Frequently asked questions — What you can do with Olie, answered one question at a time.

At Olie, work is designed in stages, and each stage can be carried out by a person, an AI agent or an API call. The same project goes through a salesperson (human), an assistant that qualifies it (AI) and a webhook that writes to the ERP (API), in the order you defined, with a restriction, a deadline and a record at every handoff. That is what orchestrating means: deciding who does what, in what order, and having the history of everything in one place.

Yes. There are no closed modules: CRM, service desk, approvals and onboarding are combinations of the same building blocks, which are pipelines, stages, forms, automations, AI agents and channels. You build the flow your operation needs and swap a block without redoing the rest. A clinic and a distributor use the same blocks in different arrangements.

It is where the company's process runs: the rule chases the deadline, locks the stage when data is missing, records who did what and shows where things are stuck. AI works inside the flow, as the executor of stages, not as a chatbot on the side. Unlike a task board, which only shows cards, Olie executes the process.

It is the operation being run by explicit rules instead of personal chasing. The algorithm is the set of restrictions, deadlines, automations and agents you configure; it decides when a project moves forward, who gets notified and what happens if nobody acts. The manager starts designing the rule and following the result, instead of chasing the team.

On a task board the task is the card and someone has to remember to chase it. At Olie the task is the stage, and the stage has a rule: a required field that locks, entry and exit restrictions not even the admin can skip, a deadline that mentions the owner and escalates to the manager. The card still exists, but it moves by the rule, not by someone's memory.

It covers what you expect from BPM, which is designing and executing processes, but without the formal modeling layer. You don't draw a diagram: you build the pipeline with stages and rules and the process is already running. The practical difference is that execution, the conversation with the customer, the forms and the AI all live in the same object, the project.

The card (project) is the live demand that moves along: the order, the ticket, the lead. The stage is where the rule, the data and the action live. When the order reaches Billing, it is the Billing stage that requires the invoice, sends the email and chases the deadline. That is why the same card can cross departments without being copied.

It is the person inside the automated flow, at the stage where the decision needs to be human. The agent qualifies the lead, the automation assigns it, and the proposal only goes out after the manager approves, even by replying on WhatsApp. AI handles the volume; the person makes the decision, and the record keeps both.

Day to day it is no-code: pipelines, forms, automations and agents are configured on screen, without programming. For those who want to go further, there is the advanced syntax (expressions to calculate and build text), the open API and webhooks. You start without code and grow when you need to.

Managing is knowing where each thing is; executing is making it move. Most tools solve the first part with a board. Olie solves the second: the deadline chases, the rule locks, the agent executes the stage, the automation moves the project forward. Visibility comes for free, because execution happens inside the platform.

It is an environment where AI agents have tools, context and a place to act. At Olie the agent brings together a model, tools and instructions; placed on a stage, it becomes a stage assistant and runs when the project arrives, writing the result to a field, moving the project forward or calling another agent. Several chained agents form an agent pipeline.

Yes. The Olie agent doesn't just answer: it acts inside the platform with the tools you allow, looks up forms and projects, searches the internet, writes to fields and triggers automations. Every execution is recorded with the model, tokens, cost, input and output.

It does, by composition. An AI agent looks up the sales, finance or operations forms; an idle-time automation runs that agent every 10 minutes; the output is saved to a form; a second agent interprets the output and decides the action, which the automation triggers. The result: the platform detects the trend and acts before anyone looks at the report. Through the official ChatGPT plugin you can also ask for a trend analysis over all the data and create a task so it repeats.

Every provider available on OpenRouter is natively available on Olie: you choose the default model of each agent, or let the automatic selection profile decide (default, economy or performance). You can use one model on one stage and another on another, and test a different model without changing the configuration.

The agent is the base: model, tools, system instruction and contexts. The assistant is the agent placed on a stage, with a specific prompt, the moment it runs and where its output goes. One agent can have several assistants, one on each stage, each asking for something different.

It is chained agents: the first qualifies the lead and records the score, the second drafts the proposal from the score, the third classifies the risk. Each runs on its own stage, and the output of one becomes the input of the next. If one fails, the failure route takes the project to a human review stage.

Operations inside the platform (reading and writing projects, forms, stages) and internet search. To call a third-party API, trigger a webhook or send through an external channel, the agent triggers an automation. You limit what each agent can do through its permissions.

In the credit statement: each execution shows the model used, the latency, the tokens, the cost in credits and the input and output texts. That is how you find out which agent is expensive and why. Short instructions and the economy profile reduce consumption.

New executions are blocked until you top up; executions in progress are never interrupted. Automatic top-up is optional: you set the minimum balance and how much to top up. There is never a surprise AI bill, because consumption is prepaid.

