How a Custom Manufacturer Reclaimed $500K of Exec Time
A PE-acquired manufacturer's COO was burning half his week on VLOOKUPs. An ML stack now pulls live data and quotes jobs automatically — reclaiming $500K of executive time a year.
~$500K
Saved annually in exec and ops time
Michael Cohen
Managing Director
Stalliant
Connect ↓
The Challenge
A Midwest custom manufacturer, newly acquired by private equity, installed a Harvard-educated COO to professionalize operations. Within months, the COO was consuming nearly half his working hours pulling data exports, running VLOOKUPs, and managing a 20-to-30-tab Excel workbook just to generate the information he needed to do his actual job. The business had no clear visibility into cash flow, labor profitability, or cost-per-job. Job quoting relied entirely on the CEO’s institutional intuition, creating extreme key-person risk. Without intervention, the organization would continue bleeding executive capacity to manual data wrangling while flying blind on profitability.
What They Built
Stalliant began with the tools the client already owned — Excel and QuickBooks — building a Power Query-based data model that connected live to the system of record, eliminating manual CSV exports. The stack expanded into Microsoft Fabric and Azure to create data pipelines and nightly snapshots, with machine learning applied to job quoting accuracy and OCR deployed for purchase orders and invoices. The entire solution was built inside the client’s existing Microsoft environment with no new SaaS licenses required. The unexpected outcome: once all operational data was connected, the team identified the most profitable customers and prioritized them — turning a cost-reduction project into a top-line revenue lever that improved customer lifetime value and reduced acquisition cost.
Stalliant began with the systems the client already owned, rather than introducing new tooling — a deliberate choice that eliminated the adoption friction of onboarding new software into an organization already stretched thin.
The first phase established a Power Query-based data model connected live to QuickBooks, the system of record. This eliminated the manual CSV export cycle and gave the COO real-time data access for the first time. The model was validated before any ML capability was layered on top.
The second phase expanded the stack into Microsoft Fabric and Azure, creating structured data pipelines and nightly snapshots across job costs, labor hours, and vendor expenses. With a governed data foundation in place, machine learning was applied to job quoting — encoding historical cost and margin data into a model that could generate quotes without relying on the CEO’s institutional knowledge. OCR was deployed for purchase orders and invoices, eliminating manual document entry.
An unexpected strategic outcome emerged: once all operational data was connected, the team identified the company’s most profitable customers, turning a cost-reduction engagement into a revenue optimization lever.
AI Role
Machine learning models are applied to job quoting, learning from historical cost and margin data to improve quoting accuracy over time. OCR technology processes purchase orders and invoices automatically, extracting structured data without manual entry. Power Query and Azure data pipelines connect live to source systems, automating the data extraction and consolidation that previously consumed the COO's time.
Infrastructure
Microsoft Excel (existing — primary source system and reporting layer) • QuickBooks Online (existing — system of record for financials) • Microsoft Fabric and Azure (expanded data pipeline and storage infrastructure) • Power BI (reporting and visualization) • Power Query (live data model layer connecting source systems)
Integration Points
Power Query connected live to QuickBooks Online, eliminating manual CSV export cycle • Azure data pipelines pulling nightly snapshots from operational source systems • OCR processor reading purchase orders and invoices and writing structured data to the data model • ML quoting model reading from historical job cost and margin data in the unified data model
Impact
~$500K Annual Time Savings
Automation returned approximately $500,000 per year in combined executive and operator time — $100,000 attributed directly to the COO’s reclaimed capacity (50% of his role), with an additional $300,000–$400,000 from employees no longer tied up in manual Excel processes.
5–15% Margin Improvement Per Job
With accurate, real-time visibility into labor hours, cost of goods, and vendor expenses, the client could quote custom manufacturing jobs with precision for the first time — translating to a 5–15% bottom-line margin improvement on individual jobs.
Key-Person Risk Eliminated
All job quoting had previously flowed through the CEO’s intuition — a single point of failure. The new system encoded historical data into an ML-assisted quoting foundation, removing the CEO as the sole dependency on the firm’s most critical pricing decisions.
Implementation Complexity
The solution was built inside the client's existing Microsoft environment (Excel, QuickBooks, Power Query, Microsoft Fabric, Azure), requiring no new SaaS licenses. However, integrating these systems into a unified data model with ML-powered quoting and OCR-based document processing required meaningful custom development and data engineering work.
Best Fit For
Lower-middle-market manufacturers, distributors, or industrials ($10M–$100M revenue) that have recently received PE investment and are suddenly responsible for investor reporting their finance function was never built to produce. Also relevant to PE operating partners who need to rapidly professionalize portco reporting without replacing existing systems.
