The starting point: 80 to 90 systems and a green screen
Madisa is not a startup. As it approached its 80th anniversary, the group's business spanned equipment sales, rentals, parts, service and logistics. Madisa CAT became the first company in the group to move to Dynamics 365, and the group intends it as a foundation for the other companies.
Before the move, the company ran across roughly 80 to 90 systems, including AS/400 and DBS. Opportunities, quotations, invoices, customer records and operational data sat in separate places. Staff worked in green-screen interfaces with function keys. IT spent its time maintaining aging technology and point-to-point connections.
Ruben Vargas, VP of Product Support, described the commercial cost. An opportunity and its quotation could carry different prices, which made it hard to reconcile the sales funnel with real transactions. Customers also needed separate records for each currency.
What was deployed
According to Microsoft, the program unified core processes for more than 2,500 users across 70-plus branches. The product mix is broad:
- Dynamics 365 Finance, described by CIO Jose Cesar Mendez as the backbone for standardizing financial processes and connecting operational and financial data.
- Supply Chain Management, Sales, Customer Service, Project Operations and Customer Insights – Journeys.
- Power Platform, Power BI and Microsoft Fabric.
- NAXT365, an embedded ISV solution from implementation partner XAPT. It supplies dealer-specific capabilities for rentals, service, work orders and Caterpillar parts. XAPT also supported Mexico localization.
The story gives no module-by-module timeline, licensing details or integration diagram.
The dual-write challenge
The most technical item is dual-write between Finance and Dataverse. ERP architect Miguel Alejandro Zapata Sanchez led it. Early customer synchronization could take as long as two days. The team also had to balance dual-write behavior against cross-company data sharing. He says synchronization is now real-time.
Microsoft's documentation explains why this is a hard problem. Dual-write is a tightly coupled, bidirectional integration between finance and operations apps and Dataverse. A data change on either side causes a write to the other. A Customer in finance and operations is reflected in the Account entity in Dataverse, and the reverse is also true.
Company data adds complexity. Business unit and company are not equivalent concepts, so they can't be mapped one-to-one for Dataverse integration. Microsoft's answer is a dedicated cdm_Company table. That is presumably the kind of cross-company issue Madisa's team was balancing, although the story doesn't say which tables, legal entities or settings were involved.
Two caveats apply to the "real-time" claim:
- Microsoft's own documentation describes dual-write as near-real-time.
- Madisa gives no measured latency.
Treat "real-time" as a qualitative description, not a service-level figure. The two-day initial sync is also a useful warning for anyone planning a large customer migration, because initial synchronization of a big customer base is its own project.
People, not just platforms
The program involved about 200 to 250 people. Mendez described a hybrid approach: Scrum-based business teams plus separate workstreams for data migration, technology and change management. On day one of production the company invoiced, paid employees and accepted customer payments.
More than 2,000 employees needed help adopting new ways of working. Change coordinator Marisol Nava's team built:
- role-based training;
- internal instructors;
- centralized videos and guidance in SGT, the company's HR platform;
- six business champions representing local needs;
- Max, a Copilot Studio support agent employees could use before and after go-live.
The story says nothing about Max's knowledge sources, usage or measured outcomes.
Reported business results
All of these are Madisa's claims, as relayed by Microsoft:
- One customer record can support transactions in both pesos and dollars.
- Sales teams can bundle equipment, warranties, maintenance and other services in one offer.
- Caterpillar's SIS parts system feeds quotation workflows, which speeds up repair estimates.
- Reporting that took months now takes hours or days, according to Vargas. The story's own summary says "weeks to hours," so the two framings differ.
- Warehouse picking time fell from a monthly average of 16 hours to five. The story doesn't define the measurement period or scope.
- Employees work from role-based tiles, and some "data citizens" do self-service analysis on live data.
Memo: the customer-facing agent
Memo is a rental quotation agent built with Copilot Studio and Power Automate. It does the following:
- Retrieves current equipment rates from Dynamics 365.
- Discusses availability and rental requirements with customers.
- Gives pricing and package information during the conversation.
- Creates the quote in Dynamics 365 once the customer approves, so a sales representative can continue with a qualified opportunity.
Rental quotes previously took two or three days while staff checked availability, validated specifications and calculated pricing. Memo has been in production since March. The story's sentence doesn't state the year. Martin Dieck Iza, AI and Analytics Coordinator, reported nearly 120 conversations in the previous seven days. A Power BI dashboard for conversions, qualified leads and business impact was still being built.
Madisa says the agent is designed to prevent unauthorized discounting. It relies on prompt engineering and testing to keep prices consistent and to filter out price-sensitive inquiries. That is a design intent. The story offers no audit, no discount-leakage data, and no detail on whether a human reviews each quote or which channels customers use.
Roadmap
Madisa plans to extend Dynamics 365 to more group companies. It is exploring agents that recommend machine purchases and configurations, steer sellers toward high-value opportunities and automate repetitive tasks. Employees have asked for a service quotation agent. The company is also piloting Model Context Protocol with Dynamics 365. These are pilots and plans, not production results.
Analysis: what to take from it
The story's strongest point is sequencing. Madisa built agents only after consolidating its data. Mendez says Dynamics 365, Fabric and Copilot Studio let the team create agents in days. That claim is plausible only because pricing, availability and customer data now sit in one connected platform. An agent on top of 80 fragmented systems would be guessing.
Readers should also keep the limits in mind:
- This is marketing-adjacent content, and it names no failures or cost overruns beyond the slow early sync.
- There are no conversion rates, revenue figures or ROI numbers for Memo.
- Security, privacy, licensing and model-governance controls are not described, and these are the questions IT administrators will ask first.
- About 120 conversations in a week is a modest, early signal, not proof of business impact.
For IT teams considering something similar, the practical lessons are to plan dual-write and company-scoping early. They should budget for initial-sync time, treat adoption as a workstream of its own, and set up measurement for any customer-facing agent before launch.