Yes. The customer's message becomes project content; the stage assistant reads it, replies through the content automation and the answer goes out from the same number. When the conversation needs a human, the rule moves the project to that person's stage. Everything is recorded in the project, not on anyone's phone.

It is the agent that builds the project summary from everything already in it: forms, conversation, attachments and history. It helps whoever joins the project halfway and the handoff between departments: finance receives the order with the briefing ready.

It can. The assistant's output can be saved to a form field, and the automation that reacts to the field change moves the project. That is how a lead's qualification becomes a score on the form and a high-scoring lead goes straight to Proposal.

Yes. With the official ChatGPT plugin you connect your Olie account and ask in natural language: "how is project 412 doing?", "which orders have been stuck for more than 2 days?", "build a sales trend from this month's forms". You can ask for analyses of forms, logs and data and even create a recurring task for it.

MCP is the protocol that connects AI assistants to systems. The Olie MCP server exposes projects, pipelines, forms and metrics; connected to Claude or ChatGPT, the assistant queries and operates the platform through conversation. The owner asks for the status, requests a custom view or creates a project without opening the screen.

Yes, Olie has a native integration with WhatsApp Business, with three connection options: the official WhatsApp Business API, a connection through an intermediary service and a connection hosted by the platform itself. Several numbers, each tied to its own pipeline; the official connection records the ad source when the conversation starts from a campaign.

Yes: a REST API with about 200 endpoints, documented at docs.olie.ai, with 160 requests per minute per user or application (a limit that can be raised per company). Projects, pipelines, stages, forms, customers, contacts, metrics and movements are available.

Through automation: any trigger (entering the stage, a field change, a date, an idle project) can have a webhook to your system as its action, with the project's data. The response body comes back to the automation, so you can save the result to a field.

Yes, through the open API and webhooks: Zapier, Make or n8n receive the event from Olie or create and move projects in it. It is the way to connect systems that don't have a direct connector yet.

Through the API or a webhook: the order closed in the pipeline triggers a webhook that creates the order in the ERP; the response saves the order number in the project. For reading, you often don't even need to integrate: an assistant connected to both systems via MCP answers cross-system questions, as long as you use the project code as a common reference.

Yes. The public form has a link and can be embedded in the website; whoever fills it in doesn't need an account, and the answer creates the project in the stage you chose. It works for quote requests, pre-consultations, sign-ups and opening tickets.

Yes, import .xlsx files of up to 50 MB to create projects and form responses in bulk, and export any form to a spreadsheet. It is the fastest way to migrate the tracking spreadsheet the team uses today.

When two systems need to talk, the Olie project becomes the meeting point: system A creates the project through the API, the pipeline rule decides the next step, and the automation calls system B. The decision is visible on the board instead of hidden in a script.

Email is an outbound channel through automation: the stage sends the email with the project's data to the customer, the supplier or the manager. Internal email within the team goes away, because the conversation happens inside the project, with mentions.

Through the API and MCP. A software customer uses the project number as the branch name; through MCP, the owner asks the assistant "how is project 412 doing?" and gets the answer from Olie and Git together, with no specific connector.

You can. Part of the team can work without opening the platform: they get the mention on WhatsApp, reply, approve with "ok" and the project moves forward. Olie runs as the engine of the flow behind the scenes, and the manager follows the board in real time.

The order reaches the approval stage; the automation sends the message with the details to the approver; they reply on WhatsApp and the reply moves the project. You can have two layers: below R$ 5 thousand the manager approves, above it goes to the board. Who approved, when and what stays in the history.

Yes, with the "create hierarchy" action: the parent project creates children per department (production, finance, logistics), each in its own pipeline, and the parent only completes when the children complete. A conditional form decides which children are created.

Yes. The project links to several pipelines at the same time, with the same code and the whole history. Sales closes, finance bills and production delivers the same project, each department in its own pipeline, without retyping anything.

With the idle project trigger: stuck for 1 day in the stage, the automation mentions the owner; 3 days, it notifies the manager on WhatsApp; 5 days, it moves the project to an escalation stage. The check runs every minute. Nobody needs to remember to chase.

Yes, by draw or by rule: the new lead gets an owner drawn from the team, or goes to the salesperson for the region set in the form. The distribution records who received it and when, and the individual goal keeps track.

Yes. Migration mode brings in the history from another tool in minutes, and the spreadsheet import creates the projects in bulk. You pick a ready-made template on first access and adjust the stages to your process.

Through the timesheet on the stage: the timer records the time, and the user's hourly rate turns it into cost. You find out that paying a bank slip costs 12 minutes of finance times the hourly rate, at every stage it went through, and the margin per project shows up.

Each stage has a weight: proposal sent is worth 40%, contract awaiting signature is worth 90%. The month's forecast adds up the values weighted by the stage each project is in. The postponement counter shows how many times the date was pushed back.

