What a forward-deployed engineering engagement looks like inside a $1.1B mid-market PE-backed manufacturer. Synthetic client (Apex Precision Holdings, industrial equipment platform formed via 4 acquisitions), real methodology, real artifacts. Eight agents shipped to production across 9 entities and 4 countries. Heterogeneous ERP environment — D365 F&O, Sage Intacct, NetSuite — bridged without rip-and-replace.
What we walked into. Three weeks of discovery, ten systems mapped, the heatmap of where time was actually leaking.
Hidden FX rules, post-acquisition crosswalks, regional VAT quirks. How undocumented finance conventions became agent-runnable rules.
What the AP Specialist, Controller, and CFO each saw on Monday morning after the agents went live.
How a single intercompany transaction flows through five sub-agents in parallel, and what the feedback loop looks like.
Before/after metrics, accuracy curve, full audit trail, and what the team kept after we left.
Apex Precision Holdings is a $1.1B PE-backed industrial equipment platform formed via four acquisitions over five years. Each acquisition came with its own ERP — D365 F&O for the original parent, Sage Intacct for the German and Mexican subs, NetSuite for the carved-out US-02 entity. The 240-person finance team had been holding the consolidation together with Excel and heroics. PE sponsor wanted close cycle from 11 days down to 6. CFO wanted SOX audit findings to zero. Controller wanted the team out of the data-archaeology business and into actual financial analysis. Three weeks in, we had the picture below.
Heatmap from time-and-motion shadowing of 18 finance professionals across two weeks (AP, AR, IC, Treasury, FP&A, Controllership). Reconciliation work consumed nearly a third of the org’s capacity — across multiple ERP boundaries.
Week-three deliverable: which agent to build first and why. P1 + P2 are the foundation — three-way match clean-up and intercompany reconciliation are how the close cycle compresses. Everything else flows from those.
The close cycle wasn’t broken because the team was slow. It was broken because the same intercompany transaction had to be reconciled three times — once in D365, once in Sage Intacct, once in NetSuite — and the FX rate inconsistencies between the three systems generated daily exceptions that someone manually triaged. Fix the FX policy alignment + intercompany matching, and the close compresses by itself. We didn’t need to make people faster; we needed to remove the work that shouldn’t exist.
Three ERPs, four countries, four acquisitions. Half of how Apex’s finance team actually closes the books was in nobody’s SOP — it lived in the heads of the senior accountants who had been there through every acquisition integration. Before agents could run, we needed those rules captured, scored, and made revisable.
Apex had a 47-page close manual maintained by the Corporate Controllership. The agent extracted 12 executable rules from it. Six of the most-cited shown below.
What lived in people’s heads. Captured through structured interviews with senior AP staff, the IC Accountant, the Treasury Director, the Controllership team, and the FP&A bench. Each rule has a named source so it can be revisited as the org changes.
Q4 intercompany settlements always require manual adjustment for FX timing differences because Oracle books at transaction date and NetSuite books at settlement date.
The German entity (DE-01) invoices get coded differently than policy says because of a legacy VAT structure that was never updated in the SOP after the 2023 acquisition.
Only Maria on Treasury handles the JPY bank account reconciliation because she understands the BOJ rate lag and the correspondent bank timing.
Ramp categories 'Software' and 'SaaS' both map to the same GL code, but 'Software—Implementation' maps to CapEx. The Ramp admin doesn't know this distinction exists.
When the Brazilian entity (when MX-02 had a Brazilian subsidiary) had a month with >$500k in intercompany charges, the transfer pricing team needed a heads-up 3 days before close or it delayed consolidation by 2 days.
The MX-02 acquisition closing GL codes haven’t been migrated to the corporate crosswalk yet — controllers maintain a side spreadsheet for consolidation.
Every agent ships with an explicit confidence threshold. Below it, the agent escalates to a named human; never silently fails. SOX-relevant decisions (vendor onboarding, FX revaluation, IC reconciliation) ship with stricter thresholds and full audit trail.
In a SOX environment, an agent that pretends to be right when it’s not is worse than no agent at all. The whole stack is built around the agent declaring uncertainty and routing below-threshold work to a named human with the full reasoning trace attached. The audit trail tracks every decision — auto-resolved or human-handled — and the PE sponsor’s auditors have the evidence pack one click away.
Three perspectives on the same Monday morning, three weeks after deployment. The AP Specialist saw a 4-item queue (down from 40+). The Controller saw a 240-person team working at 340-person effective capacity. The CFO saw the close at Day 6 of 8, on track.
Karen Mitchell, Senior AP Specialist · US-01 opened her queue Monday morning and saw four items requiring human judgment. Down from 40+ before the agents shipped. Everything else was handled overnight across 9 entities.
One intercompany transaction enters the system. The Main Finance Agent spawns five specialized sub-agents in parallel — each applying a different rule set across two ERPs (Sage Intacct DE-01 and NetSuite US-01). The orchestrator synthesises. The feedback loop captures human corrections. The next similar pattern auto-adjusts.
