How each agent works in practice — drawn from ecovium's operational reality and from deployments across PE-backed multi-site companies. Some of these you asked about. Others are scenarios we've seen create value that wasn't on anyone's roadmap until the platform surfaced it.
Month-end cutoff. 10 properties running Dynamics 365 with different chart-of-accounts mappings, billing cycles, and local tax treatments. Property 3 closes on the 28th. Property 7 closes on the 1st.
Collectors pull final period data from each property's Dynamics instance. Financial Consolidation normalizes all 10 into a unified chart of accounts, eliminates intercompany transactions, applies currency conversion, and generates a consolidated P&L and balance sheet.
Consolidated financial package ready for review 3 days ahead of the previous manual timeline. Discrepancies flagged with source property and line item identified. Finance leadership reviews instead of assembles.
What changed: At a 12-property industrial services company we deployed for, month-end close went from 14 days of manual reconciliation to 3 days of review. The CFO said the first month felt wrong because it was done too fast.
Customer signs a Warehouse Management contract at Property 4 in Germany. The Mantis WMS implementation team sits at Property 2. Invoice generated at Property 4 but professional services hours logged at Property 2. Revenue recognition split unclear.
Billing agent detects cross-property service delivery via Collector data from both properties. Flags the split. Applies the intercompany billing rules defined during scoping. Generates the customer-facing invoice from Property 4 and the intercompany transfer record to Property 2 simultaneously.
Customer receives one clean invoice. Property 2 gets the intercompany credit. Finance sees both entries reconciled in the consolidated view. No manual intervention. No delayed recognition.
What changed: Cross-property billing exceptions are one of the most common time sinks in multi-site SaaS companies. We've seen AR teams spend 15–20 hours per month just resolving these manually. The agent catches them at invoice generation, not after the customer complains.
47 overdue invoices detected across 6 properties. Payment terms vary — net 30 at Property 1, net 45 at Property 6, net 60 for enterprise accounts at Property 3. Some customers have balances across multiple properties.
Agent groups overdue amounts by customer (not by property), applies the correct dunning sequence per payment term tier, and generates communications in the customer's language. Customers with balances across multiple properties receive one consolidated reminder — not three separate emails from three properties.
Dunning sequences initiated. Escalation flags set for accounts over 90 days. Finance team sees a single collections dashboard across all properties. Customer experience stays professional — one company, one communication, not a fragmented chase.
What changed: DSO (days sales outstanding) at a PE-backed SaaS company we deployed for dropped by 11 days in the first quarter. The improvement came entirely from consistent, timely dunning — not changes in customer behavior. The system just stopped letting invoices slip through the cracks.
EU Council publishes an updated sanctions list at 08:00 CET. ecovium's Customs & Compliance division serves customers who run sanctions screening through ecovium's software. 18 customer configurations across Properties 1, 4, and 7 are potentially affected.
Compliance Monitor ingests the update, cross-references against all active customer configurations, identifies the 18 affected accounts, generates an impact assessment for each, and triggers Dispatchers at Properties 1, 4, and 7 to notify the relevant customs team leads with full context — which customers, which list entries, which configurations need review.
By 08:45 CET, every affected customer account has been flagged, every relevant team member has been notified, and the impact assessment is documented. Compliance team reviews and updates configurations. No customer runs a stale sanctions list past morning.
Why this matters: In foreign trade compliance, a missed sanctions list update exposes the customer — and ecovium — to regulatory penalties. The speed from publication to notification is the difference between a same-day update and a configuration that runs stale for days because someone didn't see the change.
Predictive Compliance agent detects a proposed regulatory change in the Bundesanzeiger (Federal Gazette) and EU regulatory pipeline that will affect customs tariff classifications effective May 1. The change hasn't been widely reported yet — it's in the consultation phase.
Agent maps the proposed change against ecovium's entire customer configuration database. Identifies 340 customer configurations across all properties that will require tariff code updates. Generates a pre-assessment report with customer-by-customer impact analysis. Dispatches early warning to the compliance team with a 6-week runway.