Yes. Each role sees only what it can request: reimbursement, purchase, vacation, access. The request is born in the right department's pipeline, with required fields and a deadline. Internal email to ask for things goes away.

Yes, with conditional fields: choose "company" and the tax ID and legal name show up; choose "urgent" and a required justification shows up. The form opens up according to the answer, and an empty required field blocks the project from moving forward.

Yes. The stage can have a required video: the person watches it before moving the project for the first time. New employees learn on their first real order, without anyone stopping to teach, and the stage's help text is always at hand.

Yes. A single pipeline for every unit, with segmented visibility: each unit sees its own projects, headquarters sees everything and compares each unit's average response time. For a new unit, you clone the pipeline or use the same one.

Yes. The WhatsApp message becomes project content, with the date, time and who replied. When the salesperson leaves, the history stays. A customer who calls to ask about their order can be served by anyone with access to the project.

Only those whose role has that permission (Managerial or Partner type). Stage restrictions apply to everyone, including the admin: if the rule says the contract doesn't go to finance without a signature, nobody moves it without a signature.

Each project has an edit log and a movement log: who changed the proposal amount, who moved it to Approved, when. AI and automation executions are recorded too. "Who did this?" is answered in one click.

You build the role with the permissions you want; the type (Operational, Managerial or Partner) is calculated by the system from the checked permissions, and the type is what defines the license. Nobody picks the type by hand, and inactive users don't take a license.

Yes. Besides the role, there is pipeline access: the person only sees the pipelines assigned to them. When the role and pipeline access disagree, the pipeline's governance prevails. The field team doesn't see revenue.

Servers on AWS in São Paulo, Cloudflare protection and daily backups kept for 30 days. AI processing happens in the United States and the managed WhatsApp runs in Europe. The public status page is status.olie.ai.

Yes. Each account has its own subdomain (yourcompany.olie.ai), its own credit balance and its own users. The same user can be part of several companies, always by invitation. Every user also has a Personal company of their own.

Yes, optional two-factor authentication per user, plus a corporate alias for login. Login is limited to 8 attempts per minute per user or IP address.

Yes. Any form exports to a spreadsheet, and the Partner role has the full export. If the subscription ends, the account stays in read-only mode for about 30 days so you can export what you need.

The agent acts with the permissions you give it, like any user: it can read one pipeline and not another, write to one field and not another. Credit permissions are separate: one to see the balance, another to buy.

It is, in the execution itself: pipelines, stages, rules, forms and history are the living process. There is no outdated manual on the side; what is configured is what runs. That counts for audits, for selling the company and for training newcomers.

8 triggers, 4 conditions and 23 actions. Triggers include entering and leaving a stage, a field change, a date, an idle project and project creation. Actions include moving, mentioning, creating a hierarchy, sending email, WhatsApp, webhook, writing a field and running an agent.

Yes, up to 5 levels of chaining. The sixth is stopped and the execution log shows why. The classic symptom of the limit is an automation that works when run manually and fails when triggered by another.

Expressions to calculate values, build text and query data inside the automation: add up fields, build the project name from the customer and the date, look up related projects. Limits: 1,000 iterations per loop, 50 queries per expression, 5 seconds of processing.

Every minute the platform checks which projects have been stuck in the stage longer than the limit, and triggers the automation. It is the basis of the deadline that chases on its own and of periodic routines, such as running an analysis agent every 10 minutes.

Yes: an order above R$ 5 thousand goes to the board for approval; below that, to the manager. An agent score above 80 goes straight to Proposal. The condition reads the field and the action decides the path.

It does. If the automation tries to move a project to a stage whose restriction doesn't allow it, the move fails and the execution log shows which restriction refused it. The rule applies to people, automations and the API.

Yes. Rule-fed goals track progress, and a goal that reaches (or misses) the target triggers an automation: notify the manager, open an action plan project, mention the team.

Through the execution log: each run keeps the trigger, the conditions evaluated, the actions executed and the error, if any. About 100 records are kept per automation, with the 20 most recent in the list.

Stages per pipeline: unlimited. Children per project: unlimited. The number of pipelines, restrictions, assistants and active automations counts against the plan's ceiling, visible under Usage and limits in your account.

Four: root, child, grandchild and great-grandchild. The automation action that creates a hierarchy creates three levels from the existing project. A residential project (parent) with architectural, structural and electrical (children) and their revisions (grandchildren) fits.

About 95 MB per form attachment, with a closed list of accepted extensions. Import spreadsheet: 50 MB, .xlsx only.

160 requests per minute per user or per application, which can be raised per company based on the subdomain. MCP: 160 per minute per user. Inbound channel webhook: 120 per minute. Public form: 60 per minute.

Yes, always by invitation, and each company has its own subdomain. On top of that, every user has an exclusive Personal company, with every feature, which only doesn't allow adding other users.

