Avonlea Systems

AI consulting and implementation

We take the pulse of the operation and put AI to work inside it.

Avonlea Systems is a Chilean firm for AI consulting and implementation. We bring together hands-on operations, the Lean method and a proprietary memory-and-agents system, 2ndBrain. We diagnose, put the process in order and implement agents that monitor the business's vital signs, with every improvement measured against a baseline the client signs.

  • Transportation
  • Energy
  • Mining
  • Construction and industry
  • Retail
  • Education
  • Technology
Measure Improve Protect Repeat Second brain nothing is lost, everything is reused
  1. Measurebaseline and leaks
  2. Improvelevers with owners
  3. Protectagents and controls
  4. Repeatin the next process
  5. Second brainnothing is lost, everything is reused

01 · The gap

Almost 9 in 10 organizations already use AI. Fewer than 4 in 10 see it in their results.

From adoption to results · % of organizations · McKinsey, 2025

  • 88%use AI in at least one business function
  • 62%are experimenting with AI agents or already run them
  • 39%see some EBIT impact from AI
  • 23%are scaling agents somewhere in the enterprise

Adoption is no longer the question. The question now is how AI moves from a pilot into everyday operations, and how its gains are measured. That gap is where we work.

MIT, 2025: 95% of generative AI pilots show no measurable impact on results. When companies work with a specialized partner, about two in three reach results; internal builds succeed one-third as often.

Sources: McKinsey, The state of AI in 2025: Agents, innovation, and transformation. MIT NANDA, The GenAI Divide: State of AI in Business 2025.

02 · Any given Monday

The answer exists. It is scattered.

Without a shared memory

Monday, 9:00. Management asks which guarantees expire this quarter and how much margin is left on the contract. The answer exists: it sits in an email from March, in two spreadsheets and in the memory of someone who is out in the field today. Putting it together takes the morning. Next Monday, it gets put together again.

With a second brain

Monday, 9:00. The same question is answered in seconds, with the source document in view. An agent has already left a draft of the weekly report; a person reviews and approves it. What was learned on this contract is available for the next one.

03 · What we observe

AI performs when it finds an orderly process and a shared memory.

Our own reading, based on the projects we work on.

Knowledge is scattered.

Contracts, minutes, spreadsheets and agreements live in emails, folders and people. An agent without access to that memory answers without a source.

Without a baseline, improvement cannot be proven.

That is why we measure before setting targets, and every target rests on a figure the client recognizes as its own.

A tool alone does not change the process.

Results appear when every lever has an owner, an indicator and a weekly review.

So we start by measuring, put the process in order, and only then put agents to work, on a shared memory from day one.

04 · The method

First we put the process in order. Then we put the agents to work.

We work in phases, with a client decision at every milestone. In our experience, the fastest return tends to sit in administration and management, before the end-customer experience: that is where we start.

  1. Phase 0 · Diagnosis

    We measure the baseline, locate cash and time leaks, and prioritize the levers. It ends with a roadmap and a baseline the client signs.

    Deliverables: process and data map · quantified leaks · prioritized roadmap

  2. Phase 1 · Automation and Lean

    We organize and automate administrative and management processes: procurement, contract control, progress payments, reporting and documents. Every lever has an owner and an indicator reviewed weekly.

    Measures: hours freed · errors avoided · cash recovered

  3. Phase 2 · Digital transformation

    Capability projects on the business. Each one is assessed with its own business case, and some may move ahead of Phase 1 if the diagnosis shows it.

    • SeeComputer vision, IoT and telemetryWhat happens at the plant, the gate or in the fleet, measured in real time.
    • UnderstandOwn data and public-source radarWhat the company knows, cross-checked against what is published: competitors, tenders, regulation.
    • AnticipateDigital twins and predictionTest an investment or a large order before committing to it.
    • AttendAI agents on WhatsApp and webAnswers in the asker's language, with the company's data, and a handoff to a person when needed.

Cross-cutting · Second brain and agentification

From the first week, project information goes into 2ndBrain and agents start preparing work. What is built in one phase is available for the next.

Working rules

  • Targets are set after measuring, not before.
  • What an agent produces is a draft until a person approves it.
  • Every answer cites its source.
  • Every improvement is measured in cash against a baseline the client signs.

05 · 2ndBrain · Proprietary system

2ndBrain: the company's memory, with agents that prepare the work and people who decide.

It is the system we implement on. It brings together what is scattered across emails, folders and people, answers with the source, and lets agents work on that same memory. We designed and built it at Avonlea, and we use it in our own operation.

  1. Load

    Contracts, drawings, minutes, spreadsheets and manuals come in classified: scope, who can see it, layer and status.

  2. Ask

    An authorized person asks in plain language and gets the answer with the source document.

  3. Propose

    An agent drafts the report, the minutes, the alert or the reply. It stays a draft.

  4. Approve

    A person reviews and approves it. Every change is logged with actor, action and date.

Illustrative

  • Which guarantees expire this quarter, and on which contract?
  • What did we agree with the supplier at the last meeting?
  • What does the equipment manual say about this alarm?
  • What is pending for each area this week?

What it includes

Memory in three layers

Information: what the documents say. Control: what it costs, what risk it carries and when it expires. Actions: what needs to be done and how.

Data kept apart from analysis

Seven statuses, from original source to agreement. What a document says is kept apart from analysis, hypothesis and proposal.

Access by level and area

Every document has a level (internal, restricted or confidential) and an area. Each person and each agent sees what corresponds to them.

Agents with approval

Versioned agents with bounded tasks. What they produce enters a drafts inbox and leaves with approval.

