--- title: "Reporting Automation: First Report in 14 Days | JustAutomate" description: "One place for the numbers, metric definitions written down once, and reports that send themselves. First report in production 14 days after the workshop." url: "https://justautomate.ai/en/services/reporting-automation/" lang: en --- Process automation # Reporting automation one set of numbers for the whole company Data lands in one layer, each metric has one written definition, and the finished report goes to the right people at a set hour. The board, the department and the client see their slice of the same numbers. A question the report does not cover goes to an AI agent in a plain sentence, instead of turning into another file request. Book a free consultation [See what the workshop looks like](https://justautomate.ai/en/services/process-workshop/) - 14 dayshow long it usually takes from the workshop to the first report that sends itself from your data - 2 to 4 weeksfor the whole reporting layer: sources, metric definitions, schedule, distribution and alerts - 5 to 10 secondshow long the AI agent takes to answer a question the report does not cover, without opening a file The problem ## Why do two reports show two different numbers? At the Monday meeting three departments quote three different lead counts. Marketing counts every enquiry, sales counts only the ones somebody spoke to, and finance does not count them at all. The first half hour goes on agreeing whose number is real, and the decision waits another week, because nobody wants to take it on data they do not trust. The cause is rarely the tool. Every report grew on its own: a different file, a different formula, a different time window, often a different person who has since left. The definition of revenue, of a lead, of margin is written down nowhere, so each report carries its own copy and nobody can see where they drift apart. Only one report in the company has a real deadline. The annual financial statement has a filing date, an auditor and an owner, and nobody misses it. The reports you actually take decisions on during the year have no deadline, no owner and no shared definition. They get made when somebody finds a Friday afternoon for them. - The board meeting opens with agreeing on numbers instead of taking a decision - The same report is assembled by hand by several people, each in their own file - How margin is calculated leaves the company together with one person - Nobody notices that a source stopped loading on Tuesday, so the number is quietly low Why now ## What changed, so a reporting layer takes weeks rather than a quarter? Every system where your numbers are born now has a read interface: the CRM, the invoicing system, the shop, the ad platforms, the task tool, the spreadsheet holding the schedule. Nobody has to export by hand, because the same query can pull the data every hour or every night. A data warehouse stopped being a quarter-long project. BigQuery is a service billed for the data a query processes, with no server to keep alive, and its [pricing](https://cloud.google.com/bigquery/pricing) includes a free monthly tier: 1 TiB of query processing and 10 GiB of storage. Management reporting at a mid-sized company sits inside that order of magnitude. Collection runs through [n8n](https://docs.n8n.io/) or Make, presentation through Looker Studio or Google Sheets. This page sits a level above any single report. We describe the individual ones separately: [the sales report](https://justautomate.ai/en/automations/automatic-sales-report/), [the weekly report for the board](https://justautomate.ai/en/automations/automatic-weekly-report/), [the financial report](https://justautomate.ai/en/automations/automatic-financial-report/), [the campaign report](https://justautomate.ai/en/automations/automatic-campaign-report/) and [client reports](https://justautomate.ai/en/automations/automatic-client-reports/). The same thing sits under all of them: a [data pipeline](https://justautomate.ai/en/automations/automatic-data-pipeline/), one metric definition and a schedule. That layer is what this page is about, because without it every new report is another file with its own formula. The other change is what happens to a question the report does not answer. It used to end as an order for a new spreadsheet. BOND, our company AI agent, answers from the same data layer in 5 to 10 seconds, in a plain sentence. The report covers the questions that come back every week, the agent takes the tail of one-off ones. What changes ## What does a company get when the reports assemble themselves? - ### One number for the whole company The definition of revenue, of a lead, of margin and of conversion is written once, in the data layer, not in a fifth spreadsheet. Every report and every dashboard counts from the same formula. A metric catalogue written in the workshop, with the period, the denominator and the moment of recognition - ### The report arrives on its own, at a set hour The schedule collects data overnight and delivers the finished set where the team already works. Nobody logs into five panels to put one file together. Email at 07:00, a Slack message, a spreadsheet refreshed at 23:59 or a Looker Studio dashboard - ### Who sees what, settled once The board gets the whole picture, a manager gets their department, a client gets only their slice. Access levels sit in the data layer, so nobody can send out somebody else's numbers by accident. Three access levels: board, department, external recipient - ### You know when the data did not arrive The worst report is the one that looks normal and has a quiet hole