Build

Your manual processes can run on their own.

We design and build AI systems that do the heavy lifting of one concrete process, integrated with what you already use.

A system in production

It comes in where the work already comes in. It goes out where it already gets loaded.

A system of ours is not one more app to open. It reads what arrives through the channels you already use, decides with your company's rules, loads into the system you already have. And when it doesn't know, it doesn't guess: it tells a person.

It comes in where the work already comes in. It goes out where it already gets loaded. WhatsApp Email Documents and PDFs Spreadsheets The system Reads, decides, loads Reads Decides Loads With your company's rules, in writing. ERP or management system CRM The usual spreadsheet Alert to the right person When it doesn't know, it tells a person.
40
systems we know how to build
3
running in companies with names
7–10
weeks to production
0
licenses we sell: everything stays in your accounts

The starting point

The symptoms are known. So is the underlying problem.

  • Inboxes sorted by hand.

    Someone reads every email and decides where it goes.

  • Data trapped in documents.

    That someone has to read, understand and type into a spreadsheet.

  • Questions that always land on the same two people.

    And when they're out, nobody knows.

  • Systems that don't talk to each other.

    Data gets copied from one to the other, sometimes wrong.

  • Inquiries that arrive after hours and get lost.

    At the end of the month nobody knows how many there were.

All of the above is the same thing: work a person does today because there was nothing to replace it with.

Not everything needs an agent

Sometimes a simple flow is enough.

Other times you need something that understands language: reading a tender, answering an inquiry, building a quote. The key is knowing which of the two it is, and that's decided in the diagnostic, not in the sale.

01

A simple flow

When the data is already structured and the rule is clear.

  • Move the order from the spreadsheet to the ERP
  • Reconcile the bank statement against invoices
  • Flag a due date that's getting close
02

Something that understands language

When something has to be read, interpreted or answered.

  • Read a 200-page tender and pull out twenty data points
  • Answer a WhatsApp inquiry with your real stock
  • Draft a quote from an email

Which of the two it is gets decided by looking at the process, not the catalog.

What we build

Three systems with names, not a catalog.

Each one with the number, the period and how it was measured.

Engineering and public tenders · INTED

From three days to a few hours.

−85% · reading time per tender, first three months

Reads tender documents and pulls deadlines, requirements, amounts and guarantees into a spreadsheet, with a link to the source paragraph.

How it was measured Hours per tender before and after, over the first forty tenders.

Paper and supplies distribution · Market Paper

Prices get reviewed every day.

1,200 · products compared against competitors per day

Watches competitor prices every day and flags when it's worth moving yours.

How it was measured Products covered and alerts per week, over the first two months.

Wholesale e-commerce · e-commerce, name with permission

The wholesale flow runs on its own.

−70% · loading time per wholesale order, first quarter

Answers and takes orders on WhatsApp, loads them into the system and flags when a person needs to step in.

How it was measured Minutes per order before and after, over the first quarter's orders.

House rule: if a number can't be defended with a period and a method, the card goes without a number and with the fact. A number without a method is worse than no number.

What we know how to build

What we know how to build.

Forty systems grouped by area of the company. Three are already running somewhere. None is sold by the unit: which one comes first is decided in the diagnostic.

Systems in the catalog
42
Running in a company
3
Areas of the company
5

Sales and customers 11 systems

01

In production

WhatsApp service and order taking

Answers, quotes, loads the order into the system and flags when a person is needed. The most common pain, and the easiest to quantify.

Integrates with WhatsApp, your price list and the system where orders are loaded.

See the case
02

Quoting and proposal assembly

Takes the order and the price list and builds the quote, ready to review and send. Replaces copy-pasting from old templates.

Integrates with Your price list, email and the management system.

03

Customer follow-up and reactivation

The one who asked and didn't buy, the one who hasn't come back in six months. With judgement, not spam. New revenue with no acquisition cost.

Integrates with WhatsApp, email and your customer base.

04

Appointment, visit and delivery reminders

Confirms, reminds, reschedules. Cuts no-shows in healthcare, workshops and home services.

Integrates with WhatsApp, your calendar and the bank statement.

05

After-sales

Order status, simple claims, warranties, returns.

Integrates with WhatsApp, email and the order system.

06

New customer onboarding

Sign-up, documents, verification, first contact. Strong in financial services and healthcare.

07

Lead qualification and routing to the right rep

By region, by product, by size. Raises conversion without adding salespeople.

08

Cart recovery and e-commerce follow-up

Timely contact with real context on what the person looked at.

09

Multilingual customer service

For companies that export or sell across the region. Multiplies the market without adding staff.

10

In production

Price and competitor monitoring

Watches other people's prices every day and flags when it's worth moving yours.

See the case
11

Voice-of-customer analysis

Reviews, complaints and conversations grouped by topic, with the trend over time.

Administration 10 systems

12

Collections

Reminders before the due date, follow-up after, escalation to a person when it's warranted. Pays for itself by speeding collection up a few days.

13

Supplier invoice reading

Pulls the data out and loads it into the system. The most universal document-extraction case: every company with a back office has it.

