7unmatched
3resolved
1for you
Needs your decisionDecided
A payment landed after the cutoff.
Returned with a noteApproved
Logged with date, owner and reason.
Your operation, automated. Every decision, still yours.
We diagnose with a fixed scope, automate what pays for itself, and every step leaves a trail: what was done, on which source, and who approved it.
The run
today · 06:04- Sources matched
- 2
- Unmatched
- 7
- Already resolved
- 3 of 7
- Rule applied
- differences-are-not-averaged
- Period
- September 2026
- Owner
- Treasury
What it did
summaryMatched the ERP file against the bank statement by reference, amount and date. The entries that did not match were set aside one by one, with their difference and likely cause. None was averaged or absorbed into a blanket adjustment.
Needs your decisionDecided
A payment landed after the cutoff.
The reference is correct, but the money arrived after the close. It proposes reclassifying it into the next period: a closed month is not adjusted.
Returned with a noteApproved · Logged with date, owner and reason.
Logged with date, owner and reason.
Log
Every step, with the source it ran on.
- 06:02
Read both sources and checked that the periods lined up.
erp-entries-sep.csvbank-statement-sep.pdf
- 06:03
Matched by reference, amount and date. Everything that lined up was closed.
- 06:04 · stopped here
Set aside the unmatched ones with their differences, and stopped on the one needing judgement.
rule: differences-are-not-averaged
- 06:07
Published the unmatched-entries file where the team already reads it.
unmatched-entries-september.xlsx
Names and cases are illustrative. The mechanism is real: the agent neither averages nor decides on its own.
We work with regulated operations and multi-country corporatesFinancial services · Insurance and InsurTech · Multi-country retail · Media · Creative industry
The budget is approved. The pilot is still a pilot.
16%
Of AI initiatives scaled enterprise-wide. The other 84% have not.
The blocker is neither interest nor money: only 25% of AI initiatives delivered the expected ROI. What's missing is the boring part: reliable reconciliation, an owner per rule, and the handover that leaves the team running without the vendor.
Source: IBM Institute for Business Value, CEO Study 2025 · 2,000 CEOs across 33 countries and 24 industries, in cooperation with Oxford Economics.
Four fronts. One rule: nothing without a trail.
Every track leaves the same trail: what was done, on which source, under which rule, and where it stopped to wait for a person.
Efficiency diagnostic
The agent executes. Your team decides.
Log
Every step, with the source it ran on.
- day 2
Inventoried the sources feeding every metric.
source-inventory.xlsx
- day 6
Found three different figures for the same receivables and traced each to its origin.
- day 8 · stopped here
Ranked the findings by impact and effort, and stopped on the one your team doesn't own.
rule: impact × effort matrix
- day 9
Published the ranked matrix with the phased plan.
prioritised-findings.pdf
Needs your decisionDecided
The credit finding depends on another committee.
High impact, but the policy that needs changing is not your team's to change.
Set aside for the credit committeeMoved into phase 1 · Logged with date, owner and reason.
Logged with date, owner and reason.
Source reconciliation
The agent executes. Your team decides.
Log
Every step, with the source it ran on.
- 06:02
Read both sources and checked that the periods lined up.
erp-entries-sep.csvbank-statement-sep.pdf
- 06:03
Matched by reference, amount and date. Everything that lined up was closed.
- 06:04 · stopped here
Set aside the unmatched ones with their differences, and stopped on the one needing judgement.
rule: differences-are-not-averaged
- 06:07
Published the unmatched-entries file where the team already reads it.
unmatched-entries-september.xlsx
Needs your decisionDecided
A payment landed after the cutoff.
The reference is correct, but the money arrived after the close. It proposes reclassifying it into the next period: a closed month is not adjusted.
Returned with a noteApproved · Logged with date, owner and reason.
Logged with date, owner and reason.
AI applied to operations
The agent executes. Your team decides.
Log
Every step, with the source it ran on.