in it. When a source goes silent for longer than the agreed threshold, the alert arrives before the report, not after the decision. An alert when a source has not fed the data layer for more than 25 hours - ### Ask about a number in a plain sentence How much revenue year to date. Which channel has the highest cost per lead today. How many invoices are past due. The AI agent answers from the same data layer, in English or in Polish. An answer in 5 to 10 seconds, with no CRM login and no new report - ### Fewer reports, not more We start from the question of which decision a report supports. The ones that support none we propose switching off rather than automating, and we say that in the workshop. Every report has an owner and a decision it leads to; after 90 days we check who opens it Proof ## Do you collect your own numbers this way? Yes. Since March 2026 the data layer and the reports inside our founders' agency have been fed by an AI agent, the same one behind the figures on our AI for business page. Below is the scale of that work, together with what the number does NOT prove. Built for ourselves. These are not client deployments but implementations inside the founders' own companies. We say so plainly, because proof you cannot verify is not proof. - ![JustIdea Agency](https://justautomate.ai/klienci/justidea.png)![JustBot](https://justautomate.ai/agenci/justbot-80.webp)JustBottheir agent JustIdea Agency, a performance marketing agency of around 30 people 23 people from the company work with the agent 38 systems wired in permanently 112 processes run without a human 475 thousand lines of the agent’s code The agent is wired into Google Ads, Meta, the CRM, the task system, the SEO tool, Analytics, Search Console and the data warehouse. 23 people use it, from salespeople to the board. 8 of 38 systems it works in - Google Ads - Meta Ads - HubSpot - ActiveCollab - Ahrefs - GA4 - Search Console - BigQuery Before A person collected the numbers for reports: an export from the ad panel, an export from the CRM, last week's spreadsheet, margin recalculated by hand. Each department counted its own way, so the meeting had to start by agreeing which version applied. After Data flows into the warehouse on its own, through one formula, and the team asks about numbers in a plain sentence instead of ordering a spreadsheet. The person comes in at the comment and the decision. 3,506 data warehouse load operations in 30 days, with no person involved And in your company ## What if our data is a mess? Four sentences we hear most often on the first call, and what we say to them. - We already have a dashboard, nobody looks at it. A dashboard is the last step, not the first. There are usually two reasons nobody opens it: it does not answer the question somebody actually has, or it lives somewhere that needs a separate login. We start from the decision the report is meant to support, and deliver it where the team already works. - We have no warehouse and we do not want another system. With two sources and small data, Google Sheets fed by n8n or Make is enough, and that is what we will tell you. A warehouse starts to make sense when a report joins three sources or history has to be counted backwards. BigQuery is not a system to maintain, it is a service billed for queries and paid to Google directly. - Our data is in a real mess, we have to tidy it up first. Mess is the rule, not the exception, and it only becomes visible when two sources stand side by side. The workshop ends with a list of discrepancies: where a client name is written three ways, where a date is missing, which metrics are counted differently. The pilot runs on real data, not on a cleaned sample. - The board will ask about something the report does not have anyway. It will, and that does not go away. This is why we keep a data layer under the reports rather than finished files: one source can answer a question nobody anticipated. That tail goes to the AI agent, which answers in 5 to 10 seconds, instead of becoming another permanent report. Data and warehouse BigQuery, Google Sheets, SQL databases, GA4 CRM and sales HubSpot, Pipedrive, Salesforce, BaseLinker, Allegro, Shopify, WooCommerce, PrestaShop Accounting and finance Xero, sevDesk, Fakturownia, Comarch Optima Presentation Looker Studio, Google Sheets, a dedicated dashboard, PDF export Distribution and teamwork Slack, email, Google Workspace, Microsoft 365, Notion, ClickUp, Jira, Asana Automation layer n8n, Make, Zapier, custom API integrations ## The metrics that get argued about, and the definition settled once Seven places where reports most often drift apart. This is a difference of definition, not of tooling, which is why no dashboard fixes it on its own. | Metric | Why two reports disagree | What we settle in the workshop | | ------------------- | ------------------------------------------------------------------------- | --------------------------------------------------------------------- | | Revenue | the CRM counts from the signature date, accounting from the invoice date | the moment of recognition, net or gross, the rate used for currencies | | Lead | marketing counts every enquiry, sales only the ones with a conversation | what a lead is, when it stops being one, how duplicates are counted | | Margin | some subtract ad cost, others labour cost, some subtract nothing | which costs go in and at what level: product, order, client | | Conversion | GA4 counts sessions, the CRM counts opportunities, the shop counts orders | what sits in the denominator and over what period | | Active client | the shop counts an order, the CRM counts an open contract | the time window and the event that counts as activity | | Cost of acquisition | the ad panel knows spend, finance knows labour cost and commission | whether team time, commission and tool cost go in | | Lead time | one person counts from the order, another from picking | the start and stop points, working days or calendar days | These rows are a starting point. In most companies the first rollout settles three to four metrics, because that is enough for the sales report and the management report to show the same number. How we do it ## Five steps from five versions of a file to one set of numbers Every step ends with something you can judge before you decide on the next one. 1. ### Workshop We write down every report that gets made today: who assembles it, how long that takes, who reads it and what decision comes out of it. Reports with no decision behind them go on a separate list. Along the way it comes out which metrics are counted two ways inside one company. 2. ### Metric catalogue and data layer design Every metric definition goes on paper together with its period and its denominator. We agree the sources, the refresh frequency, where the data sits (BigQuery, a SQL database or Google Sheets) and the access levels: what the board sees, what a department sees, what an external recipient sees. 3. ### Pilot on real data The first report, usually the one that hurts most, runs alongside the existing one for two weeks. We compare number by number and explain every discrepancy, rather than assuming the new report is the one that is right. 4. ### Production and distribution Schedule, distribution by email, to Slack, to a spreadsheet or to a dashboard, alerts when data is missing and error monitoring. We teach your person to add a column and change recipients on their own, without raising a ticket with us. 5. ### Measurement after 30 and 90 days We check who actually opens the reports, whether the definitions held, and what the board asks about outside the report. Reports nobody reads get switched off. Questions that keep coming back get added to the report or handed to the AI agent. Before and after ## How the board's Monday changes ### Before - Somebody assembles the report by hand on Friday, on Monday you discuss data that is three days old - The meeting opens with the question of whose number is real - A request for a new cut of the data turns into several hours of work - Nobody knows that one source stopped loading on Tuesday ### After - The report is in the inbox at 07:00, built from last night's data - Each metric has one definition, written down and visible next to the report - A new cut of the data is a question to the AI agent and an answer in 5 to 10 seconds - The alert about a silent source arrives before the report, not after the decision The first report in production usually shows up 14 days after the workshop. Whether the definitions hold and whether anyone opens those reports we check after 30 and after 90 days. For the person signing this off ## What the board and the CFO get out of it You stop opening meetings by agreeing on numbers. The report arrives at a set hour, each metric has one definition, and when the data did not load you know before the meeting rather than during it. That changes the shape of the conversation: it starts from a conclusion, not from an argument about sources. The second thing matters when you weigh risk. How margin is calculated stops living in one person's head and in one person's file. The metric catalogue is written down and versioned, so a holiday or a resignation does not stop reporting for a month. - Metric definitions written down and versioned, independent of any one person - An alert about missing data instead of a report with a quiet hole in it - The pilot runs alongside the existing report, so the old way works until testing ends - After 30 and 90 days you know which reports get opened and which can be switched off What it costs ## The workshop has a fixed price, the rollout is quoted after it The process workshop costs **EUR 3,500 net per week**. With reporting it usually runs one to two weeks, because every report made today has to be walked through and the metric definitions agreed across departments. We quote the rollout after the workshop, once we know how many sources are in play and what state the data is in. Tool cost goes to the vendor directly: Google for BigQuery at list price, n8n or Make for scenarios, or nothing at all when n8n runs on your own infrastructure. The AI model cost is a vendor cost as well, in the range of EUR 0.005 to 0.07 per query depending on the model. No margin from us and no reselling of invoices. - ### Process workshop EUR 3,500 net per week, with reporting usually one to two weeks A self-contained piece of work. It ends with a metric catalogue and a list of reports with owners, which you can execute without us. - An inventory of every report made today, with the time it takes and its owner - A metric catalogue with the definition, the period and the denominator - A list of data discrepancies and the sources that have to be fixed first - A recommendation for the data layer: Google Sheets, BigQuery or a SQL database - The indicators to measure after 30 and 90 days Start with the workshop - Most popular ### Workshop and reporting layer rollout Quoted after the workshop scope follows from the number of sources and the state of the data The option most companies choose. Working reports 2 to 4 weeks after the workshop, the first one usually after 14 days. - Everything in the workshop - Collection through n8n, Make or an API into BigQuery, a SQL database or a spreadsheet - Scheduled reports, distribution by email, to Slack and to a dashboard - Access levels and alerts when data is missing - Pilot alongside the existing report, plus training for your person Talk about a rollout - ### Reporting and an AI agent Quoted after the workshop for companies where questions about numbers come back every day Everything in the rollout plus BOND, the company AI agent that answers questions about numbers from the same data layer. - Everything in the reporting layer rollout - AI agent wired into the warehouse, the CRM and the invoicing system - An answer to a question about a number in 5 to 10 seconds, in English and in Polish - Alerts on deviations from plan and on sources that have gone silent - Agent rollout usually takes around 2 months, running in parallel with the reports Ask about the agent In the workshop we count how many of your reports lead to a decision. If half of them lead to none, we say so and propose switching them off rather than automating them. The workshop ends with a document that stands without us: the metric catalogue and the list of reports with owners. ## Frequently asked questions What is reporting automation? It is building a layer from which reports assemble themselves: data flows in from your systems on a schedule, each metric has one written definition, and the finished set goes to agreed recipients by email, to Slack, to a spreadsheet or to a dashboard. The person adds a comment and takes the decision, instead of collecting numbers. How long does a reporting rollout take? One to two weeks of workshop and 2 to 4 weeks of rollout. The first report usually runs 14 days after the workshop, because we start with the one that hurts most and add the rest on top of a data layer that already stands. The timeline stretches when a source has no read interface and the data has to be pulled another way. Do we need a data warehouse to have automatic reports? Not always. With two sources and little history, Google Sheets fed by n8n or Make is enough, and that is what we will tell you in the workshop. A warehouse starts to make sense from three sources, from counting history backwards, or when one metric has to feed several reports at once. That is usually where we put BigQuery, billed for the data a query processes. How is this different from a Looker Studio dashboard? A dashboard is a way of showing numbers, not a source of truth. Wire it straight into five systems and every chart still counts its own way, so the argument about definitions stays. We settle the definitions first and collect the data into one layer, and Looker Studio or a spreadsheet is the last step on top of it. How do we know the numbers in an automatic report are right? From three things. The pilot runs for two weeks alongside the existing report and we compare number by number, explaining every discrepancy. Each metric definition is written down, so you can see exactly what a given column counts. And when a source has not fed the data layer for more than 25 hours, you get an alert before you open a stale report. What if the board asks about something the report does not have? This is why we keep a data layer rather than finished files: one source can answer a question nobody anticipated. One-off questions go to BOND, our company AI agent, which answers from company data in 5 to 10 seconds. A question that comes back every week gets added to the report permanently. Who can change a report when the needs change? Your person, and that is part of the rollout. We teach adding a column, changing recipients, moving the send time and switching off a report nobody reads. We stay with changes in the data sources and with new integrations, not with every cosmetic tweak in a template. Can a report for an external client come from the same data layer? Yes, and that is the point of a single source: the client sees the same revenue and the same cost as your board, only in a narrower slice and in their own template. Access levels sit in the data layer, so nobody can show somebody else's numbers by accident. We describe the details on the page about automatic client reports. ## Let us check which of your reports lead to a decision The free consultation takes 30 minutes. You say which reports get made today, who assembles them and what the board asks about outside them. We show what can be built from what you already have, and what needs the definitions agreed across departments first. You do not have to share any systems at this stage. We run at most two rollouts at a time, so that someone is actually sitting on each of them. [Book a free consultation](https://justautomate.ai/en/contact/) **PS.** If somebody at your last meeting asked where that number came from, the problem is not in the report. A report that sends itself is the easy part. The hard part is getting every department to agree what exactly they are counting, and getting that definition written down outside somebody's head. That is where the workshop starts. ## Related - [Automatic sales report](https://justautomate.ai/en/automations/automatic-sales-report/) - [Automatic weekly report](https://justautomate.ai/en/automations/automatic-weekly-report/) - [Automatic financial report](https://justautomate.ai/en/automations/automatic-financial-report/) - [Automatic data pipeline](https://justautomate.ai/en/automations/automatic-data-pipeline/) - [BOND: business AI agent](https://justautomate.ai/en/services/bond-business-ai-agent/) - [Process workshop](https://justautomate.ai/en/services/process-workshop/) --- **JustAutomate** · contact: contact@justautomate.ai · form: https://justautomate.ai/en/contact/