews scope.

Negotiation — AI generates counterproposal language for flagged clauses. Redlining and version tracking happen in real-time.

Execution — Final approval triggers e-signature workflow. Signed contracts are stored with extracted obligations, renewal dates, and automated alerts.

Impact: Reduce contract cycle time from 3.4 weeks to under 1 week. Catch 95% of risky clauses automatically. Eliminate contract bottlenecks that delay revenue.

Healthcare (Prior Authorization)

Automate the PA process that delays patient care:

Clinical data pull — Dorcas integrates with EHR systems (Epic, Cerner via FHIR) to pull patient records, diagnosis codes, and procedure details automatically.

Payer rules matching — Each payer (Aetna, Cigna, UnitedHealthcare) has different rules. Dorcas maintains payer-specific policy configurations and checks medical necessity against the right rule set.

Auto-population and submission — PA forms are generated with clinical evidence auto-populated. Submissions go to the payer portal automatically.

Denial management — When denials come back, Dorcas analyzes the denial pattern, identifies the best appeal strategy (e.g., "73% appeal success when PT records are included"), pulls supporting documents, generates the appeal letter, and submits.

Impact: Reduce per-PA processing time from 20-30 minutes to 2-3 minutes. Automate 50-75% of PA tasks fully. Improve appeal success rates with data-driven strategies. Reduce care delays that 94% of physicians cite as harmful.

Insurance (Claims Processing)

From first notice of loss to settlement:

FNOL intake — Claims arrive via phone, web, email, chat, or fax. Dorcas normalizes all channels into a structured claim record.

AI triage — Each claim is automatically classified by complexity, fraud risk, and expected handling requirements. Simple, low-risk claims are fast-tracked for auto-adjudication.

Document intelligence — Police reports, medical records, repair estimates, and photos are processed automatically. Key facts are extracted and cross-referenced against the claim.

Fraud detection — ML models analyze claim patterns for anomalies: suspicious timing, duplicate claims, inconsistent narratives, known fraud indicators.

Adjudication — Straightforward claims are auto-adjudicated against policy terms. Complex claims are routed to adjusters with AI-generated summaries and recommended settlements.

Impact: Reduce average claims processing time by 60%. Auto-adjudicate 40-50% of straightforward claims. Catch fraud earlier with pattern-based detection.

HR (People Operations)

Hiring, onboarding, leave, and compensation — all approval workflows unified:

Headcount approval — Requisitions include AI-generated market data: salary benchmarks, time-to-fill predictions, internal mobility candidates.

Offer management — Offer letters are generated from approved parameters with compensation details, equity, and benefits auto-populated. E-signature and acceptance tracking built in.

Onboarding orchestration — Employee acceptance triggers automatic IT provisioning, benefits enrollment, training assignment, and equipment ordering. Each step is an approval that flows through the system.

Leave management — Leave requests are checked against team coverage, project deadlines, and policy limits. Straightforward requests auto-approve. Conflicts trigger manager review with context.

Impact: Reduce time-to-hire by 40%. Eliminate onboarding delays caused by manual IT and benefits provisioning. Standardize compensation decisions across the organization.

---

Architecture: 29 Composable Primitives

The power of Dorcas is in its architecture. Instead of building rigid, niche-specific tools, Dorcas provides 29 primitives organized across 8 layers:

Workflow Engine (6 primitives) — Define workflows, act on requests, evaluate rules, build forms, schedule actions, query data.

AI Brain (5 primitives) — Decide (approve/reject/escalate with reasoning), Extract (universal document intelligence), Analyze (risk scoring, anomaly detection), Generate (documents, messages, reports), Converse (natural language interaction).

Data and Documents (4 primitives) — Document management with versioning, e-signature integration, data transformation and validation, universal data queries.

People and Access (3 primitives) — Org chart resolution, authentication (SSO/MFA), system provisioning.

Money (3 primitives) — Budget transactions, payment execution, card feed ingestion.

Communication (3 primitives) — Multi-channel notifications, external portals, omni-channel intake.

Integration (3 primitives) — Read from, write to, and subscribe to events from any external system. Adapter registry supports 40+ systems via YAML configuration.

Compliance (2 primitives) — Policy checking and immutable audit logging with configurable policy stores for SOX, HIPAA, GDPR, and more.

Why this matters: Adding a new industry vertical — say, construction change orders — means adding adapter configurations for Procore, policy files for AIA standards, and workflow templates for change order approval chains. Zero new code. Days, not months.

---

Why Dorcas

One platform, every approval type. Stop buying separate tools for procurement, expenses, contracts, HR, legal, and compliance. Dorcas handles all of them with the same 29 primitives.

AI that decides, not just routes. Auto-approve 50-75% of routine approvals with full confidence scoring, reasoning, and audit trail. Humans focus on the decisions that actually need judgment.

Configuration, not code. New workflows, policies, integrations, and compliance rules are YAML configuration files. Deploy changes in minutes, not months.

Enterprise-ready from day one. Multi-tenant architecture, SSO/MFA, immutable audit logging, HIPAA-capable compliance infrastructure, and role-based access control — not bolted on later.

Implementation in weeks, not years. Where SAP Ariba takes 12-18 months, Dorcas deploys in weeks. Where Coupa costs hundreds of thousands, Dorcas prices for the mid-market.

---

Built by Skunkworks

Dorcas was built because we saw the same problem everywhere: organizations spending millions on rigid approval tools that don't talk to each other, don't understand the content they're routing, and take years to implement.

We asked a simple question: what if one intelligent platform could handle every approval workflow in every industry? Not by building 300 tools, but by building 29 primitives that AI assembles into solutions?

That's Dorcas. One platform. Any approval. Any industry.