It can: a solar energy customer runs 23 franchises and 19 thousand customers in a single pipeline, with visibility segmented by unit and an automatic WhatsApp message sequence. What scales is the rule, not the number of managers chasing.

No. Deactivating someone who left frees the seat right away, with no downgrade. The role type is recalculated when permissions change.

No. Olie works in the phone's browser, and part of the operation can be done over WhatsApp alone, without opening the platform.

It is a complete copy of an environment's configuration: departments, pipelines, stages, tags, restrictions, automations and agents. It doesn't carry projects, customers, contacts or users. You deploy it and adjust it your way.

No. Right after signing up you pick a template for your type of operation (operational CRM, omnichannel support, service desk, internal operations, project management, onboarding, compliance) and you already have pipelines, stages, forms and automations set up.

The showcase of templates created by Olie and by the community, ready to deploy to your account. Each deployment is an independent copy; one purchase is valid for one company. The Marketplace is coming soon to the website.

Yes, subject to review by Olie. You build it in a builder company (an isolated environment just for that), generate the snapshot, fill in the name, description, categories and images, and submit it. The creator sets the price.

Yes, with its stages, rules and forms. It is the way to open a new unit or a second team with the same process.

1) Create the account and pick the template for your type of operation. 2) Adjust stage names, owners and the deadline that chases on its own. 3) Turn the spreadsheet you use today into a required form on the right stage. The first order already runs by the rule.

No. Each deployment is a copy of the snapshot at the time of purchase. Publishing a new snapshot doesn't change the accounts that already use the template; this protects your environment from changes you didn't ask for.

It brings the history from another tool into Olie in about 5 minutes, creating the projects in the equivalent stages. Together with the spreadsheet import and the ready-made templates, it saves you from starting from scratch.

Per person and per role: Operational R$ 35, Managerial R$ 119 and Partner R$ 199 per user/month, with no tiered plans and no locked features. AI credits are prepaid, separately. Over 100 users, talk to sales.

No: 30 days with full access, no credit card. At the end, you either buy licenses or the account goes into read-only mode for about 30 days so you can export. Reactivation is automatic when you subscribe.

Not on the monthly plan: cancel whenever you want and access lasts until the end of the paid cycle. On the annual plan, the 30% discount applies to the first subscription; renewal is at full price, negotiable by volume.

Yes. There are no feature tiers: what changes from person to person are the role's permissions. What the plan limits is quantity (pipelines, restrictions, assistants, active automations), visible under Usage and limits.

Adding one generates an invoice prorated to the remaining days and is released upon payment. Removing one takes effect at the next renewal, with no prorated refund. Changing type means adding one and removing another.

Prepaid credits, separate from the license, with a statement per execution. You can set automatic top-up if you want. When the balance runs out, new executions stop until you top up; there is never a surprise bill.

No. The ready-made template on first access, the three-step guide and the documentation at docs.olie.ai cover the start. If the operation has several departments or an ERP integration, a 30-minute meeting with the team builds your flow on screen.

Whoever refers up to 3 companies that subscribe to the annual plan earns AI credit worth one Partner user (R$ 199) per referral. The Affiliates program, with recurring commission, is coming soon.

A process designed as a sequence of stages: sales, support, purchase approval, production. It has a name, a prefix (the code of the projects created in it), an optional business area and the project term (lead, order, ticket).

The unit of work and of rules. It holds the form, restrictions, checklist, training video, AI assistants, automations and time metrics. Moving the project from stage to stage is what makes the process move.

The unit that moves through the pipeline. Each pipeline chooses what to call it: lead, ticket, order, work order, case, request. A project can be in several pipelines at the same time, with the same code and the same history.

The grouping of pipelines: Sales, Finance, HR, Operations. It is optional and helps organize the board and permissions. A project crosses areas by passing from pipeline to pipeline.

A tag marks the project and can call in another department (the "legal" tag takes the project to the legal pipeline). The status is per pipeline and tells the situation in that process (in progress, completed, canceled); a final status ends the passage.

A restriction is the entry and exit rule (it only enters Billing with a signed contract). A lock prevents any movement in the stage until it is released. The check when moving follows a fixed order, and the first rule that refuses shows up in the message.

The list of items that need to be checked for the project to leave the stage. Together with the form's required field, it is what ensures nothing moves forward incomplete.

The documentation pages and subpages inside the project, with version history. It is for minutes, specifications and notes that need to stay with the work, not in a loose document.

The list of what is with you: mentions, assigned projects and requests, with a state (unread, replied, completed). The person opens Olie and sees their own queue, sorted by priority, without what isn't theirs.

Model Context Protocol: the standard that connects AI assistants (Claude, ChatGPT) to systems. Olie has an MCP server: the assistant queries projects, pipelines, forms and metrics and runs operations through conversation, with a limit of 160 requests per minute per user.