01Service · Data
Data Architecture and GovernanceA number nobody disputes.
We map where the data travels, from where it enters to where it is used, show where it diverges and what the LGPD requires of it, and deliver the data organised, checked and ready to use, through the channels your team already works in.
What arrives
- A charter of indicators, with an owner and an origin for every number
- The automated check that fails the number when two sources disagree
- The ready data served through integration, agent access and a dashboard
How it is boughtOne project, from the assessment to assisted operation.
02Who it is for
Does any of this sound familiar?
- You take a number to the board and cannot say where it came from.
- Each area brings its own number to the same question.
- Nobody can say who sees the personal data.
- When two sources disagree, the decision waits for the next committee.
The person who signs under that number is you.
03At the end
Where the company stands when the work is done.
Every indicator that reaches the board has an owner, a declared origin and the rule that invalidates it.
When two sources disagree, the number is stopped before the dashboard, not discovered in the meeting.
People, systems and AI agents read the same data, each with the access its purpose justifies.
04Instead of
What usually gets tried first, and where it stops.
Another dashboard
Where it stops
It shows the number more clearly, and still cannot say where it came from.
A reconciliation spreadsheet before every committee
Where it stops
It works for as long as the person who builds it stays.
Replacing the system
Where it stops
The data keeps arriving in disagreement, now in a new system.
05In your market
The same problem, under the name it has in your sector.
Real Estate
The portfolio sits in the developer's system, the collector's spreadsheet and the administrator's report, and the three do not close on the same balance. We deliver one base for the portfolio, and the risk analysis starts from a balance with one version.
Non-profit Sector
Donations arrive through more than one gateway, a recurring donor's charge is declined and nobody sees it until the month closes below target. We deliver one base of donors and donations, and fundraising and marketing start counting the same donor the same way.
Franchises and Multi-Unit Retail
Each unit runs a different point of sale, the stock does not close with the day's till, and the revenue the franchisee reports is not what the franchisor sees. We deliver one base for the network, and the store and the franchisor start seeing the same number.
06The proposal
One project, from the assessment to the data in use.
The assessment decides the size of the rollout. We build nothing before the map shows where the data diverges, what it is used for and who answers for it, and the board approves the plan before we start. Then we build the load that delivers the governed data, and each area receives the data in the form its use justifies.
When you also bring in agents or AI products, they read the data through this same delivery and follow the same access policy.
07Delivery
Data Governance Rollout
For when every area has its own number and nobody knows which one counts.
A project in five stages, from the assessment to assisted operation, with the plan approved before anything is built.
Delivery
What stays with you
To know where the problem is
- A map of the data's path, from where it enters to where it is used, with the person who maintains each source
- An LGPD assessment of every flow: what the data is used for, on what legal basis, and who sees it
- A quality profile of each source, with the register of where the numbers diverge and where they have gaps
See the rest
- A charter of indicators: every number that drives the business with a named owner, a declared origin, an update cadence and the rule that invalidates it
- The rollout plan, ordered by what a wrong number would cost, with the record of every decision to cut
For the number to arrive right
- A filling standard with instructions per form, and the team trained on it
- The automated data load in production on Google Cloud, in layers that separate raw, standardised and ready-to-use data, with validation at the door (an ELT pipeline in Medallion layers)
- An AI step that puts each record into the standard, and sends what it cannot resolve to human review
See the rest
- The automated check that fails the number when two sources disagree (a gate)
- The trail from each indicator back to the point where its data was entered (lineage)
For each area to see only what it should
- An access policy by purpose of use, with anonymisation, masking and restriction by area
- An LGPD and GDPR framework where every control exists because a named rule requires it
For the data to reach where you work
- The ready data served through integration with other systems (an API), access for AI agents (an MCP server) and an AI dashboard, all three under the same policy
- The operating manual for the load (runbook)
Stages
How it works, stage by stage
Stage 1: Assessment
We map where the data travels, what it is used for and who answers for it, measure the quality of each source and assess what the LGPD requires of each flow.
Done when the map shows where each indicator diverges and who answers for it.
Stage 2: Plan and design
For each divergence we choose the fix: a filling standard with the team trained on it, an automated rule, or an AI step with human review. The board approves the plan.
Done when the board approves the plan and the order of the rollout.
Nothing moves on without your approval.
Stage 3: Building the load
We declare each indicator and the check that protects it before we build the load that feeds it.
Done when every indicator in the charter arrives through the load, with its check running.
Technical detail
The load is ELT on Google Cloud, or leaner into a data lake when the assessment shows that is enough.
Stage 4: Access by purpose
Each area receives the data in the form its use justifies: anonymised, masked or withheld.
Done when each area receives the data with the access policy applied.
Stage 5: Handover and assisted operation
The data arrives through integration, agent access and a dashboard, and we stay with the operation until your team runs it from the manual.
Done when your team runs the operation from the manual, without us.
Participation
Who from your team takes part
- A sponsor on the board, who approves the plan
- An hour of each area's time in the assessment
- An owner for each indicator
- One of your engineers at each stage
Measurement
How you know it worked
Measured in your own operation, before and after.
- Sources with an identified maintainer, over all sources
- Charter indicators with a traceable path back to source, over the charter's total
- Charter indicators with nobody yet answering for them
- Times two sources disagreed, counted per quarter, before and after the automated check
- Invalid records that reach the people using the data, before and after
- Reads of sensitive data outside a declared purpose
- Time from the board's question to the number it takes to the committee
08After the handover
What keeps running without us.
After assisted operation, your team runs the load from the manual, every indicator stays with its owner, and the automated check keeps failing the number that diverges, without depending on us.
09Questions
Common questions
Will we have to replace our systems?
The map starts from the systems you already use, from where the data enters to where it is used, and the load reads from them. What changes is the data's path between them, not the system your team works in.
How long does it take, and what does it cost?
The assessment decides the size of the rollout, and with it the timeline and the investment. You approve the plan before we start building, so the decision to go ahead is made with the scope on the table.
What about the LGPD? Who gets to see personal data?
Every flow gets an LGPD assessment: what the data is used for, on what legal basis and who sees it. Then an access policy by purpose decides what each area receives: anonymised, masked or withheld.
Who from our team needs to take part?
A sponsor on the board, an hour of each area's time in the assessment, an owner for each indicator and one of your engineers at each stage. That is what keeps the data governed after we leave.
10Other services
The other services we offer
11Contact
Bring the indicator that does not add up
Tell us which indicator does not add up and where its data comes from. Giulliano Soares is who answers.
- Speak with the microphone on
- Type whenever you prefer
- The contact is recorded at the end
Where would you like to start?
Type, or tap Talk by voice.