Portfolio
Real operations.Measured results.
Eight operations running in production right now. Numbers that were measured, not promised.
automated runs in just 13 days of measured operation
The robot that reads occupational exams and feeds the official system on its own
The problem
Every exam arrived as a scanned PDF and someone had to open it, identify the employee, find the record in the official system, create the entry and attach the document. One by one, every day.
What changed
The team drops PDFs into a shared folder and the robot does the rest: groups files by employee, creates the entry in the official system, attaches the documents and files everything in the matching company folder. Anything that cannot be processed safely goes to a visible review queue with the reason logged, and the robot never creates records on its own.
Measured results
- 271 employees had exams processed automatically within 14 days
- The team now drops the file in the folder and handles exceptions
- Every exception is visible and monitorable, with the reason logged
Timeline and pace: In continuous production, with iterative improvements week after week
first response time to a lead that used to wait hours
The carrier with 300+ trucks that now answers leads in minutes
The problem
Leads arrived by email, WhatsApp and the website and waited hours for a manual reply. The off the shelf CRM could not model freight bids or seven branches, and salespeople spent the day qualifying raw leads.
What changed
A proprietary sales platform with an AI agent at the center, handling email and WhatsApp: she replies in minutes, qualifies, routes the deal to the right track and only hands over leads that are ready. Management follows everything on dashboards by branch and salesperson.
Measured results
- Automated service by email and WhatsApp, 24 hours a day
- Salespeople receive only AI qualified leads
- Automatic follow-up on stalled leads every 7 days
- Every change logged with author, date and previous value, at the group ISO standard
Timeline and pace: About 10 weeks from day one to the AI answering in production
recorded deliveries across 12 months of continuous partnership
Six departments of a trading company on a single platform, at ISO 9001 standard
The problem
Each department lived in its own tool: leads in a spreadsheet, approvals by chat message, imports in another spreadsheet and manual bank reconciliation. ISO 9001 certified, the company could not trace who decided what.
What changed
A single platform where the business flows end to end: the lead becomes a tax simulation, the approved simulation becomes an import process and the process feeds finance on its own. With bank reconciliation, a document hub for the accountants and an immutable log of every change.
Measured results
- Six departments on the same platform, with permissions by area and role
- Reconciliation that never balanced before now matches the statement
- Accounting receives the monthly package through a protected link
- Full decision traceability for the ISO 9001 audit
Timeline and pace: 12 months of continuous evolution, with improvements every week
real defenses from the firm became the AI knowledge base
The AI that drafts legal defenses trained on the firm own case archive
The problem
Each defense took hours of research and writing, with a permanent risk of missing a deadline published in the official gazette. Hundreds of pieces produced over the years sat idle in archives.
What changed
A case management platform with three layers of intelligence: a drafting assistant trained on the firm real defenses, with traceable citations and a hard rule against inventing legal grounds; an agent factory where the lawyer creates specialist assistants; and deadline surveillance with an executive digest at 6pm every day.
Measured results
- 32 thousand searchable passages grounding every new piece
- 458 official gazette publications captured and structured in 30 days
- 100% of publication dates identified for deadline counting
- The daily digest eliminated the manual end of day review
Timeline and pace: Three months from kickoff to full operation, with 76 verified deliveries
interactions processed in just 4 days of measured operation
The virtual attendant who welcomes, qualifies and never forgets a follow-up
The problem
The vaccination clinic WhatsApp depended 100% on the nurses, from triage to repetitive questions. And the follow-ups that build loyalty, like birthdays and post-vaccine check-ins, depended on someone remembering.
What changed
A virtual assistant with her own name on the clinic WhatsApp: she welcomes the client, checks the vaccine catalog, validates coverage by neighborhood and hands the conversation to a nurse when it takes a real person. One non-negotiable rule in the flow: pricing and clinical guidance always belong to the human team.
Measured results
- Peak of over 250 interactions handled in a single day
- Birthdays, reminders and post-vaccine check-ins going out every day
- Nurses focused on what only they can do
Timeline and pace: Three versions of the assistant evolved within weeks
walls extracted and linked to rooms in a single reference project
The digital estimator that reads the technical floor plan and quantifies the job
The problem
In the client own words: the more complex the project, the more time invested, the fewer projects delivered. Measuring and material takeoff were 100% manual from the drawing, taking days to weeks.
What changed
The technician uploads the floor plan, the system extracts every wall, suggests the typology of each one and generates the measurement sheet and the bill of materials, including the structural reinforcements that used to be counted by hand. Formulas configurable by the team and the final decision always human.
Measured results
- Project validated before contract with a proof of concept on a real floor plan
- Roughly 4 weeks ahead of schedule: 6 of 8 contract blocks delivered and audited
- Contracted goal: cut estimating from days or weeks down to minutes
Timeline and pace: A 16 week contract in biweekly sprints, running ahead of schedule
days from zero to the system running sales, inventory, routes and finance
The system where the invoice commands the inventory
The problem
Real inventory never matched the records: invoices issued in a third party system, movements noted separately, and managers unable to see revenue, margin or receivables without building a spreadsheet.
What changed
A commercial system where the fiscal document is the source of truth: the invoice file comes in and stock movements happen on their own, traceable by warehouse. On top of that: orders, delivery routes, automatic receivables, commissions on real margin and a monthly P&L. Validated on mobile too.
Measured results
- Inventory in sync with fiscal reality, with zero manual entry
- 11 feature blocks delivered in 22 days of project
- Managers with real time visibility of sales, margin and receivables
Timeline and pace: Three weeks from zero to the full system, evolving continuously
companies in the sector mapped across the state, 11,204 active in the federal registry
The market radar that mapped 15 thousand potential clients on its own
The problem
Prospecting was manual, clinic by clinic, with no idea how many potential clients existed in the state and selling time spent on research instead of conversations.
What changed
Every week, a robot sweeps search results and online maps for the three client personas across 28 municipalities, extracts contacts, removes duplicates and cross-references every lead against the federal registry, using only official public sources. The result feeds a living database that keeps itself up to date.
Measured results
- Around 190 new leads per month coming in automatically
- 7 out of 10 leads automatically matched against the federal registry
- Next contracted phase: a WhatsApp sales agent, 24 hours a day
Timeline and pace: Data phase designed, built and validated in 3 weeks
The next case on this page could be yours.
Free diagnosis, first delivery within weeks and continuous evolution after go-live.
This page shows a selection of the projects in production; new cases are added as clients approve publication. Client names withheld for confidentiality, and every number comes from internal project measurements and live operation data.