Skip to content
AI Board

For general managers and BU directors

AI for general managers and business-unit leaders: the P&L is yours, the staff functions aren't

You carry the full P&L at one site or one unit, and everything crosses your desk: margin, cash, the order book, the people. What you don't carry is a staff function of your own, so the analysis you need for Thursday's call with the group waits on one shared controller. An AI grounded in your own site's files takes the waiting out, and stays straight about what it doesn't solve.

The month closes, the variance file lands, and you have a call with head office where you are expected to explain a gap you haven't had time to take apart yourself. The pressure behind that is not imagined: in Deloitte's 2026 Global Human Capital Trends survey, 60% of executives say they regularly use AI to support their decisions, across more than 9,000 business and HR leaders in 89 countries. It is reasonable to assume the people on the other end of that call have the same tools you do.

The standard answer is to wait for the group programme. That answer has a poor record. Gartner predicts that through 2026, organisations will abandon 60% of AI projects unsupported by AI-ready data (a prediction, not a measurement), and RAND found more than 80% of AI projects fail, roughly twice the rate of comparable non-AI IT projects. Meanwhile your numbers are due on the 12th, and the files that hold the answer are already on your machine: the variance workbook, the debtor aging, the order book and capacity sheet.

AI Board reads those files and answers from them, naming the exact document and version behind every figure, so you can check it in seconds rather than trust it blind. It reasons at the altitude your seat actually needs, from a margin bridge down to which four accounts to call before payroll. This page is about the week-to-week use. For the action itself, asking your own files and getting a grounded answer, see chat with your data.

The week you recognize

Explaining the EBITDA gap before head office explains it to you

Revenue held, so it isn't volume. It's margin, and somewhere in Variance_June_2026_final.xlsx is 1.8 points of gross margin lost on mix, with the Coating line carrying most of it and Assembly the rest. On its own, 1.8pt is abstract. Put a euro figure next to it (roughly 120k of EBITDA on a site your size) and it becomes physical, which is exactly how the group will read it. The fear isn't the analysis. It's being ambushed on a call with a number you haven't taken apart yet.

Four customers are sitting on your cash, and there's no treasury here

There is no treasury function at your site, so working capital is yours too. Four accounts are 60+ days overdue, around 40% of AR, with Van Leeuwen alone at EUR 180k and Bektas Metaal behind it. DSO sits at 61 against a 45-day target, and payroll doesn't move. What you need on Monday morning isn't a report on debtor days. It's a call list: name, amount, days, in that order, before the credit controller picks up the phone.

A CEO's job and none of the staff to do it with

You have the accountability of a company director and the support of a shared services line that answers three other sites first. So you do the assembly work yourself, and the cost of that is measurable: McKinsey estimated the average interaction worker spends nearly 20% of the workweek looking for internal information or tracking down colleagues who can help (2012, and file sprawl hasn't improved since). A day a week of finding things is a day a week not spent on the floor. The missing piece isn't another headcount request. It's the staff function you never got.

Live demo

Your personal AI assistant, thinking

How much revenue can I safely commit upward for Q3?

Confirmed order book covers 1.8M and capacity is booked to 82%. Project P-2043 (0.3M) is still unsigned. Commit 1.8M and hold P-2043 back for the review.

Order_Book_Capacity_Q3_2026.xlsxQ3 order book & capacity
Ask AI Board…

What changes

The margin bridge, in lines and euros, before the call

Ask why EBITDA came in under budget and you get margin separated from volume, the specific lines that carry the gap, and the euro impact next to the percentage points, read from the variance workbook on your drive with the file and version named. You still own the conversation with the group. What changes is that you arrive at it having already taken the number apart, instead of hearing it explained back to you.

A call list, and an honest ceiling on what calling fixes

Ask what is choking cash and you get the accounts, the euro amounts and the days, ordered so your credit controller can start at the top. It also tells you what four phone calls realistically move: DSO towards 50, not to your 45-day target, because the rest of that gap is structural (payment terms, invoicing lag, one customer's own approval cycle). A tool that promises the clean jump is a tool you'd stop believing after the first month.

A bandwidth you can commit upward, not a single number

Before a group review, ask what you can safely commit for the quarter and you get the confirmed order book, booked capacity at 82%, and what that occupancy means for quoted lead time, because that is the question the group asks back. Unsigned work is held out and named (P-2043, 0.3M), with both sides shown: what the quarter looks like if it signs, and what falls out if an order slips. You commit a range you can defend, not a point estimate you have to walk back.

Answered on demand

Overtime and hired labour

Where did overtime and hired labour hours land above plan last month, and on which line did they concentrate?

Downtime against plan

Which machines cost me the most unplanned downtime hours this quarter, and what did that do to the output plan?

Purchase price drift

Which supplier price increases have actually landed in cost of goods, and which product groups absorb them?

Open actions from the group review

Which commitments from the last two group reviews are still open, and what is the status of each in my own notes?

Questions, answered

What does AI actually do for a general manager?
It answers the questions your seat already asks every month, from your own files instead of the open internet: where the margin gap sits and which lines carry it, which accounts to chase before payroll, what the order book and booked capacity let you commit upward. You point it at the workbooks you and your controller already maintain, and it reads them and answers, naming the document and version behind each figure. It doesn't run the site and it doesn't make the call to head office. It removes the two days of cross-reading that currently sit between you and the answer.
How is this different from the group BI dashboard or a P&L agent from a BI vendor?
A group dashboard shows what made it into the group's data model, on the group's refresh cycle, in the group's definitions. Half of what you need for the call is not in there: the manual tab your controller keeps, the capacity sheet, the version marked final after five revisions. Vendor P&L agents are the closest functional comparison, and they are real products, but they are platform installs that go through IT and BI procurement and run on the vendor's infrastructure. AI Board reads the files as they actually exist on your machine and answers there. It complements a dashboard rather than replacing one.
Do I need an analyst or an IT project to set this up?
No, and that is the point of the design: there is no data warehouse to build first and no queue to join. It is set up on your own laptop against the folders you already use, and it starts answering from those. Be straight with yourself about the limit, though: it can only be as good as what is in your files. If your line-level cost data is thin, the answer about that line will be thin, and it will say so rather than fill the gap with something plausible.
Does my site's financial data leave my laptop?
No. That question matters here because the alternative is already happening in most companies: [MIT's NANDA research found employees at over 90% of companies regularly use personal AI tools for work while only about 40% of companies have an official AI subscription](https://fortune.com/2025/08/19/shadow-ai-economy-mit-study-genai-divide-llm-chatbots/), which is how a debtor aging list ends up pasted into a personal chatbot. The consequence is priced: [IBM's Cost of a Data Breach 2025 found breaches involving a high level of shadow AI cost on average $670,000 more than the $4.44M global average](https://www.ibm.com/think/insights/data-matters/cost-of-a-data-breach). AI Board is private by design: it runs grounded in the files on your own machine. See [security](/en/security) for how that holds up.

The questions don't change from month to month: the gap, the cash, the number you can safely commit. What changes is whether you have taken them apart before someone else does it for you. AI Board is your personal AI assistant that makes you AI-native: it runs on your own laptop, grounded in your company's data, and gets sharper as your knowledge grows. Private by design, reasoning at CEO, CFO and CTO level, so you ride the AI wave instead of swimming behind it. It will tell you when four phone calls only get you halfway, and that is the reason it belongs on your laptop: a tool that rounds numbers to please you is one you'd walk away from.

Put your own site's files to work

See what a second brain grounded in your variance, aging and order book files answers before the next call with head office, private by design, with the source named every time.

Runs on your own laptop. Your data never leaves it.