Autonomous agents with memory and context — patent filed with the USPTO

Hal-AI Systems · 2CW group

An AI company built on memory and context

Hal-AI Systems was founded in 2022 inside 2CW — Cloud & Inteligência Artificial, a group with more than 35 years behind it in enterprise infrastructure and software. The company builds autonomous agents that serve customers and run processes across aviation, financial services, retail, logistics, healthcare, HR, education and real estate.

NVIDIA Inception · Meta Business Partner · AWS Partner

The company in numbers

+35 years of experience in the 2CW group
2022 the year Hal-AI launched
USPTO patent filed in 2025
IEEE Senior Member since May 2025
The thesis

An AI agent is only useful to operations when it can hold on to context

That realization shaped what Hal-AI built — and what it chose to patent.

An assistant that forgets what was said two messages ago resolves nothing on its own. It hands the problem back to a person. That is why most customer service automation ends up as a tree of menu options: the flow is drawn to compensate for the system's missing memory.

Hal-AI went after the root cause instead. Rather than scripting the path a conversation has to follow, the company built the layer that gives the agent what it needs in order to decide: memory of what has already happened, retrieval of the relevant context at the right moment, and access to the tools that carry out the action inside the customer's systems.

This is not a chatbot running a flow. It is an operational agent trained to do the work.

  • Context that carries across channels and sessions instead of resetting with every message.
  • A real decision at every turn: answer, query a system, or bring in a person.
  • One foundation serving both customer service and internal operations.
An AI humanoid reaching out to touch a Hal-AI chip seated on a circuit board USPTO 2025
USPTO · 2025

The memory that keeps an agent in the conversation

Patent Autonomous Agent Memory Framework Inspired by Human Cognition

A memory and context framework for autonomous agents, inspired by human cognition. It covers what the agent holds in the short term, what it keeps for the long term, how it retrieves the context that matters at each turn, and how a task stays continuous from one conversation to the next.

That is what separates an agent that holds the thread from an automation that starts over with every message. It is Hal-AI's central piece of technology: the reason an agent can hold on to context instead of depending on a scripted flow.

Already in use at companies in logistics, aviation, healthcare and banking.

How it works on the platform
2CW group

Three values Hal-AI inherited

These are the values 2CW publishes, and they apply to Hal-AI just as directly.

Honor and ethics are obligations

Not a differentiator, and not a line in a pitch deck. They are the precondition for operating on another company's data — and for stating plainly what the technology does and what it does not do yet.

Technology has to create value

Technology that moves no number in the operation is a cost. Every project starts from a real process at the customer and from what the agent will actually resolve there.

Grit, passion and expertise

The combination that sustains original research in a field that shifts every quarter — and that led the company to file a patent rather than simply integrate what already existed.

Timeline

From launch to patent

Hal-AI's public milestones, in order.

2022

Hal-AI launches

The 2CW group's artificial intelligence practice gets a brand and a product of its own, built on the group's track record in cloud and enterprise systems.

2024

NVIDIA Inception

The company joins NVIDIA's program for organizations developing their own AI technology.

Apr · 2025

Patent filed with the USPTO

Autonomous Agent Memory Framework Inspired by Human Cognition — the memory and context framework for autonomous agents is filed with the United States Patent and Trademark Office.

Apr · 2025

Cases in aviation, pharmacy and payroll-deductible credit

The cases are published: a major Brazilian airline handling check-in and support in more than 40 languages over WhatsApp; one of the country's largest pharmacy chains, integrated with Alpha 7; and the partnership with PHNIX Card in payroll-deductible credit.

May · 2025

Marcos Alves elevated to IEEE Senior Member

The founder and CEO receives the IEEE Senior Member grade, awarded through peer review.

May · 2025

International press recognition

ValiantCEO names Hal-AI a competitive alternative to global enterprise automation platforms.

The product

The patent ends in a working screen

Memory and context aren't a paper: they're the dashboard the customer opens in the morning.

This is the first screen of the Hal-AI console, the one a manager lands on after signing in. Up top, the day's numbers: the channels connected through the official Cloud API, the agents on the air and the conversations handled. Below it, what the Squads did on their own since the small hours — with the run still in progress in plain sight and the stop button right next to it.

