Finance transformation and AI results, with the numbers

The situation

Every month the FP&A lead audited the profit cube by hand before it went to leadership: 7,215 rows by 144 columns, 19 months and two costing methods. The audit took two hours on a good day and up to two days on a bad one, and the full cycle to the CEO ran about 10 business days. By the time leadership read it, the month was old news.

What we did

• Set up a dedicated project with written instructions, slimmed-down cube files and a role prompt built for steel transfer pricing.
• Defined the audit rules and tolerance thresholds so the model flags what matters and documents what is by design.
• Built a reusable "Profit Cube Audit" skill that runs the variance analysis by year, month, country and SKU and writes a one-page executive summary.
• Added CEO, CFO and pricing lenses so the output answers the questions leadership actually asks, and shared it with two teammates.

The result

The monthly profitability read now takes about 30 minutes and reaches the CEO while the month is still actionable. The first run surfaced insights the manual review had missed, including four product segments that turned negative after tariffs, and caught real data errors before leadership saw them.

By the numbers

The situation

The team did the work by day and the paperwork by night. Every client wanted different reports and closeout packages, estimating output was re-mapped by hand, and the company's procedures lived in people's heads.

What we did

• Started with a seven-question survey so the training used the team's real tools and real pain points.
• Had every participant set up an AI assistant for their role and adopted one prompting standard for the team.
• Built one project per client and skills for the repeatable work: proposals, closeout packages and estimating prompts.
• Turned a live incident in the room into a new SOP the same afternoon.

The result

Every team member left with a personalized assistant and one build assigned to their role. The CEO now turns resolved scenarios into SOPs as they happen, and the corporate manual was already 47 pages and growing during the session.

By the numbers

The situation

Every Monday the owner compared exports from about 10 carrier portals against the CRM to find actives, cancellations and renewals. He was paying for a top-tier AI plan but using it for one-off chats, and three hours of Monday went to building a report before he could act on it.

What we did

• Taught a prompting structure and a model-per-task policy so routine work stops burning the expensive model.
• Wrote five SOPs with him for the recurring work.
• Moved the weekly audit into a reusable skill: drop the export, get the audit and a dashboard.
• Built a scheduled Monday task that downloads the portal exports, with the rule that the AI drafts and he decides.

The result

The Monday report now takes about 10 minutes and comes with a dashboard of how the book and the sales team are doing week over week. He only works the exceptions, and his assistant audits about 30 policies a week from the same output.

By the numbers

The situation

Financial data lived in three places: the ERP, a project tool and Excel. A first reconciliation showed 184 errors, the owner had paid company subscriptions on his personal card since 2019 without reimbursing himself, and 18 million pesos of past-due items were unidentified.

What we did

• Explained when to use chat, when to use an agent that works in files, and when to build a skill.
• Set up a reconciliation pattern: read every PDF statement in a folder, classify each charge, reconcile, produce one Excel.
• Used quick prototypes to define what the BI dashboards should show.
• Put guardrails in place: default to the efficient model and never put passwords in the tool.

The result

The owner ran all his statements from 2019 to date through the pattern himself and found about 400,000 pesos the company owed him. The finance lead cut the unidentified past-due balance from 18 million pesos to 7 million, and a three-system reconciliation now takes about 15 minutes.

By the numbers

The situation

After every discovery call, handwritten notes went to an assistant, and the advisor wrote each client summary and recommendations by hand. It added up to about seven hours per client, and in one case $171K in his notes was keyed as $17K.

What we did

• Started with security and compliance: manual approval by default and clear rules on what data goes where.
• Built a client profile skill trained on five real examples across life stages.
• Set up one project per client, a notes connector, a morning briefing and calendar booking.
• Made data validation part of the skill so a mistyped figure gets flagged, not published.

The result

The skill now builds a complete client profile with recommendations in under an hour, with deeper recommendations than the handwritten version. About six hours per client are freed up for time with clients, and the work continues on a weekly cadence.

By the numbers

The situation

Collections, billing, credit and cash application ran country by country, each with its own team, process and cost base. Invoicing was still on paper in several markets, and credit decisions took a week, so sales waited on credit before it could ship.

What we did

• Designed and ran an outsourcing strategy for collections in the US and Mexico, with clear SLAs and governance.
• Centralized collections across eight countries into one operating model.
• Automated credit approvals so a decision that took seven days took one.
• Implemented paperless invoicing across the region.

The result

The outsourcing strategy delivered $23M in savings over five years. Centralizing collections saved a further $1.4M a year, paperless invoicing $450K a year, and faster credit approvals enabled $4M in annual revenue growth.

By the numbers

The situation

Payment terms had grown vendor by vendor with no policy behind them, so the company paid faster than it needed to. At the same time AP, master data, T&E and other back-office functions had to move to a new operating model, under SOX, without missing a payment run or a close.

What we did

• Co-led payment-term renegotiations with Procurement, backed by one clear policy.
• Partnered with Treasury on banking and credit programs so vendors were paid on their terms while the company held cash longer.
• Led the transition of AP, master data, T&E and other back-office functions to the new model in three months.

The result

Working capital improved by about $30M a year, days payable outstanding improved by 22 days, and the back-office transition finished in three months with zero disruption and 100 percent SOX compliance.

By the numbers

The situation

Four percent of a $450M portfolio was more than 90 days past due. Collections practice varied by market and did not line up with the lien laws that protect a building materials supplier, and credit committees were slow and inconsistent from state to state.

What we did

• Standardized collections with market-specific strategies aligned to each state's lien laws.
• Chaired the nationwide credit committees and set one credit policy with clear thresholds.
• Led a team of 73 credit and collections professionals across more than 7,000 customer accounts.

The result

Receivables more than 90 days past due fell from 4.0 percent to 1.6 percent of the portfolio. Collections ran the same way in every market, and credit decisions became faster and consistent.

By the numbers