The problem: a mature process trapped in spreadsheets
The clearest symptom is an operation that works only because a few people hold it together by hand. The signs repeat across industries:
- Quoting depends on one expert. The cost spreadsheet, with materials, labor, factors and regimes such as Colombia's AIU, is mastered by two or three people.
- Purchasing is not checked against the quote or the warehouse. Teams buy what is already in stock, or more than was approved.
- There is no traceability. Nobody knows for sure what left, for which job, at what cost and who received it.
- Information is scattered. Approvals by email, quality forms in folders, prospects on one advisor's phone.
- Reports are assembled by hand. Sales, conversion and billing get added up in a separate sheet that mixes quoted with invoiced.
- Compliance is manual. ISO forms, signatures, origin and transport documents are filled in outside any system.
Every error in quantities, prices or traceability is a risk with demanding clients. Our article on when a custom ERP makes sense instead of spreadsheets goes deeper.
When custom software makes sense, and when it does not
Custom software makes sense when your business rules are your advantage and an off-the-shelf ERP would force you to change them or keep parallel spreadsheets. It does not make sense when your process is standard and an existing tool already handles it well.
Signs that it does:
- Your way of quoting or costing is your own and hard to configure in a generic system.
- Several roles (sales, purchasing, warehouse, technicians, management, suppliers, clients) need the same information with different permissions.
- You need local integrations: electronic invoicing, regional payment gateways, digital signatures or official rate tables.
- You want AI inside the workflow, on your data and your rules, rather than as a separate chat.
Signs that it does not, yet:
- The process is not defined yet, or it changes every month.
- A commercial product covers the process without parallel spreadsheets.
For a fuller comparison, read off-the-shelf ERP or custom software: how to decide.
AI in quoting and workflows: AI proposes, rules decide
AI pays off in operations when it reads, interprets and proposes, while prices, dates and approvals come from rules and people. Here is how that works in our projects:
- At DYMM Ingeniería, Gemini returns structured change proposals that are validated on the server. Prices for new items come from the catalog, not the model, and the user ticks what to apply.
- At SWAP Translators, the website, portals, WhatsApp agent and staff all call the same pricing function. Rates are injected live and never left in the prompt.
- At ANT Mobility, Gemini only interprets the customer's request. The price comes from the deterministic engine and travels as a signed token, so the browser cannot alter it.
- At Kila, Gemini receives the tenant's real catalog of fields, so it cannot invent fields, and nothing lands on a document without a click.
Our article on how to use AI in quoting without inventing prices explains the pattern.
What to measure before you build
Measure your process and your digital presence before anyone writes code. A process diagnosis defines the rules the software has to reproduce. The WebMinds free assessment reviews your website and tools and sends a written action plan in 1 business day, reviewed by Julio Castro, with automation and integration opportunities and an honest view of where custom software would pay off and where it would not.
The assessment covers nine areas of your digital presence, which matter for industrial B2B companies too:
- Speed: mobile performance and Core Web Vitals.
- Design & UX: design audits, headings and the first screen.
- SEO: title, meta description, main heading, canonical URL and crawl files.
- AI readiness: structured data, content depth, llms.txt and AI crawler access.
- AI visibility: whether AI assistants mention or cite your company when asked a buyer's questions.
- Google Maps: listing, reviews and local rank, if you serve customers in person.
- Conversion: form, booking, phone, WhatsApp, email, chat and contact page.
- Analytics: measurement tags installed.
- Trust & tech: security, quality and accessibility signals.
What we built for operations companies in Latin America
Each project is documented as a case study with verifiable numbers.
DYMM Ingeniería: from spreadsheets to a single source of truth
DYMM Ingeniería quotes, buys, installs and maintains HVAC systems in Colombia. In about four months we built Dymm OPS: a nine-role ERP 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. Twelve systems in production, 1,200+ catalog items wired into quoting and 66 of 66 tables under row-level security.
Global Sensors: the operating system for a calibration company
Global Sensors runs ISO 17025-accredited calibration for the cold-chain industry in Colombia. We built one platform with a commercial CRM, a quoting engine with signed PDFs, service orders and invoicing, the client-signature compliance form required before calibration, structured WhatsApp intake and reporting. Fifteen systems in production, live with real data about seven weeks after the first commit.
SWAP Translators: the agency that runs itself
SWAP Translators quotes in seconds with AI, charges in three currencies across seven payment methods with automatic DIAN e-invoicing through Siigo, routes every job to a translator with free capacity and runs each delivery through quality control. Twelve systems in production and 1,600+ quotes generated.
Loski Internacional: EUDR-ready coffee traceability
Loski exports Colombian specialty coffee. In four weeks we built LOSKI TRACE: every lot gets a structured code and an immutable chain of events tied to geo-referenced farms and satellite-drawn plots, with a public QR any buyer can scan to verify the history with SHA-256 integrity checks.
ANT Mobility: freight without the phone calls
ANT Mobility is a Colombian freight platform. Eleven systems, including a SICE-TAC pricing engine, a load marketplace, B2C Express with AI quoting and online payments, OCR-driven onboarding and e-signed trip summaries, are now in final acceptance testing before launch.
Kila DocGen AI: trade documents that fill themselves
For Kila, Kontroll's logistics platform, we built a module in eight days where freight forwarders design their own Bills of Lading, Packing Lists and Arrival Notices, fill them from 162 live data fields and export branded PDFs.
IRS VIAL and IPS Aleso: expert and clinical operations
IRS VIAL moved from reports in Word to a forensic platform with PAdES digital signatures and a SHA-256 chain of custody. IPS Aleso moved from a spreadsheet schedule to a multi-company portal, a floor-by-floor back office and an AI agent that books and escalates by topic.
See more on internal automation, manufacturing, logistics and WebMinds in Colombia.
How we work: Hybrid AI Coding
WebMinds works with Hybrid AI Coding: senior engineering owns the architecture and every decision with the client, AI agents implement, and independent QA checks every change. That method is behind timelines such as eight days for the Kila module, four weeks for Loski and twelve DYMM systems in about four months. Read how long it takes to build a custom operations platform and how secure a custom business platform is.







