DYMM Ingeniería is an HVAC engineering company in Colombia that quotes, buys, installs and maintains air-conditioning and ventilation systems for clinics, laboratories, offices and industrial plants. Its whole process lived in Excel, SharePoint and email. In about four months we helped it replace spreadsheets with a custom ERP: Dymm OPS, a nine-role platform that takes a job from the quote, priced with DYMM's own cost analysis and assisted by AI, to the requisition, the purchase order, the scanned warehouse release and the work order on the technician's phone.
Key facts
- Client: DYMM Ingeniería, HVAC engineering in Colombia
- What we built: Dymm OPS, 12 systems in production after roughly four months of construction
- Users: 9 roles on one platform, from management and engineering to warehouse, field technicians and external suppliers
- Catalog: 1,200+ items wired into quoting, hundreds of them carrying DYMM's full cost-analysis breakdown
- Security: 66 of 66 tables under row-level security, with 80 policies isolating data by organization and by supplier
- Quality: 148 end-to-end tests on desktop and 360 px mobile, run before every deployment
The context: a mature process trapped in spreadsheets
DYMM runs a mature process, but it lived in spreadsheets, SharePoint folders and email. With demanding clients, every error in quantities, prices or traceability was a risk.
It is a common pattern in engineering and contracting firms: the knowledge is there, the method works, but it is spread across files that only a few people can operate safely. The company grows, and the spreadsheets do not.
For DYMM, the problems were specific:
- Quoting depended on an expert Excel. Material, accessories, labor, fixed costs, sales factor and the Colombian AIU regime were mastered by only a few people.
- Purchasing had no check against the quote. Requisitions were not matched to the approved quote or to stock already in the warehouse.
- No traceability in the warehouse or tool room. It was hard to know what left, for which job, at what cost and who received it.
- Scattered information and slow decisions. Approvals happened by email, quality forms sat in folders, and there was no dashboard and no audit trail per cost center.
Why a generic ERP did not fit
When companies decide to replace Excel with software, the first instinct is an off-the-shelf ERP. For DYMM, the hard part was never storing data. It was the rules: the cost-analysis formula, AIU with VAT only on profit, quantity control against the quote and the warehouse, concurrent document numbering, stock maintained by triggers and permissions by row.
A generic ERP would have forced DYMM to bend its method to the software: different quote types, different release formats, different numbering. And the cost-analysis formula, the core of how DYMM prices a job, would have stayed in Excel on the side, which is exactly where the risk lived.
So we built software around DYMM's real way of working. Dymm OPS is not a website; it is the company's operating system. This is what custom web application development means for us: the business defines the rules, and the platform enforces them.
How we replaced spreadsheets with a custom ERP: Dymm OPS
Twelve systems in production, built around DYMM's real way of working and still evolving with the team. You can browse the screens on the DYMM case study page.
A spreadsheet turned into a pricing engine
The quoting engine supports DYMM's six quote types, its cost-analysis formula reproduced to the peso, the AIU regime with VAT only on profit, per-item currency with the official exchange rate, and branded PDFs.
DYMM's macro-driven cost analysis is reproduced exactly, including its round-up to the thousand, and the component breakdown of hundreds of catalog items was loaded from the original workbook. The result: the platform reproduces the list price, so any engineer gets the same number the expert would.
The AI proposes, the engineer decides
AI lives inside the workflow, not on top of it. Users can adjust a quote by conversation, generate a per-m² pre-design that learns from approved quotes, and ask a system assistant that knows the record they have open.
In the quote editor, Gemini returns structured change proposals that are validated on the server. Prices for new items come from the catalog, not the model. The user ticks what to apply, and stale proposals are discarded if the quote changed. This is the approach to AI automation we trust in operations: AI helps, but it never invents prices.
Client lists matched to the catalog
Clients can send their own Excel with the list of items they need. The quote importer reads it, keeps their numbering and suggests catalog prices through fuzzy matching on tokens, trigrams, units and sizes, each with a confidence level.
Purchasing against the quote and the warehouse
Requisitions are prefilled from the approved quote. Overages must be justified and are emailed to management. Then comes a three-supplier comparison, the purchase order, the receipt and the invoice.
When a requisition is approved, each line is checked against inventory. What is in stock ships, and only the gap gets bought. The release order reserves stock, partial deliveries send the difference back to purchasing, and stock moves only through database triggers.