14

Automatic reports with alerts

Pulls data from several sources, builds the periodic report and flags when a metric drifts. Gives administration a day back every week.

15

Invoice-to-purchase-order reconciliation

Cross-checks, flags differences and leaves only the mismatches for review.

16

Bank reconciliation

Matches bank movements with invoices and payments. High value for companies with many small transactions.

17

Expense control and reimbursements

Reads the receipt, categorises it and loads it.

18

Margin analysis by product and by customer

Brings together real costs, discounts and logistics. Many mid-sized companies don't know which customer loses them money.

19

Data quality control in the ERP or CRM

Duplicates, empty fields, made-up categories. The mandatory first step for almost everything else, and a good small first project.

20

CV screening and ranking

With explicit, auditable criteria. For companies that hire seasonally.

21

Employee onboarding and HR answers

Leave, payslips, policies. Lowers the admin load on the HR team.

Operations 6 systems

22

Email and inquiry triage and routing

Sorts, routes to the right team and answers the simple ones. Attacks the overflowing inbox, the symptom everyone recognises.

23

Internal assistant over company documentation

Answers the team's questions about policies, processes, products and customers. Removes the dependence on “the two people who know”.

24

Meeting minutes and follow-up

Summary, decisions and tasks with an owner. Cheap to build and adopted immediately: the team notices on day one.

25

Data entry between systems that don't talk

What today gets copied by hand from a spreadsheet to the ERP, from email to the CRM. Unglamorous, and one of the biggest hour savers.

26

Search across the company's history

Past quotes, projects and cases, searchable in plain language. “What did we quote for something like this in 2023?”

27

Demand forecasting and replenishment

How much will be needed and when to buy it. Hits working capital directly.

Documents and tenders 5 systems

28

In production

Tender document reading

Pulls deadlines, requirements and amounts into a spreadsheet, with a link to the source paragraph.

Integrates with Email, the folders where documents arrive and your spreadsheet or system.

See the case
29

Technical proposal drafting

From the tender and previous proposals, builds the draft. The natural next step after reading, and it multiplies how many tenders you can bid on.

30

New tender monitoring

Scans public procurement portals and flags the ones that fit the company's profile. Sells itself to anyone who lives off tenders.

31

Contract and policy data extraction

Expiry dates, clauses, amounts, automatic renewals. Avoids the fine or the renewal nobody saw.

32

Contract version comparison

What changed between the version you sent and the one that came back signed.

By industry 10 systems

33

Car dealerships and used-car agencies

Handles the leads that arrive through MercadoLibre and WhatsApp, qualifies them, books the visit and leaves a count at month end. That's Raudo, launching in October.

34

Routing and logistics planning

Loads, time windows, fuel costs. For distribution and last mile.

35

Predictive maintenance

On sensor data or failure history. Industry and fleets.

36

Visual quality control

A camera on the production line to catch defects. Food, plastics, auto parts.

37

Export traceability and documentation

Certificates, customs, per-country requirements. Agriculture and food.

38

Claims management

Report, documents, assessment, follow-up. Insurance and healthcare.

39

Appointments and billing for healthcare services

Reminders, confirmations and the billing cycle. Medical and dental centres.

40

Construction site tracking

Progress, certifications, deviations from the schedule. Construction.

41

Recipes, stock and waste

Real costing per dish or per product. Restaurants and food.

42

Compliance and regulatory reporting

Assembles the periodic filing from the operational data.

Which one comes first is decided in the diagnostic, not in the sale.

How we work

We don't sell AI. We sell that it works.

We understand how your company works today and build the bridge so AI does the heavy lifting, with your team inside.

  1. 01 Weeks 1–2

    Discover

    We talk for 45 minutes. Then we sit down with the people who do the work. You leave with the prioritized page.

  2. 02 Weeks 3–6

    Pilot

    One thing, small, with real data. We measure before and after. No slide decks.

  3. 03 Weeks 7–10

    Adopt

    What worked goes to production and the team takes ownership: training on their cases, use policy, a visible metric.

  4. 04 Weeks 11+

    Scale

    What's next. New processes, new areas. AI stops being a project.

Every stage can be the last one: at each step you decide whether we continue. We give the numbers in the first conversation.

Questions

The four that always come up.

What tools do you use?

Whatever the problem calls for. We don't sell licenses and we don't have a stack to push. Everything stays in your accounts and your repository.

Will you break the system that works today?

No. We build alongside it, integrated through APIs, without touching what runs. What shouldn't be touched goes in writing.

How long does it take?

Weeks, not months. One to two to discover, three to six to pilot small with real data, seven to ten to go to production with your team inside. Every stage can be the last.

Who maintains it afterwards?

At first we do, with your team watching. Then whoever you decide: everything stays documented and in your accounts, and it's built so you can keep going without us.

Let's talk

Are we the right fit for your company?

Tell us what's going on and we'll tell you frankly whether we can help, and how.

  • We reply within 24 business hours
  • The first conversation is free
  • We come prepared: you leave with three concrete things for this quarter, whether you continue with us or not.

Rather write?

We use your data only to reply. We don't share it or add you to any list unless you ask.