Compliance team has 6 weeks to prepare customer communications, schedule configuration updates, and stagger the rollout. When the regulation takes effect, ecovium's customers are already compliant. Competitors' customers are scrambling.
What most companies don't have: Reactive compliance monitoring is standard. Predictive compliance — scanning regulatory pipelines and assessing impact before rules take effect — is a capability most companies don't even staff for. This agent does what a senior regulatory analyst would do if they read every government publication in every jurisdiction ecovium serves, every day. Nobody has that person.
External auditor requests a complete record of all sanctions screening activity, tariff classification changes, and compliance configuration updates across all ecovium properties for the past 12 months. Request arrives on a Tuesday. They want it by Thursday.
Compliance Monitor pulls the full audit trail from the hub — every screening event, every configuration change, every regulatory update response, timestamped and attributed to the specific property and customer account. Generates a formatted audit package with chronological compliance activity across all properties.
Audit package delivered within the hour. Every action traceable. Every response time documented. The compliance team didn't pull a single report manually. The auditor has what they need before lunch.
What we've seen: At a PE-backed logistics company, compliance audit prep used to take 2–3 weeks of team time across 4 properties. After deployment, the same audit package generates in under an hour. The compliance lead's reaction: "I didn't know we had all of this data." They did — it was just scattered across 4 systems in 4 properties.
Over a 10-day period, 23 customers across 4 properties file tickets that look unrelated at the property level — different descriptions, different products, different severity. But they all involve the same Mantis WMS module after a recent update.
Support Intelligence aggregates tickets across all properties, runs pattern analysis against product module, version, and error type. Identifies the common root cause. Generates a pattern alert with the 23 affected accounts, the specific WMS module and version, and the recommended fix. Dispatches to the Mantis product team and all affected property support leads simultaneously.
Mantis team pushes a hotfix. All 23 customers are proactively notified with the resolution before most of them escalate. 4 properties that would have investigated the same issue independently get the answer from one analysis.
Why this can't happen manually: When tickets are siloed by property, nobody sees the pattern. Each property's support team investigates independently, burns hours on the same root cause, and the product team hears about it late. The agent sees across all properties simultaneously — that's the structural advantage of centralized intelligence.
A regulatory change affects how a specific commodity is classified for export. The Customs & Compliance division at Property 6 identifies the issue. But the customer's Shipping configuration — which determines how the commodity is declared on shipping documents — lives at Property 2.
Workflow Engine detects the cross-division, cross-property dependency. Creates a linked workflow: Customs team at Property 6 provides the updated classification, Workflow Engine routes it to the Shipping configuration team at Property 2 with full context and the customer account details. Tracks both sides of the resolution.
Customer's Customs and Shipping configurations are updated in sync. No manual phone call between properties. No email thread that gets buried. Resolution time: 2.1 hours. Without the Workflow Engine, this type of cross-property, cross-division coordination typically takes 2–5 days because it depends on someone at Property 6 knowing who to contact at Property 2.
The hidden cost: Cross-property workflow failures don't show up as incidents. They show up as delays. Nobody logs a ticket that says "I couldn't find the right person at another property." The work just takes longer. The Workflow Engine eliminates the search — it knows the org structure, escalation paths, and communication channels at every property.
A German-speaking customer files a WMS support ticket in German through Property 3's ticketing system. The Mantis WMS specialist who needs to handle this issue is based at an English-speaking property.
Multi-Language agent detects the language, translates the ticket context for the English-speaking specialist, routes it with full technical context preserved. When the specialist responds in English, the agent translates the response back to German for the customer-facing communication. All within the existing ticketing workflow — no separate translation step.
Customer receives support in their language. Specialist works in theirs. Resolution quality isn't compromised by language barriers. When Greece goes live in Phase 2, the same agent handles Greek without a new build — just a language model addition.
Why this scales: ecovium serves 30+ countries. As properties expand into new geographies, language routing becomes exponentially complex. The agent handles it structurally — not by hiring multilingual support staff at every property.
