Olie Execution Engine · for COOs and IT — One step of the process, from the inside. Ten parts move the work, with no manual chasing.

Every process at Olie is a sequence of steps. Each step runs on the Olie Execution Engine, with 10 parts: from the request that comes in with a deadline and a goal to the record of the cost of each run. The simulation below shows the engine at work.

  • 10 parts in each step
  • 4 layers: rules, team, connections and measurement
  • Nobody skips the rule; whoever changes it is on record
The engine drawing

From the request to the next step. Everything that happens in between has a rule.

The orange card is the request: a refund request, a new hire, a proposal. At Olie, each request is a card (the project's record on the board). It goes through two gates, one on entry and one on exit; in between, the step works with people, AI agents, training, deadlines and goals. The 10 parts are numbered on the drawing and in the list below.

RequestSLA + KPIPRE-REQUISITECONNECTIONSAPIMCPWebhookWhatsAppSTAGEOLIE EXECUTION ENGINEAIAIPeople + AI agentsTrainingDeadline + triggerGoalsCost and recordR$ 7.85GOVERNANCEnextOVERFLOWsame card, in parallelLogistics funnelFinance funnel
One step in the Olie Execution Engine. Illustrative values.
In short

What is the Olie Execution Engine? It is how Olie makes each step move by the rules.

It is how Olie makes each step of a process move: you design the step once, with its rules, and the platform applies those rules to every card, every day, without anyone having to chase. To change the behavior, you edit the rule, and the edit is recorded.

Whoever executesKnows what to do now

The step shows what is missing, the deadline and how to do it: a new hire follows the same standard as someone who has been here for years.

Whoever managesStops chasing

What used to depend on someone remembering happens by the rule, on every card, even when the manager is away.

Whoever auditsFinds everything on record

Each run and each rule change stay in the history, with author and date.

Updated on · Source: Olie technical base · Guide to the 77 features

The engine map

Four layers, ten parts. Each part links to the page that details it.

This page is the map. Each part is covered in depth on its own page, with the platform's screens.

Layer 1

Rules

What is allowed, what isn't and when
  1. Piece 1 The rule receivesRequest with a deadline and a goal

    Each request comes in with the SLA (time to start) and the KPI that measures whether it was met.

    See in Algorithmic management
  2. Piece 2 The rule locksPrerequisite on entry

    Without the required data, the request is not distributed, and the AI asks for what is missing.

    See in Governance
  3. Piece 3 The rule locksGovernance on exit

    Without the required items, the card doesn't leave the step: not even the super admin breaks the flow.

    See the two guardians
  4. Piece 4 The rule chasesDeadline and trigger

    Once the deadline passes, the trigger alerts, swaps the owner or escalates to the manager.

    See in Algorithmic management
Layer 2

Team

Who does it, knowing how to do it
  1. Piece 5 The rule teachesTraining at the step

    Video, checklist and the right questions inside the step; the refund example is in Governance.

    See in Governance
  2. Piece 6 The rule coordinatesPeople and AI agents

    The AI prepares, the person decides, and both follow the same rules.

    See in Orchestration
Layer 3

Connections

The step talks to the rest of the company
  1. Piece 7 The rule distributesExecutes or overflows

    With Overflow, the same card starts running in other funnels at the same time, without being duplicated.

    See in Orchestration
  2. Piece 8 The rule connectsConnects with your systems

    Four doors (REST API, MCP, webhook and WhatsApp) take to your systems whatever the rule says.

    See in Integrations
Layer 4

Measurement

What happened, how long it took and how much it cost
  1. Piece 9 The rule showsGoals at each step

    Time, volume and quality with target and limit, and each goal can trigger an action.

    See in Algorithmic management
  2. Piece 10 The rule recordsCost and record of each run

    Who did it, when, how long it took and how much it cost, person and AI separately.

    See in Governance

You design the step once. The engine makes the rule happen on every card, every day.

Simulation

Pick a request and follow it step by step. See the engine's 10 parts at work while the card moves.

Kanban simulation with three examples: sales proposal, customer refund and employee hiring. The card moves through the steps and, beside it, the engine's 10 parts fill in as they act. The demo moves on its own and pauses when you interact.

Loading the simulation…

Text version of the simulation

Sales proposal (Sales and Board)

The quote request comes in and the card is born with a 30-minute SLA. The quantity is missing: nothing is distributed until the AI asks the customer. The rule hands the card to the sales rep for the customer's region, who has the step's training video and two AI agents: one builds the price and margin, the other checks the scope. The customer asks for a 12% discount, above the 10% authority limit, and the card overflows to the Board, which approves it on the card itself. Without the PDF proposal, governance blocks the move, not even the super admin can skip it. Once it moves on, the proposal goes to the customer with the acceptance link and the order goes into the ERP. The customer goes quiet, the 1-day limit runs out and the trigger alerts the manager, has the AI message the customer and flags a priority. Deal closed: the step's goals are measured and the cost of people and AI is recorded with the history.

Customer refund (Customer service, Logistics and Finance)

The customer asks for her money back on WhatsApp and the card is born with a 15-minute SLA. The Pix key is missing: nothing is assigned until the AI asks the customer. Júlia gets the case because she has the shortest queue; one AI agent checks the photo, another looks at the history, and the assistant confirms the policy covers the case. Without a recorded reason, governance blocks the move, and only Customer service can move the card. Once approved, the card overflows to Logistics and Finance. The pickup is late, the 8-hour limit runs out and the trigger alerts the coordinator, hands the case to someone else and has the AI tell the customer the new date. With the product received, the bank sends the Pix, the ERP records the refund and the customer gets the receipt on WhatsApp.

Employee hiring (HR, IT and Finance)

The manager asks to hire an approved sales rep and the card is born with a 4-hour SLA. The start date is missing: nothing is assigned until the AI asks the manager. Carla, in HR, gets the case; one AI agent reads the documents and flags the outdated proof of address, another writes the welcome email. The new hire does not send the new document within 3 days and the trigger follows up, alerts the manager and flags a priority. Without the medical exam scheduled, governance blocks the move, and only HR can move the card. Once approved, the card overflows to IT and Finance, the contract goes out for e-signature and the employee is added to payroll.

In all three: each step measures its own goals and the cost of people and AI is recorded with the history. The same engine runs any process. Names, deadlines and amounts are illustrative.

That is the engine. The next step is to see your process in it.

Execution that doesn't depend on people: the 10 parts run at every step, every day. Then, see how goals and rules decide and follow up.

Next step: 30 minutes to see your process on screen; then, a diagnosis with your leadership.