- 09:15
Read the month's invoices and checked them against the written rules.
invoice-review-month.pdf
- 09:18 · stopped here
Flagged one for new account and out-of-range amount, then stopped for judgement.
rule: supplier-threshold-01
- 09:18
Attached the explanation of why that one and not the others from the same supplier.
Needs your decisionDecided
An invoice falls outside the supplier's pattern.
If you approve, the pattern updates and the rule stops flagging this case.
Evidence requested from the supplierApproved: the pattern updates · Logged with date, owner and reason.
Logged with date, owner and reason.
Automation and handover
The agent executes. Your team decides.
Log
Every step, with the source it ran on.
- 05:40
Loaded the three sources for the daily report.
- 05:52
Detected an incomplete source, notified the receivables owner and waited without publishing.
rule: do-not-publish-incomplete
- 06:04
Published the report where the team already reads it and logged the run.
operations-report.pdf
Nothing pending
One source arrived incomplete. It notified the file's owner through the agreed channel and waited for the complete file before publishing. The report went out at 06:04 and the run was logged: what came in, what was validated, and who was notified.
Logged with date, owner and reason.
Efficiency diagnostic
Interviews with every process owner and a source inventory. You leave with findings ranked by impact and effort, estimated ROI, and a phased plan your team can run alone.
6interviews9findings1for youNeeds your decisionDecided
The credit finding depends on another committee.
Set aside for the credit committeeMoved into phase 1
Logged with date, owner and reason.
Source reconciliation
Deterministic matching across ERP, banks and operational systems. When sources disagree, it names which one and why — traced back to the source file.
7unmatched3resolved1for youNeeds your decisionDecided
A payment landed after the cutoff.
Returned with a noteApproved
Logged with date, owner and reason.
AI applied to operations
Assistants and models over your own data, with documented criteria, not a black box. We start with one scoped case and a written metric, then expand once it pays off.
41invoices reviewed40within pattern1for youNeeds your decisionDecided
An invoice falls outside the supplier's pattern.
Evidence requested from the supplierApproved: the pattern updates
Logged with date, owner and reason.
Automation and handover
Ingestion, validation, alerting and reporting running unattended, with SOPs and training for your team. It connects to your core: it doesn't replace it.
3sources loaded1self-resolved0pendingNothing pending
Logged with date, owner and reason.
Names and cases are illustrative. The mechanism is real: the agent neither averages nor decides on its own.No figure on this screen comes from a client.
Building is no longer the hard part. What's scarce is the method.
Any team with the right tools can ship something functional in weeks. The hard part is understanding what your operation truly needs and turning it into a plan that pays off, well below traditional consulting. Most companies don't have the internal bandwidth; we do, at your team's pace, not ours.
Diagnose
Interviews with process owners, source inventory and findings ranked by impact and effort.
Closes with your decisionPilot
One scoped case, a written success metric and an assigned counterpart on your side. If it underperforms, we fix it before scaling.
Closes with your decisionBuild
Redesigned processes, integrated tools and gradual rollout. Nothing continues on inertia: each phase is approved or corrected.
Closes with your decisionHand over
Documentation, SOPs and training. Your team runs the full cycle; the retainer stays optional.
Closes with your decision
The order is not negotiable; the calendar is. We work in short sprints or in quarterly phases with a committee, depending on how your organization buys and operates.
The savings start before we build.
Half the value of a diagnostic lies in what you decide not to build. Real figures from regulated operations, multi-country retail, media and the creative industry.
Lead caseCOP 135Midentified saving
Identified at the diagnostic stage, before building anything.
The scope was cut to what moved the decision: the saving came from not building what wasn't needed.
Supervised financial services- Digital insurer
Efficiency diagnostic over an already defined corporate stack
14 findings ranked by impact and effort, each with an owner and a phased plan.
- InsurTech · the partners' own project
Conversational AI service MVP
11 agents on WhatsApp, with a phased roadmap and clear escalation to human.
- Regional news outlet
Monetization strategy and AI in the newsroom
New revenue lines built on the audience they already had, plus AI assistants in the editorial workflow.
- Independent record label
Efficiency audit and process redesign
Annual savings identified and quantified, with administrative processes redesigned and documented.