Revenue Expansion agent analyzes the full customer base across all properties. Identifies 47 customers who only use ecovium's Warehouse Management product but whose operational data (pulled by Collectors from CRM and ticketing systems) shows high outbound shipping volume — volume that's currently processed through competitor shipping solutions.
Agent scores each account by expansion potential based on shipping volume, contract size, relationship tenure, and support satisfaction. Generates a prioritized cross-sell pipeline with account context already assembled — current products, contract value, key contact, last interaction, and the specific shipping volume that makes them a candidate.
Sales team receives 47 qualified cross-sell opportunities with full context. No cold outreach. These are existing customers with a demonstrated need and an existing relationship. Account managers have the conversation armed with data instead of intuition.
The math: 5,000+ customers × 5 product divisions = tens of thousands of expansion combinations. Even at a conservative 5% conversion rate on identified opportunities, the pipeline this agent generates represents significant annual expansion revenue — revenue that exists in the customer base today but requires systematic identification that no manual process can sustain across 10 properties.
Sales Pipeline agent scans CRM across all properties. Flags 12 deals that have been in "negotiation" stage for 30+ days with no stage change. Rep notes say "waiting for decision" on all 12.
Agent cross-references CRM data with actual engagement signals from Collectors: email activity from Exchange, document access from SharePoint, support ticket activity, and meeting history. 8 deals show zero engagement in 30 days — no emails opened, no documents accessed, no meetings. 4 deals show active engagement despite no CRM stage change — the rep just hasn't updated.
Sales leadership gets a reality-based pipeline: 8 deals flagged as likely dead (with evidence), 4 deals confirmed active (with engagement timeline). Forecast accuracy improves because it's based on behavior, not rep optimism. The 8 dead deals stop inflating the pipeline. The 4 active deals get proper attention.
What we've seen at other PE-backed companies: Pipeline inflation is universal. Reps don't close stale deals because they might come back. Forecast accuracy at one deployment went from 62% to 91% in 90 days — not because reps changed behavior, but because the agent scored deals on engagement instead of stage labels.
Marketing Intelligence agent consolidates campaign performance across all properties. Property 3's WMS campaign is generating leads at €4.20 CPL with a 14% conversion to qualified opportunity. Property 9's identical campaign format is running at €28 CPL with 2% conversion.
Agent generates a reallocation recommendation: shift €12K of Property 9's Q2 budget to Property 3 and replicate Property 3's targeting and messaging at Property 9. Includes the full ROI comparison, lead quality metrics, and projected pipeline impact of the reallocation.
Marketing leadership sees one consolidated view of what's working where — not 10 separate property reports. Reallocation decision is data-driven and property-specific. Projected impact: €12K redirected from 2% conversion to 14% conversion generates an estimated 40+ additional qualified leads.
Why centralized marketing intelligence matters: When each property runs marketing independently, nobody sees the cross-property comparison. The best campaign and the worst campaign can run simultaneously for months without anyone knowing. The agent makes the comparison automatic and the recommendation actionable.
R&D Intelligence agent pulls Azure DevOps metrics from Collectors across product divisions. Transport division's sprint velocity is stable, but bug backlog has grown 34% over 8 weeks. Code review turnaround time has increased from 4 hours to 18 hours. The team is shipping features but accumulating debt that isn't visible in sprint reports.
Agent correlates velocity, backlog growth, review latency, and deployment frequency. Generates a technical health assessment showing the Transport division is 4–6 weeks from a point where accumulated debt will force a sprint-zero cleanup that delays the next major release.
Product leadership gets the early warning with enough time to allocate a partial sprint to debt reduction — preventing the full sprint-zero scenario. The data was all in Azure DevOps. The agent connected the dots that no single sprint report shows.
What we've seen: At a PE-backed SaaS company, a technical debt situation exactly like this caused a 6-week release delay that impacted two customer commitments. The CTO said afterward: "The data was there — we just weren't looking at it the right way." The agent looks at it continuously.
Property 5's Collector initiates a scheduled financial data pull at 14:00. Dynamics 365 API returns a timeout — the instance is under heavy load from an end-of-day batch process running simultaneously.