Extracted 3 line items from DE-01 journal entry. Currency: EUR. FX rate: 1.0842 (ECB date). US-01 counterparty entry located in NetSuite.
DE-01 (Sage Intacct, 5-digit GL: 51000) matched to US-01 (NetSuite, 4-digit GL: 5100-200). Crosswalk applied. FX variance detected: $3,412 (DE-01 used daily rate, US-01 used monthly average).
Line 1 → Oracle 71000 / NetSuite 7100-100 (Mgmt Fees). Line 2 → Oracle 62000 / NetSuite 6200-300 (IT Services). Line 3 → Oracle 65000 / NetSuite 6500-200 (R&D).
Exception Type 4: FX Timing Variance. DE-01 booked at transaction-date rate (ECB daily), US-01 used monthly average. This pattern recurs monthly with EUR entities. Recommend: adjust US-01 to match transaction-date rate.
Draft elimination JE assembled for both NetSuite and Oracle. FX adjustment of $3,412 included. Awaiting human approval for FX variance resolution.
The IC FX variance pattern recurs monthly because of an unresolved policy inconsistency between DE-01 and US-01. After the first human correction, the agent absorbed the rule change. The next 8 similar IC transactions auto-applied the corrected FX policy.
Three real changes from Q1 — entity restructuring, taxonomy updates, expense-tool migration. Each detected automatically and absorbed into the agent rule set without manual reconfiguration.
Brazilian entity restructuring introduced new GL codes. Crosswalk table auto-updated for 6 codes. 2 flagged for human mapping.
Ramp changed category names for 14 merchant types. Agent auto-mapped 12 to existing GL crosswalk. 2 need confirmation (Software vs SaaS distinction).
Concur expense rules now apply to shrinking volume. Ramp rules expanded to cover migrated expense types. Dual-charge detection activated.
Month 1 vs Month 4 at Apex Precision Holdings. The agents got smarter, the close got faster, the team shifted from processing to strategic finance.
Agents get smarter every week. Human corrections and SOP changes are absorbed automatically. Overall accuracy lifted from 84% in week 1 to 97.4% by week 16.
Every agent action with timestamp, reasoning, confidence, and human approvals. Searchable. Filterable. SOX-compliant. Exported for PE sponsor quarterly auditor review.
Apex owns the agents, the data, the rules, the methodology. We did the work; they keep everything. Including the SOX-ready audit pack the auditors love.
Every workflow, every rule, every model. Deployed on their infrastructure, inside their VPC, within their security perimeter. Full PE portfolio company ownership.
Processing happens in their environment. No financial data sent to external servers. Full compliance with PE sponsor security policies and SOX requirements.
Zero vendor lock-in. They keep everything if the engagement ends. The IP is in the methodology, not the output.
The PE sponsors own the building. We designed and built it. The blueprints, the structure, the systems. All theirs.
Range, not point estimate. PE sponsor CFOs read these numbers carefully and our methodology has to survive their auditor review. Below is how the value gets created — each line tied to a specific agent and a specific measurable outcome.
DSO down 8 days (47 → 39) drives ~$840k of working capital release on $42M average AR. Plus early-payment discount capture lifted from 12% to 67% — adds $280–460k/yr in captured discounts. Net release deployed to debt reduction per PE sponsor instruction.
Close cycle 11 days → 6 days. PE sponsor receives monthly financials 2+ weeks earlier — material for sponsor reporting cadence and covenant compliance reporting. Annualized value: faster operating decisions, reduced restatement risk, audit prep compression.
16 AP/AR positions previously running 60%+ on transaction processing have been redeployed to vendor negotiation, cash forecasting, and M&A diligence support. No layoffs — capacity moved up the value chain. Net savings vs hiring strategic finance externally: $1.4–2.0M.
SOX audit findings YTD: 0 (vs 7 prior fiscal year, with $200–400k of remediation cost each). Audit prep compression saves ~30% of Big-4 audit fees. Error-rate reduction on three-way match (84% → 96.4%) eliminates duplicate-payment recovery costs.
All baselines are pre-engagement (the prior fiscal quarter at Apex). DSO benchmarks against APQC top-quartile. Headcount equivalent uses fully-loaded comp ($120k median for senior accountants benchmarked against 2025 Robert Half data). Audit-savings methodology cross-checked with PE sponsor portfolio company benchmarks. Range exists because cohort sizes are small (n=2 quarters of post-engagement data); we tighten the band each quarter. We never bill more than the lower bound of created value.
Saying this matters more in finance than anywhere else. SOX is real. Auditor scrutiny is real. Knowing the limits is the only way to deploy something that survives an audit.
Quarter two: scope expansion to expense audit agent, lease accounting (ASC 842) automation, and tax-provision compiler. Quarter three: BOT (build-operate-transfer) optionality. The methodology is portable; the agents are theirs.