See if this fits your operationThirty minutes, no proposal attached: we review your case and tell you whether there is a fit.
We come in through your technology function.
We sign confidentiality before seeing a single record and register as a vendor through whatever channel your company requires. No exceptions requested from the committee.
Commitments in writing
- Your data trains no modelsIt goes in the contract.
- Segmented, auditable accessEvery query is logged.
- Deployment where you decideYour cloud, your private network or dedicated infrastructure.
- Board-ready materialEach phase closes with the slide your team presents.
Stack and integrations
We build on what you already chose: we connect to the core already running and licenses stay in your name.
Data and documents
- SharePoint
- Oracle ERP
- SQL Server
- PostgreSQL
Automation
- Power Automate
- Make
- Zapier
- WhatsApp Business API
Models and AI
- Claude
- Gemini
- Copilot
Dashboards
- Power BI
- Looker Studio
- Metabase
Three ways to engage. None open-ended.
Every model is quoted against a written scope. We also structure an annual budget with attached scope when your organization buys by period rather than by deliverable.
Diagnostic
Fixed scopedefined deliverables
- Inventory of sources and owners
- Gap matrix by system
- Findings ranked with estimated ROI
- Phased plan and closing session
Phased implementation
Per phaseat the pace you set
- Pilot with a written success metric
- Processes and workflows in production
- Deployment within your policies
- Documented handover to the team
The diagnostic fee is credited in full when the implementation is contracted within the same quarter.
Request a proposalAdvisory retainer
Monthlyrenewable each quarter
- Defined hours and usage report
- Optimization driven by metrics
- Quarterly review with the committee
- Priority on new fronts
What gets asked before signing.
If something is missing, message us on WhatsApp and we answer without booking a call.
How long does it take?
It depends on your organization's pace. Some companies want a pilot running within weeks; others work in quarterly phases with committee approval. The scope and calendar of each phase are set in writing before we start.
Who does the work?
Vectrum's partners lead every project end to end, with a support team behind them. The relationship is never handed off: the same person designs the solution and defends it before your committee.
Where does our data live?
In your infrastructure and under your access policies. We never train models on your information, access is segmented by role and every query is logged. When data cannot leave the internal network, we work on dedicated infrastructure.
Can Vectrum register as a vendor?
Yes. We go through vendor registration, document review and whatever compliance bond your company requires, and we sign confidentiality before seeing a single record. We start that process in parallel with the diagnostic so the quarter is not lost.
Do you invoice in pesos or dollars?
In the currency you budget with: Colombian pesos or US dollars. We structure by deliverable, by phase or as an annual budget with attached scope, depending on how your organization buys.
Who pays for the licenses?
You do, directly with the vendor. We choose and configure, but licenses stay in your name so no part of the operation depends on us the day we leave.
Whoever interviews you is whoever builds.
Four partners, two fronts: whoever designs the technical solution and whoever owns the business side. Both on the technical front are lawyers with quantitative training: regulatory judgment is there from the design, not after.
Technical front
Pedro Ardila
Founding partnerData architecture, models and implementation. A lawyer trained in data science: one judgment covers both the model and its regulatory framework.
LinkedInMax Duque
Founding partnerTechnology strategy and client accompaniment. Political scientist and lawyer: translates the operation into architecture, reading the institutional and regulatory context it has to work in.
LinkedIn
Business front
Mateo Gutiérrez
Founding partnerBusiness structure and financial backing. Defines how each engagement is priced and measured, and sustains the operation so no pilot depends on an advance from you.
LinkedInMateo Uribe
Founding partnerClient relationship and business development. Fixed counterpart from first diagnostic to handover: answers your emails and takes the project's pulse to your committee.
LinkedIn
Tell us what is breaking.
We book a thirty-minute call, check whether there is a fit and leave with the diagnostic scope in writing. If you prefer the direct channel, message us on WhatsApp.
Received.
We will write to book the thirty-minute call. If you need to move on this today, WhatsApp is the fastest channel.