Agentic AI across the WYRM platform
Seven modules — led by the flagship engineering products WYRM MEP and WYRM Data, with Procure, Ledger, Cyber, and Legal add-ons and a standalone Healthcare line — running on one shared agentic engine. Specialist agents per module run parallel research against live data sources, an ensemble layer fuses their verdicts with explicit confidence weights, and every step is logged for audit. The flagship products and add-ons ship to SMEs; WYRM Healthcare adds a clinical-safety layer for UK NHS providers.
Definition
What is agentic operating intelligence?
A system that performs a full analysis autonomously across a defined domain, against a documented cadence, with human-reviewable output.
Most AI products in the market today are reactive. A user types a prompt, the tool returns a generic answer. Chatbots wait for input. Copilots suggest completions. They are assistants that respond when prompted.
Agentic operating intelligence is different. Specialist agents operate within defined polling and scoring schedules, produce outputs on a documented cadence, and log every action. The sanctions agent re-checks designations on every revision. The tender-match agent re-scores notices on every publication tick. The CVE agent re-prioritises against the operator footprint on every new exploitability signal. The clause agent re-runs the playbook comparison every time the operator updates it.
The distinction matters because operating decisions — across procurement, finance, security, and legal — require continuous awareness of changing inputs. A question-answering interface is insufficient: the data changes between questions, and the audit trail depends on recording every intermediate state.
Comparison
Chatbot vs copilot vs agent
| Attribute | Chatbot | Copilot | Agent |
|---|---|---|---|
| Initiation | User prompt | User action | Scheduled + event-driven |
| Persistence | Session-based | Session-based | Continuous, audit-logged |
| Planning | None | Suggestion only | Goal-directed, bounded |
| Multi-step execution | No | Limited | Yes, with checkpoints |
| Cross-domain reasoning | No | No | Via orchestrator ensemble |
| Proactive updates | No | No | On material change only |
Architecture
The specialist agents, by module
Two bounded agents per module plus a cross-module orchestrator. Each agent is colour-tagged with the module it belongs to; the orchestrator routes events between them.
Sanctions + Supplier Agent
Screens every counterparty and declared parent ownership against UK OFSI, US OFAC, EU, and UN consolidated lists via OpenSanctions. Traces ownership via OpenCorporates across 200M+ entities to detect common ownership across declared-independent suppliers.
Carbon + Commodity Agent
Computes embedded lifecycle emissions using Climatiq factors and country grid-carbon intensity. Returns CBAM liability at the decision point. Prices HS-coded commodities against LME and Pink Sheet with regime-aware volatility bands.
Tender Match Agent
Polls Find a Tender, Contracts Finder, SAM.gov, and BDUK feeds. Matches notices to operator capability profile and historical win patterns, returning a ranked shortlist with a fit score per opportunity rather than a raw firehose of notices.
Bid Drafting Agent
Drafts response sections against the operator's evidence library and prior winning bids, surfacing where claims are supported and where they require new evidence. Tender Writer mode produces a structured response aligned with the awarding authority's evaluation criteria.
Threat Intel + CVE Agent
Continuous triage across NVD CVE, EPSS exploitability, and CISA KEV. Correlates against the operator's declared technology footprint and surfaces the small subset that actually requires action this week. Monthly sweeps on Pro, weekly on Enterprise.
OSINT + Recon Agent
Username, phone, email, and image lookup across breach indices and OSINT sources. Attack-surface mapping and recon against operator-owned assets. Dark-web watch for credentials and brand mentions, with results surfaced as defensible evidence rather than raw scrape.
Clause Review Agent
Clause-by-clause green / amber / red verdicts on MSAs, NDAs, frameworks, DPAs, and SOWs against the operator's own playbook. Highlights deviations from negotiated baselines with diff-style evidence and a suggested redline.
Regulator Watch Agent
Continuous monitoring of regulator publication APIs (FCA, ICO, CMA, HMRC) and case-law indices. Filters by operator domain so the weekly digest is material updates the operator's legal function should actually act on, not generic regulatory noise.
Entity Resolution Agent
Unified entity graph across every WYRM feed — suppliers, counterparties, regulator-named entities, breach actors, contract parties. Same legal entity surfaces under one canonical record regardless of which module first ingested it, with provenance per claim.
Cross-Module Orchestrator
Routes events between modules so a signal in one surfaces in the others. A sanctions hit in Procure triggers a Cyber supplier-attestation refresh and a Legal contract-review flag. A regulator update in Legal triggers a Procure compliance recheck. Shared audit log throughout.
In practice
Cross-module correlation
A regulator publication triggers four agents across three modules in parallel.
Legal · Regulator Watch Agent
ICO publishes updated guidance on processor sub-contracting. Watch agent flags it within minutes of publication and tags every active DPA in the operator's contract library as needing re-review.
Procure · Supplier Agent
Cross-references the affected DPAs against the active supplier network. Surfaces 14 suppliers whose data-processing footprint falls inside the new guidance — including two whose self-attestation is now stale.
Cyber · Threat Intel Agent
Re-checks the two stale-attestation suppliers against breach indices and CVE exposure. One has a known unpatched exposure in the affected processor product line. Risk score raised.
Data · Cross-Module Orchestrator
Issues a composite update to the operator: 'ICO guidance change invalidates 14 DPAs in scope; 1 supplier has compounding cyber exposure; recommended action is renegotiate DPA + request remediation evidence. Full evidence trail at /decisions/{id}.'
Each agent saw one piece. The orchestrator produced the defensible answer. Every step is logged in the shared audit trail across all three modules.
Impact
Why this matters
OpenSanctions consolidates 40+ global lists updated daily. UN COMTRADE exposes over three billion bilateral trade records. Find a Tender publishes dozens of notices per day. NVD adds hundreds of CVEs per week. The ICO, FCA, CMA, and HMRC publish regulator updates on overlapping cadences. No human analyst can maintain live awareness across all of this — let alone across procurement, finance, security, and legal at once.
The viable approach is specialist agents per domain that do the research and initial analysis, with operators applying judgement to the flagged material changes. Not replacing the operator — augmenting them. An agent re-checks every revision, tick, and registry change. The operator decides what to do about the ones that matter.
For the per-module deep-dive — agent details, comparison against domain-specific incumbent tooling, and defensibility under the relevant regime — see Procure, Ledger, Cyber, Legal, and Data. For terminology, see the platform glossary. For specific applications, see the use cases.