Webinar: Power BI, Excel - AI & Copilot

Summary

Vektus A/S demonstrated practical ways finance teams can use Excel, Power BI, and AI to reconcile data, generate summaries, and build formulas and measures. The central message was that AI can speed up routine analytical work, but its results should always be checked against the source data.

Tools and workflows demonstrated

• Copilot for Finance in Excel

  • The presenter showed how to add the preview extension through Excel add-ins.
  • Its reconciliation workflow identified invoice and payment tables, suggested matching fields (invoice ID and amount), and grouped transactions into matched, potentially matched, and unmatched categories.
  • Users could review the results, investigate exceptions, and append a summary to the report or save it as a PDF.
  • At the time of the webinar, reconciliation was available in the preview; variance analysis was described as a planned future capability.

• ChatGPT for Excel tasks

  • The presenter supplied a screenshot of city-level revenue data to ChatGPT to generate a region mapping and a revenue-by-region table.
  • The first result did not reconcile to the expected total, illustrating the need to verify AI-generated tables.
  • The presenter then requested an Excel formula to calculate revenue by region. After correcting the assumed column references, ChatGPT produced a formula compatible with Danish Excel conventions, including localized syntax and semicolon separators.

• ChatGPT for Power BI

  • The presenter used AI to expand a date table with fields such as week and year-month.
  • It also helped draft DAX measures for last year’s amount and year-to-date amount.
  • The measures were tested in Power BI against the report’s date selections.

Practical playbook and recommendations

• Use AI to draft reconciliations, formulas, and measures, then validate the output against source data and expected totals.
• Give AI precise context—such as the correct column names, table structure, and intended calculation—when its first answer makes assumptions.
• In reconciliation workflows, prioritize reviewing potential matches and unmatched transactions rather than assuming all suggested matches are correct.
• Maintain a proper date table in Power BI to support time-based analysis such as year-over-year and year-to-date measures.
• The presenter recommended choosing an AI tool and experimenting with it rather than trying to track frequent model rankings. ChatGPT, Gemini, and DeepSeek were mentioned as examples.
• A signed-in ChatGPT account was recommended to retain conversation history.

Metrics and examples

• The reconciliation demo showed about 21 transactions matched directly by invoice ID and amount.
• One example displayed a payment of DKK 84,030.30 as a potential match.
• The regional revenue example used an expected total of 30,000 to check the AI-generated table and formula.
• A Power BI example returned a last-year amount of −291 for a selected period; this was a demonstration value, not a company KPI.
• No company performance targets, growth rates, or financial KPIs were presented.

Other business updates

Vektus promoted upcoming events: AI and Business Central Cloud seminars planned for the end of April, production-focused Business Central seminars at the end of May, and introductory Business Central Cloud courses in May. The company directed attendees to its website, newsletter, and LinkedIn for event information.

Presenters and sources: Ole Hansen, Sales Manager at Vektus A/S; Nikolaj, Vektus BI Department. Tools referenced included Microsoft Excel, Copilot for Finance, Power BI, ChatGPT, Gemini, and DeepSeek.