Project management that stays current

Items, dependencies, open items with an owner and a Gantt chart, updated by the same agents doing the work. The record stays alive when the one doing the work keeps it.

Auditable

Every change writes an event with actor, action and content, inside the client's instance.

It does not replace the ERP or the document repository: it uses them as sources. Each client has its own instance, which connects over MCP, the open connection standard, to the AI assistants the team already uses.

06 · Vital signs

Living automation: agents that take the pulse of the operation every day.

Once the process is in order and the memory is shared, agents monitor the business's vital signs against their baseline. They flag what changes and leave the proposal ready for a person to decide.

Agentic AI, in concrete terms

An agent has a bounded task, access to the memory it is allowed to see, an indicator to watch, and a person who approves what it produces. It is not an open chat: it is assigned work, with an owner.

Examples. Each client defines its own during the diagnosis

Illustrative

Cash and margin

Margin by contract, line or store, against the baseline.

Illustrative

Deadlines and expirations

Guarantees, contracts, permits and certifications coming due.

Illustrative

Staffing and shifts

Overtime, absences and coverage against demand.

Illustrative

Procurement and suppliers

Prices, lead times and deviations from what was agreed.

Illustrative

Incidents and risks

Failures, complaints and safety alerts, each with its status.

Illustrative

Market and competition

Tenders, competitor moves and published regulatory changes.

How it beats

  1. Measure

    Each vital sign has a baseline and an expected range.

  2. Watch

    Agents review the data as often as each sign needs: hourly, daily or weekly.

  3. Alert

    What falls outside the range reaches the person responsible. What stays within range interrupts no one.

  4. Propose

    The agent leaves the action ready: the email, the purchase order, the report. A person decides.

Living automation is measured, reviewed and adjusted together with the operation. When the process changes, the agent changes with it.

07 · Proprietary solutions

When a problem repeats across clients, we turn it into a product.

Alongside consulting, we build and run our own products for industries where we already work. They share one standard: verifiable calculation, data in view, and a person who decides.

All industries

2ndBrain

Company memory and agents with approval.

The foundation of every Avonlea implementation: classified documents, answers with sources and agents that prepare the work.

See 2ndBrain

Retail · Services · Security and facilities

Turnomix

Staffing, shifts and the cost of every scenario, in one calculation.

It calculates the required headcount, builds the shift roster and prices each scenario, with each country's labor law modeled. One example: in Mexico, reducing the work week from 48 to 40 hours means 20% more people to cover the same hours.

Calculation: 480 weekly hours require 10 people on a 48-hour week and 12 on a 40-hour week. Work week: DOF (Mexico's Official Gazette).

turnomix.com

Education

Eduk3

Pedagogy at the center; AI handles the rest.

School management with auditable pedagogical AI for Latin American schools: class register, attendance, assessments, lesson planning and communications in one place. AI supports teachers, school leadership and families, and student data is pseudonymized before it reaches the model.

eduk3.ai

08 · Who we are

We run contracts, build product and measure results in cash.

Avonlea Systems is a Chilean firm founded by two partners with more than 35 combined years leading operations, transformation and technology companies in Latin America.

Rodolfo Cuevas

CEO and Founding Partner

Today
CEO of Avonlea Systems.
Background
15 years leading technology companies in Latin America; former General Manager of ICAR, a digital automotive ecosystem.
Education
Industrial Civil Engineer (Ingeniero Civil Industrial), Pontificia Universidad Católica de Chile.

Brings: Digital product, applied AI, and shift and staffing design.

Cristian Astorga

Founding Partner

Today
Global PMO at VOY, a bus operator in Santiago's public transport system: restructuring and technology.
Background
More than 20 years leading operations, transformation and Lean in urban transport, energy, logistics, industry and higher education: LATAM Airlines, INACAP, renewable energy and Metropol (Operations Manager).
Education
MIT Professional Education, digital transformation with AI and technology leadership (G. Westerman).

Brings: Regulated contract operations, portfolio PMO, data-driven Lean management and applied AI.

How we work with each client

Confidentiality

We do not publish client names, figures or cases without their express authorization. That is why this site shows no logos or cases. Each client's information lives in its own 2ndBrain instance, with access by level and area.

Exclusivity by industry and market

In some industries and markets we agree to exclusivity: while the agreement lasts, we do not work with the client's direct competitors in that market.

What gets reused is the method

The method and the product improve with every project. Each client's data, findings and solutions stay with the client.

Our symbol is a heartbeat. The peak and the valley draw an A and a V that share a stroke, like a typographic ligature: AV, for Avonlea. The dot is the signal that keeps transmitting.

Straight answers

Frequently asked questions

Do we need our data in order to get started?

No. Organizing the information is part of the diagnosis. We start with what exists: documents, spreadsheets and exports from the systems the company already uses.

Does 2ndBrain replace the ERP or other systems?

No. It uses them as sources and works on top of them. What changes is that the information becomes searchable in one place, with its origin in view.

Who decides what an agent does?

A person on the client's side. What an agent produces stays a draft until someone with permission approves it, and every approval is logged.

How long until we see a result?

We propose 90-day pilots: two weeks of data and baseline, six of building with a client counterpart, and four of operation and measurement.

How is it priced?

The diagnosis has a fixed fee. Later phases are quoted with their own business case and may include a component tied to the result measured against the baseline.

The first step is a two-hour visit.

We take the pulse of the operation: we look at the data and the processes, and leave with a diagnosis proposal and a first pilot to choose from.

  1. Visit · 2 hours
  2. Diagnosis · 3 weeks
  3. Pilot · 90 days