Management control: how to combine Excel, Power BI dashboards and AI?

"Why did margin drop in the Western region this quarter?"

A CEO asks this in thirty seconds. Answering it — with proper analysis and clean formatting — takes several hours. Management control teams know this gap well: between the question and the figure, there's a connection to a system, an export, VLOOKUPs and SUMIFs, a file to reconcile, and a polished slide or email to explain the results.

Faced with this kind of request, two reflexes come naturally:

• dig into the Power BI dashboard (or Tableau, or similar) to see whether the answer is already there;
• reopen an analysis Excel and rebuild the calculation.

When all you'd really like is to ask the question in a chatbot and get the inputs you need for the analysis. That's what Ask This Guy's Talk to Data solution enables, for example.

The three tools remain complementary, though. This article offers a simple framework to know which one to use, and when.

Three families of tools, three uses

Before choosing, you need to see clearly what each tool does best — and where it falls short.

Excel: the Swiss Army knife of analysis

Excel is unbeatable for modelling: building a budget, simulating a scenario, presenting it with maximum transparency on assumptions and inputs. That's why it's everywhere in finance.

But as soon as it's used for recurring data analysis, the limits show up fast:

• no automated mechanism to import up-to-date data;
• every new question requires new processing and formulas;
• formulas pile up, become fragile, and one misaligned cell throws everything off;
• the result is hard to reproduce: six months later, nobody knows how the file was built (because it's rarely documented);
• the analysis depends on whoever masters the workbook. The day they're away, the information is too (same root cause).

Excel remains essential and will for a long time — and it even has its own e-sport.

Dashboards (Power BI, for example): tracking KPIs over time

A dashboard is built for one precise thing: tracking indicators over time. Revenue, margin, cash, receivables. You see the trend, in a stable, shared visual format. That's exactly what regular financial reporting needs.

Its main advantage over Excel is that it updates automatically, and offers a more intuitive, guided interface for its users.

But as soon as a question falls outside the grain modelled by the report developer ("rank sales reps by order loss this quarter vs. Y-1"), you have to go back to them for an answer — maybe right away, maybe in a week.

Conversation with AI: querying your data in natural language

This is the newest tool. The idea: query your data in natural language, the way you'd ask a colleague, and get a figure, a chart, even an export, right away.

It replaces neither Excel nor dashboards. It fills the gap between the two: one-off, unplanned questions, the ones that have no dedicated slot in a report and that, today, end up in an Excel or in IT's queue.

The real question: "track" or "understand"?

It all comes down to a simple distinction.

A dashboard answers the "what?" question: what is revenue, what is margin, how is cash evolving. It shows what's moving.

Conversation answers the "why?" question: why did margin drop, which customers slipped, what combination of factors explains the gap. It helps you understand.

The dashboard tracks your KPIs over time; the conversation explains the gaps and explores the drivers.

Recurring financial reporting clearly belongs to the dashboard world. One-off financial analysis questions are still handled in Excel today. But they can already be handled more efficiently with AI.

A simple decision framework

Rather than pitting the tools against each other, ask yourself one question: is my need recurring and structured, or one-off and exploratory?

The same question, three paths

Let's take a concrete case. Atelier Norca, a fictional industrial SME, sees its margin fall in the Western region in Q3. The management controller has to understand why. Here are the three possible paths.

Excel path. They export sales from the ERP, the customer base from the CRM, paste it all into a workbook, build a model and reconstruct the impact of the different factors. Thirty-odd minutes later, they have a lead — provided they didn't pick the wrong tab or period.

Dashboard path. They open the dashboard. They see the drop, clearly, on the regional curve. But the report is built by region, not by customer or product. To dig deeper, they have to request a new view from the report designer. The answer will wait.

Conversation path. They ask the question in natural language: "Western region margin in Q3 vs Q2, explore each factor that could explain the gap." They get the figure, a chart, the Excel/CSV export if they want to rework the data, and a summary email. In under a minute they identify the key causes.

"What about security?"

It's the first legitimate objection from a finance department and an IT team: letting an AI query the ERP or the SQL database, is that wise?

The right approach rests on three safeguards:

• Read-only. The assistant queries; it never modifies the data.
• Permissions. The AI only has access to very specific views, useful for its users' queries.
• No copying. Data stays where it is; the AI only handles query generation and interpretation of the results.

Properly framed, conversation doesn't widen the risk surface: it simply makes accessible data the user was already entitled to.

How Talk to Data brings the "conversation" to life

That's precisely the role of Talk to Data, Ask This Guy's solution. It connects to your real sources — SQL, ERP and CRM such as Sage or Salesforce, Excel, MCP, documents — and answers your questions in natural language, with your business context.

Concretely, it brings what's missing between Excel and the dashboard:

• business-context understanding (your terms, your rules, your reference data), thanks to configuration and RAG over your documents;
• chart generation and Excel/CSV export to rework the results;
• respect for permissions and read-only operation, under IT control.

The goal isn't to replace your tools, but to make your teams — management control, finance, operations — autonomous on day-to-day questions. For detailed use cases, see the Talk to Data page, and our article on accessing enterprise data with AI.

Frequently asked questions

Can you do financial reporting with AI?

Yes, with a useful nuance. For recurring, standardised financial reporting, the dashboard remains the reference tool. Conversational AI shines at the next step: exploring a gap, cross-referencing sources, answering an unplanned question without rebuilding a report. The two complement each other more than they replace each other.

Which AI for a CFO?

A useful AI for a CFO isn't a general-purpose chatbot: it's an assistant connected to your real financial sources (ERP, CRM, SQL, Excel), that understands your business context, cites its sources and respects permissions. The priority isn't model performance, but the reliability of data access and rights control.

Who should use AI to query their data?

Far beyond management control. Operational teams who need quick information to move forward (sales, supply chain, support, business unit leaders) can ask questions without going through IT. Top management benefits too: more transparent, direct access to the real numbers, without waiting for an export or an intermediate report. AI doesn't replace domain experts: it makes data accessible to everyone who needs it to decide or act.

Should you abandon Excel?

No. Excel remains excellent for modelling, simulating and building. The idea isn't to abandon it, but to stop making it do what it does badly: answering, by hand and repeatedly, one-off questions. For that, querying your data in natural language saves considerable time — and you can always export the result to Excel for what comes next.

---

Management control doesn't need to pick a side between Excel, dashboards and AI. It needs to put each tool in its place — and reclaim the time lost on questions that deserved an immediate answer.

Book a demo to see Talk to Data answer, live, a real question from your management control team.