Hal-AI · Home dashboard Company console
Home dashboard The day's numbers, the connected channels and what is running right now.
Connected channels
5 on the airOfficial Cloud API
3 official WhatsApp numbers, 1 webchat and 1 voice channel.
Aurora Store WhatsApp Vitoria Clinic WhatsApp Meridian Transport WhatsApp Horizon Academy webchat and voice
Agents on the air
6 active1 paused
Seven agents configured; six answering right now.
4 with long-term memory 3 with APIs connected last change yesterday 5:20 pm voice on
Conversations today
In progressthrough 4:40 pm
512 conversations — 448 resolved by the AI and 64 handed to a person.
Conversations per day — last 7 daysall channels combined
Conversations today512 Resolved by the AI448 Handed to a person64
Run #47 · Daily Collections on schedule · 2m14s api_get_invoices → 18 cases read · 9 reminders drafted
Squad runs so far today
#SquadTriggerOutcomeCases readStarted
47Daily Collectionson scheduleRunning18today 8:00 am
46Appointment confirmationson scheduleActed24today 7:30 am
45Inventory sweepby buttonChecked, took no action31today 6:00 am
Cases read today — 18 + 24 + 3173

Home dashboard: connected channels, agents on the air, the day's conversations and the Squad runs, with the one still going highlighted.

What you are looking at

The three cards at the top are the state of the operation right now: five channels connected through the official Cloud API — three WhatsApp numbers, one webchat and one voice channel — seven agents configured with six of them answering, and the day's 512 conversations split into 448 resolved by the AI itself and 64 handed to a person. Below that, the table lists the Squad runs since the small hours: #47 still going, #46 which acted, and #45 which checked and concluded there was nothing to do — 18 + 24 + 31, 73 cases read on the day.

Why it matters

The manager opens one screen and knows whether the operation is standing, without asking anyone for a report. A run that checked and took no action shows up as a legitimate outcome, not as a failure swept under the rug; and the one in progress comes with the stop button right beside it, not buried in a settings menu. That is the difference between trusting an autonomous agent and being held hostage by one.

Only at Hal-AI

Notice that the dashboard mixes two things that normally live in different products: the 1:1 conversations on the channels and the autonomous work that runs by itself overnight. Both sit on the same memory and context layer — the one filed with the USPTO — which is why a Squad can send a plain-language instruction to the channel agent and get back "this customer already confirmed yesterday, so I didn't send anything."

  • The same history serves both live conversations and the scheduled routine
  • Every run stays in the history, with its trigger and outcome stated
  • The stop button is always in sight on a run in progress
Leadership

Who runs the company

Founder and CEO

Marcos Alves

Founder and CEO of Hal-AI. Author of the patent Autonomous Agent Memory Framework Inspired by Human Cognition, filed with the USPTO in 2025, and an IEEE Senior Member since May 2025 — a grade conferred by the world's largest technical professional association through peer review.

IEEE Senior Member Author of the USPTO patent

The present is theirs; the future, for which I really worked, is mine.

Nikola Tesla

The line the company keeps as a statement of posture: build the technical foundation that carries tomorrow's product, not today's demo.

Credentials

What each badge actually means

Partnerships and recognitions that change what Hal-AI can deliver to a customer.

NVIDIA Inception

NVIDIA's program for companies developing their own AI technology — access to technical and compute resources for Hal-AI's research.

Since 2024

Meta Business Partner and Tech Partner

Hal-AI runs on the Official WhatsApp Business Cloud API. In practice: a verified number, approved templates and compliance with Meta's rules — no workaround with a phone plugged in somewhere.

Official WhatsApp Business Cloud API

AWS Partner

Cloud infrastructure to run the agents at enterprise scale, drawing on the 2CW group's 35 years of hands-on work in mission-critical environments.

Cloud and scale

IEEE Senior Member

A grade conferred on Marcos Alves through peer review at the world's largest technical professional association — a sign that the company's research is read by people in the field.

May 2025

Patent filed with the USPTO

The memory and context framework for autonomous agents is the company's own intellectual property, not a third-party component wired in.

2025

2CW group

Cloud & Inteligência Artificial. More than 35 years of experience in enterprise infrastructure and software standing behind Hal-AI's operation.

Parent company
Meta Business Partner AWS Partner Network NVIDIA Inception Program

Programs Hal-AI takes part in. Trademarks belong to their respective owners.

Aviation Financial services Retail Logistics Healthcare HR Education Real estate

Sectors where Hal-AI is already at work.

Press

How Hal-AI reads from the outside

ValiantCEO named the company a competitive alternative to global enterprise automation platforms — the reading that a Brazilian company, with research and a patent of its own, is contesting the same ground as the large international automation suites.

The cases published in 2025 cover aviation, pharmacy and payroll-deductible credit. The aviation and pharmacy customers are not named in the source; the payroll-deductible credit partnership is with PHNIX Card, running on the WhatsApp Business API with full traceability of the operation.

Open-plan office with rows of workstations and customer service headsets

Talk to the people who built the platform

Bring us a real process from your operation. Our team will show you where an agent's memory and context change the outcome — and what would make sense to automate first.