Ben VanOuse | Product Management, Business Systems and AI Adoption

My operating belief

AI should work like a cyborg suit, not an autonomous robot.

It should expand what a capable person can see, reason through, and execute while leaving judgment, control, and accountability with the human. I use AI for research, synthesis, drafting, and repeatable execution, then earn trust by shaping that capability into useful workflows people can understand.

What stays humanPurpose · Context · Judgment · Accountability

What AI amplifiesResearch · Synthesis · Drafting · Execution

Start with reality

I begin with how people actually work, where the dependable information lives, and why the existing tools are not giving them what they need. Then I connect the sources and use AI to make that reality easier to act on.

Evidence from the work

Giving teams real-time answers across business systems

Ecommerce product operations | Cross-system product intelligence

Marketing and Customer Service regularly needed product answers about cost, margin, inventory, and likely stockout timing. The source data existed in Salesforce and Shopify, but neither system gave these teams a simple, complete answer, so the questions kept coming back to me. I used Coefficient to bring live Salesforce and Shopify data into a structured spreadsheet that acted as the engine under the hood. I organized the fields and calculations around the questions people actually asked, then built a focused Gemini interface so Marketing and Customer Service could get current answers in plain language without learning the underlying systems.

Proof pointEstimated · 30 to 40 hours/year recoveredEstimated from 30 to 45 minutes of repeated questions avoided each week

The critical reframe

“The system closest to a transaction is not automatically the best place for someone to get an answer. The source systems needed a dependable logic layer and a simpler doorway for the people using the information.”

I bring

Defined the useful questions, selected the authoritative sources, designed the data model and calculations, and established what a trustworthy answer needed to contain.

AI contributes

Provided the approachable question-and-answer layer over the connected, continuously updated product data.

Together

I built and governed the information system; Gemini made it easy for other teams to use without depending on me.

How I approached it

Started with the recurring questions Marketing and Customer Service needed answered
Connected live Salesforce and Shopify data through Coefficient
Built a controlled spreadsheet layer for cost, margin, inventory, and stockout logic
Created a focused Gemini interface that made the resulting answers approachable and immediately available

Where judgment mattered

I determined which sources were authoritative, how the data needed to be joined, which calculations could be trusted, and what the answer needed to show before Gemini ever presented it.

What changed

Marketing and Customer Service could get product answers in real time instead of waiting for me to investigate, and I recovered an estimated 30 to 40 hours per year for higher-value product work.

Make risk explicit

I design around the cost of being wrong, not only the appeal of being fast. That means identifying failure modes, preserving human review where consequences matter, and making uncertainty visible.

Evidence from the work

Avoiding an estimated $4K test environment refresh

Enterprise ERP environment | Controlled transaction testing

A complex ERP match-settings project could not be tested safely in production, but the available test tenant contained information that was roughly three years out of date. A full refresh would have required substantial additional storage and an estimated investment of about $4,000. I used Claude Cowork to analyze the production environment, identify why live testing was unsafe, and design a selective-data plan for the test tenant. Instead of copying everything, we pieced out only the scenarios and records needed to evaluate each match setting.

Proof pointEstimated · $4K refresh path avoided · 12 to 15 hours savedEstimated infrastructure path avoided and estimated project time saved

The critical reframe

“The choice was not between unsafe production testing and an expensive full refresh. The real question was what minimum production-like data each setting needed in order to be tested credibly.”

I bring

Defined the business objective, coordinated with IS, evaluated the cost tradeoff, and decided what level of evidence was sufficient for a trustworthy test.

AI contributes

Analyzed the unfamiliar production setup, surfaced the safety constraint, and helped design efficient test scenarios and data requirements.

Together

I owned the risk and investment decision; Claude accelerated the technical diagnosis and helped turn it into a practical test plan.

How I approached it

Used Claude Cowork to analyze the production configuration and testing risks
Confirmed that production transactions could not be tested safely
Validated the age and limitations of the test tenant with the head of IS
Designed targeted datasets for each setting instead of pursuing a full environment refresh

Where judgment mattered

I accepted neither the fastest unsafe option nor the most expensive technically complete option. I focused the test environment on the evidence required to make the settings decision.

What changed

The targeted plan avoided an estimated $4,000 refresh path, enabled the team to test and improve the match settings, and saved me roughly 12 to 15 hours of independent system learning and test setup.

Prove before scaling

I look for the narrowest experiment that can answer the important question. A small, representative proof creates real evidence while keeping the cost of failure controlled.

Evidence from the work

Proving a Trello migration before scaling it

Corporate accounting operations | Workflow platform migration

A focused Trello board was being created to manage Accounting opportunities against a 1,000-hour efficiency target. The first migration approach created linked tracker cards, which meant comments, attachments, updates, and decisions could split across two versions of the same work. I revised the approach around moving the original Trello records, then tested it end to end with one information-rich card containing multiple owners, labels, and three substantial checklists. Once I verified that the original record could become the destination record while preserving its working context, I applied the method across the scoped Accounting backlog.

