Menlo Tech

Go beyond chat.Enterprise Autonomy

Menlo combines applied mathematics, AI agents and systems engineering to turn operational processes into measurable outcomes, with clear rules and people where there is uncertainty.

RE/MAX and Clave: optimization with AI and agents.

How intent algorithms and AI agents turn a property search into a conversation with context, from the first request to the keys.

A Menlo film about the RE/MAX and Clave cases: the optimization flow with AI and agents, from search intent to the right page and support for the broker. Scenes from the application in a demonstration environment.

RE/MAX and Clave: the optimization flow with AI and agents

Follow an invoice through the system.

Choose a stage to see what goes in, which rule decides, who executes and when a person steps in.

Demonstration record: invoice T-0421

Invoice T-0421 arrives from the ERP and from a store spreadsheet: two sources, the same due date.

Published indicatorSimulated values
What goes in
Invoices from ERP, spreadsheets and email, with each store’s amount, due date and reference.
Systems and context
ERP, spreadsheets and the already structured portfolio. More than 1,000 invoices per month in the published operation.
Who executes
Data connectors and entity resolution.
Rule or decision
An exact identifier resolves first. The rest is scored: match_score = Σ(wᵢ × similarityᵢ).
What comes out
A structured portfolio, with a quality score per invoice.
How it is verified
Duplicates and incomplete data become pending items and stay out of collections.
When a person steps in
Pairs with an intermediate score go to review.
See this stage in the tour

A demonstration with synthetic data, based on the Subway Brasil tour. The values are not measured results.

The mathematics behind the decisions.

Two examples to play with: how a decision changes under constraints, and how every fact in the system carries its origin.

Deciding under constraints

More contacts recover more, but each one yields less and costs money. Drag the point to see where it pays off, and where the network’s policy sets a limit.

Everything connected, with provenance

Every fact in the system has a source and relationships. Tap a node to see what it is, where it comes from and what it connects to. Drag nodes to rearrange.

  • Franchisee — operates — Store 0421
  • Invoice T-0421 — issued to — Store 0421
  • Invoice T-0421 — governed by — Franchise agreement
  • ERP — source of — Invoice T-0421
  • Collection policy — constrains — Next best action
  • Next best action — acts on — Invoice T-0421
  • AI assistant — executes — Next best action
  • Collection policy — limits — AI assistant
  • Pix payment — pays — Invoice T-0421
  • Reconciliation — reconciles — Pix payment
  • Reconciliation — closes — Invoice T-0421
  • Analyst — reviews exceptions — Reconciliation
  • Analyst — approves beyond limit — Collection policy

Autonomy with limits.

The AI acts on its own within the company’s policies and calls a person when a limit is exceeded. Change the request and watch the decision.

A franchisee’s request

Discount up to 10%: within the limit

Up to 6 installments: within the limit

Confidence of at least 80%: within the limit

Decision

The AI executesEvery rule in the policy was met.

Event log

Send a request to see the log.

A simulation with illustrative limits (10%, 6 installments, 80%). Each client sets its own policies; Menlo executes within them.

Proof over promises.

Each case has an interactive tour that shows the algorithms, formulas and parameters behind the system. What is real data, simulation or proposal is always labeled.

Real Estate

RE/MAX

Programmatic pages and real estate market intelligence

Demand data, geographic entities and eligibility rules become a coordinated digital operation: from demand discovery to staged publishing and property inventory management.

Illustrative example
Open the interactive tour

Operations & Finance

Clave

From a mortgage request to handing over the keys

A mortgage application moves through identity, documents, three banks in parallel, contract and registration, coordinated by agents and supervised by people. Clave checks; only the bank approves.

Proposed architecture
Open the interactive tour

Franchise Revenue

Subway Brasil

Collections across a network of about 1,500 stores

  • About 1,500 stores
  • More than 1,000 invoices per month
  • More than 500 hours of manual work avoided per month
  • Satisfaction of 4.5/5
Published indicatorSimulated values
Open the interactive tour

Published numbers

Indicators from the Subway Brasil collections operation, published by Menlo.

  • ~1,500stores in the network (approximate)
  • +1,000invoices per month
  • +500 hof manual work avoided per month
  • 4.5/5satisfaction

Source: menlopagamentos.com.br/subway. Also published: first response within 1 minute in 100% of cases, 24/7 service and positive ROI in the first year, with no public percentages.

Go beyond chat. Enterprise Autonomy

A system’s advantage lies in understanding how a process works, where its exceptions are and how to prove the result.

Talk to Menlo