Self-Healing agent detects the failed pull within seconds. Checks the failure type (timeout, not auth error). Implements exponential backoff — retries at 14:02, then 14:06 if needed. Second retry succeeds. Logs the exception with timestamp, failure type, retry count, and resolution. Updates the Collector's schedule awareness so future pulls at this property avoid the known batch window.
Data arrives 2 minutes late. No human intervention. No alert fatigue. The exception is logged and the schedule adjusted. If the same failure occurs 3 times in a row, then it escalates. Single transient failures are handled silently.
Why this matters: 34 agents running continuously across 10 properties will generate occasional transient failures — API timeouts, rate limits, temporary network issues. Without Self-Healing, each one of those becomes a support ticket for someone on your team. The platform watches itself so you don't have to staff an AI operations team to manage the AI operations platform.
Monday 08:00. Leadership opens the Roll-Up Dashboard. No reports to request. No emails to chase. No spreadsheets to assemble. The data is already there because it flows continuously.
R&D velocity across all product divisions. Active implementation status. Marketing campaign ROI by property. Sales pipeline health with real engagement scoring. Support ticket trends and SLA compliance. Compliance status across all jurisdictions. Financial consolidation status. All updated as of the last Collector pull — not last Friday's export.
Identify the 2–3 items that need attention. Act on them. Skip the 45 minutes of "let me pull up the numbers" that used to start every Monday. The dashboard doesn't replace leadership judgment — it eliminates the data assembly that delays it.
What a PE operating director told us: "I used to spend the first two hours of every Monday getting a picture of where things stood across the portfolio companies I manage. Now I walk in and the picture is already on the wall. Those two hours go to the companies that need my attention — not to finding out which ones need my attention."
Scheduled Collector run at Property 8: tickets pull every 30 minutes, financials every 4 hours, compliance daily. The Collector connects to Property 8's specific Dynamics instance, Exchange environment, ticketing system, and CRM using encrypted, scoped credentials.
Collector pulls only the data objects defined during scoping — nothing more. Normalizes ticket data into the hub's standard format regardless of Property 8's local ticketing configuration. Financial records mapped to the unified chart of accounts. Compliance items tagged with jurisdiction and regulatory source. All pushed to the hub. Nothing stored locally.
Hub now has Property 8's latest operational data in a format identical to every other property. Support Intelligence can analyze tickets across all 10. Financial Consolidation includes Property 8's numbers. Compliance Monitor covers Property 8's jurisdiction. The hub doesn't care about Property 8's local system quirks — the Collector abstracts them away.
The design principle: Each Collector is configured specifically for its property's systems. Property 8's Dynamics instance is different from Property 3's. The Collector handles the translation. The hub handles the intelligence. That separation is what makes the architecture scalable — adding a new property means configuring a new Collector, not rebuilding the hub.
Financial Consolidation agent detects a €4,200 billing discrepancy at Property 4 — a customer's invoice doesn't match the service delivery records pulled from Property 4's CRM and ticketing data.
Dispatcher at Property 4 receives the alert from the hub. It knows Property 4's org structure — who handles billing exceptions, which communication channel they use (Teams, email, or ticketing), and what escalation path to follow. Delivers the alert to the AR manager with full context: customer name, invoice number, discrepancy amount, source data from CRM and ticketing, and recommended resolution.
AR manager at Property 4 has the full picture within 3 minutes of detection. No generic alert. No "please investigate." The Dispatcher delivers the intelligence, the context, and the recommended action — all routed to the right person through the right channel.
Why Dispatchers are per-property: Every property has a different org structure, different team responsibilities, and different communication preferences. A centralized notification system sends generic alerts to a distribution list. A Dispatcher knows exactly who handles what at that specific property and delivers through the channel they actually check. That's the difference between an alert that gets acted on and an alert that gets buried.
Once the platform is operational, the Enhancement Report after 30–60 days will surface use cases we can't predict today — patterns that only emerge when real data flows through real agents across real properties.
Back to Proposal →