Portfolio (prioritization)

Where your implementation stands, and which single next step unlocks the most. Progress doesn't come from a checklist ticked by hand: it comes from counting your own data. If the number says there are three customers, it's because there are three customers saved — and that's why it can't be fooled, neither up nor down.

The request becomes a traceable record in the same channel as "Talk to BMCode" — you follow the answer in My feedback —, and it goes with a snapshot of the implementation attached: current phase, percentage, what is pending, the batches in memory and the steps that got stuck. It's the difference between support that starts with "tell me what happened" and one that starts with the answer.

The accountant comes in as a read-only external user — they check the mirror month and the reports at no seat cost. Without them the mirror month doesn't close, because their closing is what serves as the answer key.

Everything you send goes into a holding area, becomes a line-by-line preview and is only applied when you approve. Reapplying updates instead of duplicating, and undoing reverts the whole batch. That's why it's possible to send the spreadsheet as it is, without tidying anything up first.

When the same step is attempted twice without completing — the same reading undone twice, or the same screen opened twice with nothing saved —, a notice appears offering a specialist, without you asking. This exists because whoever gets stuck in an implementation rarely clicks "help": they close the tab. The offer needs to arrive before that, and the number of attempts is configurable in JORNADA.AJUDA_APOS_TRAVADAS.

Only a step attempted in the last 48 hours counts. A screen opened five times three months ago and never filled in isn't someone stuck right now: it's a matter you've already solved another way or abandoned on purpose, and offering help there would just be noise.

After uploading the old system's spreadsheet, the "Suggest mapping with AI" button reads the column names and three sample rows and proposes the destination for each one. Nothing is saved in this step: the suggestion only fills in the screen's fields, as if you had chosen. The path remains the same — check, generate the preview, apply.

The request made to the model is explicit: with no clear match, return empty. A plausible guess here would become wrong data in the records of someone just starting out — the worst moment to lose trust. When the AI suggests with low confidence, the screen warns and asks for a check; when it invents a field that doesn't exist in Tenka, the suggestion is discarded and that appears in writing.

The plan is built from the company's verticals, the account modules and the country of operation, in the order the modules depend on each other. Each screen appears only once: whatever your vertical adds about a screen goes into the same step. The plan suggests; the account's subscription is still what unlocks the screens. With no vertical chosen, the plan shows the steps that apply to every company.

None. The request goes out the same way, and the attached context says explicitly that the proactive rule had not fired — whoever handles it knows the request came from you, and not from the system.

Because it doesn't exist yet. A link to an empty community would be worse than a declared absence: you'd click, find nobody, and start distrusting the two paths that do work as well.

The column NAMES and up to three sample rows are sent — without seeing values you can't tell "document" from "code". If your company requires masking personal data and the configured AI provider is external, the suggestion is REFUSED and the screen explains: names and documents from your spreadsheet don't leave here. With a local model, nothing leaves the network.

Because the system only marks a step on its own when it can measure it from the data: customers, suppliers, products, contracts, bank accounts, financial categories, cost centers, chart of accounts, transactions, reconciliation, projects, tickets, employees and a second person on the team. The other steps depend on you checking the screen — the plan does not pretend to know.

Implementation playbooks became guides built from the catalog: in the Help Center, under Implementation guides, each vertical has its own, with the steps of the modules it uses, in the order one depends on another. An operation that combines several verticals (the group's textile chain, the software house with HR consulting) has a single guide, without repeating a screen. Strategy, Processes and Tenkai have their own guide. The old playbook address still opens, now with the new guide.