Warehouse, tools and field work
Warehouse releases follow DYMM's own format with five sign-offs. Barcode scanning works from a USB reader or a phone camera, and work orders reach the technician's phone with the job, equipment, materials and tool scans in the field.
Cost centers, supplier portal and audit
Cost centers are created automatically on quote approval, using DYMM's nomenclature. A supplier portal lets each supplier manage its documents and orders, and an audit log records every change.
Numbering that survives concurrency
"Read the last code and add one" collided under simultaneous creation and broke past 999. Row-locked counters in the database fixed it, verified with twenty simultaneous creations.
Hybrid AI Coding
A senior WebMinds architect owns the design and every decision with DYMM. AI agents (Claude Code) accelerate implementation, and an independent end-to-end QA suite runs before every deployment. That is how twelve systems reached production in about four months.
Is your quoting or purchasing still living in an expert's spreadsheet? Get your free assessment and we will show you which rules are worth turning into software first.
Under the hood
The hard part is the rules, and the numbers reflect it:
- 66 of 66 tables with row-level security
- 42 database triggers
- 46 versioned migrations
- 120 server actions in 34 modules
- 148 end-to-end tests, desktop and 360 px mobile
- 82 screens
The stack: Next.js with server actions, TypeScript end to end, Supabase (Postgres, Auth, Storage), PostgreSQL row-level security and triggers, Gemini for AI in the workflow, Resend for email, React-PDF for branded documents, Playwright for testing and Vercel for hosting and cron.
Timeline
- April–May 2026: Process diagnosis, proposal approved, kickoff and design mockups.
- June 4, 2026: Construction starts: roles, cost centers, projects, quotes and the first AI features.
- June 22, 2026: Process session with management: quoting rebuilt around DYMM's real method.
- July 2026: Six quote types, branded PDF, AI pre-design, full purchasing cycle, inventory, mobile work orders, tools, quality documents, supplier portal and AI assistant.
- August–September 2026: Role-by-role testing with the client, then exchange rate per item, quote importer, atomic numbering, scanner releases, cost-analysis formula and "Adjust with AI".
- Ongoing: The platform keeps evolving with DYMM.
Results: from spreadsheets to a single source of truth
The impact is about control and consistency, and we report it as such:
- Quoting no longer depends on one expert. DYMM's cost analysis, AIU regime and price list live in the platform. Any engineer produces a consistent quote, and AI helps adjust it without inventing prices.
- What gets bought is what was quoted. Every requisition is measured against the quote and the warehouse: overages are justified and reported to management, and only what is missing gets purchased.
- Traceability end to end. From the quote to the cost center, the purchase order, the signed warehouse release and the work order on the technician's phone, with row-level permissions and an audit trail.
If your company runs on similar processes, see our work on internal automation. For a sibling project, read how we replaced Excel and email with a single platform for Global Sensors, a cold-chain calibration company, or how we built a forensic platform for IRS VIAL.
FAQ
How do you replace spreadsheets with a custom ERP without losing what works?
We start with the process, not the screens. For DYMM, a process diagnosis came first and a session with management rebuilt quoting around the real method. The cost-analysis formula was then reproduced exactly, so the software gives the same number the spreadsheet did, without depending on one person.
Can the platform handle the Colombian AIU regime?
Yes. Dymm OPS applies AIU with VAT only on profit, supports per-item currency with the official exchange rate and produces branded PDFs for each quote type.
Does the AI set prices?
No. Gemini proposes structured changes that are validated on the server. Prices for new items come from the catalog, the user chooses what to apply, and outdated proposals are discarded if the quote changed.
How does purchasing avoid buying what is already in stock?
Requisitions start from the approved quote and each line is checked against inventory. Stock already available is reserved for the job, and only the difference goes to purchasing.
How long did it take to build?
Construction started on June 4, 2026, and twelve systems reached production in about four months, with role-by-role testing alongside the client. The platform keeps evolving.
Does WebMinds work with companies in Colombia?
Yes. WebMinds is an AI and software development company based in Atlanta, Georgia, that works with companies in the United States and Latin America, and several of the platforms it has built run in Colombia.
Ready to retire the spreadsheet?
Read the full DYMM Ingeniería case study or tell us about your process.
Get your free assessment or book a call with our team.