Proof pointVerified · 23 original cards consolidated · 27 temporary duplicates retiredVerified scoped migration counts; 1,000 hours was the initiative target, not a completed saving

The critical reframe

“The real migration unit was not the visible card. It was the card's accumulated context and its authority as the single source of truth.”

I bring

Set the scope, recognized the source-of-truth risk, chose what the pilot needed to preserve, validated the result, and remained accountable for the working board.

AI contributes

Accelerated board inspection, source-to-destination mapping, metadata comparison, repetitive record handling, and cleanup while surfacing inconsistencies for review.

Together

I supplied the business context and migration judgment; AI made the inspection, validation, and repetitive execution faster.

How I approached it

Identified the duplicate-record risk in the initial linked-card approach
Selected one information-rich card as the smallest production-relevant test
Verified retained context, ownership, workflow placement, and destination-specific fields
Moved the scoped original records and retired the temporary trackers

Where judgment mattered

I decided that preserving one trustworthy record mattered more than finding the quickest way to copy every visible card.

What changed

Twenty-three original project cards were consolidated into the focused Accounting board, while 27 temporary linked-tracker records from the earlier approach were retired.

Challenge the answer

An output can look polished, pass technical validation, and still be wrong in the way that matters. I trace important results back to their source and challenge anything that does not make business sense.

Evidence from the work

A valid file can still be financially wrong

Industrial supplier invoicing | ERP import validation

While automating supplier invoices into an ERP-ready import, the source document reported Price / Hundred. Treating that field as a unit price would produce a technically valid file with materially incorrect costs. I caught the pricing mismatch, worked with AI to revise the parser so the source value was divided by 100, and then validated the resulting unit costs against quantities and extended amounts on real invoices before allowing the data into the ERP workflow.

Proof pointValidated · $268.45 potential overstatement prevented on one invoice lineValidated comparison · $271.16 uncorrected versus $2.71 correct

The critical reframe

“A file passing technical validation does not mean the business meaning inside it is correct.”

I bring

Brought the process knowledge, financial skepticism, and validation rule that every generated cost needed to reconcile to the source invoice.

AI contributes

Accelerated parser revisions, extraction logic, and comparison of generated records against invoice data.

Together

I caught and defined the financial risk; AI helped turn that judgment into corrected, validated transformation logic.

How I approached it

Questioned an output that was technically valid but financially implausible
Traced the discrepancy to the supplier's Price / Hundred convention
Revised the parser to convert the value into a true unit cost
Reconciled the generated costs against real invoice quantities and extended amounts

Where judgment mattered

I did not treat a successfully generated file as proof that its financial logic was right. Every calculated cost still had to reconcile to the source invoice.

What changed

The corrected transformation produced unit costs that reconciled correctly before reaching the ERP workflow. On one real invoice line, the uncorrected interpretation would have produced $271.16 instead of $2.71.

Reduce the burden

I absorb the complexity so the person receiving the work does not have to. The goal is not to display every step. It is to preserve the meaning, risk, and next action in a form someone can use.

Evidence from the work

Turning fragmented evidence into an executive decision

Product portfolio management | Executive lifecycle recommendation

The case for exiting an underperforming third-party accessory line was scattered across sales, margin, inventory, quality, customer, and supplier records. Leadership needed one clear recommendation before the next replenishment decision. I recommended exiting the line through a controlled sell-through instead of committing to another replenishment cycle. I used AI to synthesize the source files and project history, then directed a Bottom Line Up Front presentation around the evidence that mattered most: revenue down 28%, unit volume down 37%, and weighted gross margin near 29% versus roughly 56% for a benchmark accessory group.

Proof pointProjected · 44% reduction in potential scrap exposureModeled sell-through outcome, not realized savings

The critical reframe

“The work was not to make slides faster. It was to absorb the complexity and give leadership a decision they could understand, challenge, and act on without rebuilding the analysis themselves.”

I bring

Defined the decision, gathered and validated the data, interpreted the tradeoffs, formed the recommendation, influenced leadership, and led the resulting phase-out.

AI contributes

Accelerated synthesis and presentation of the business data and recommendation I supplied and directed.

Together

I owned the decision and its consequences; AI helped turn fragmented evidence into a clear path leadership could act on.

How I approached it

Combined commercial, financial, inventory, customer, quality, and supplier evidence from multiple sources
Formed the recommendation before asking AI to structure the presentation
Directed a concise Bottom Line Up Front narrative for the executive audience
Translated the decision into supplier communication and a coordinated sell-through plan

Where judgment mattered

I decided which evidence mattered, formed the recommendation, adapted the implementation when the team preferred full-price sell-through, and remained responsible for the customer and operational consequences of exiting the line.

What changed

Leadership aligned on the direction, and ten minutes after the scheduled meeting ended I informed the supplier that we would not reorder. I then led the cross-functional phase-out across inventory, ecommerce, Marketing, Sales, Customer Service, and continued warranty support.

Keep following the thread

Thanks for taking the time to see how I work and where I can be useful.

If your team needs someone who can connect product management, cross-functional project delivery, business systems, and practical AI adoption, I’d be glad to talk.

Start a conversation ↗Connect on LinkedIn ↗Download résumé ↓

benvanouse@gmail.com