AI document generator for Word, PDF, PowerPoint and Excel

Initial request

The attached CSV contains entirely fictional data for a demonstration. Generate a downloadable report in DOCX and PDF, in Spanish, titled “Informe de actividad: ejemplo ficticio, enero–junio de 2026”. Include the original table, total sales and operating expenses, and the arithmetic difference between them; a comparison of June with May; three questions for the client meeting; and a visible note about missing data. Do not call that difference profit, cash flow or a bank balance: the file does not include taxes, depreciation, inventory, receipts or payments. Do not invent reasons for the change. No more than three pages. Include “Ejemplo con datos ficticios; no es un informe de un cliente real” on the first page.

Follow-up 1

The review found an error in both files: you changed the June row. The CSV I attached contains sales of 21800 and operating expenses of 15700 for 2026-06, but your documents show 23000 and 15800. The PDF has also lost accents and opening question marks. Read the attached CSV actividad-cliente-ficticio-2026.csv again and regenerate the DOCX and PDF from its six exact rows. Do not replace data or reconstruct the table from memory. Before generating the files, check that the rows are: mes,ventas_eur,gastos_operativos_eur 2026-01,18000,14000 2026-02,19000,14200 2026-03,17500,14100 2026-04,21000,14900 2026-05,22500,15300 2026-06,21800,15700 Verify these totals: sales 119800, operating expenses 88200 and arithmetic difference 31600. From May to June, sales fall by 700 and expenses rise by 400; sales minus expenses goes from 7200 to 6100. Do not invent a causal explanation. Keep the full title “Informe de actividad: ejemplo ficticio, enero–junio de 2026”, the three questions for the meeting and the visible note about the missing taxes, depreciation, inventory, receipts and payments. The difference is not profit, cash flow or a bank balance. Include “Ejemplo con datos ficticios; no es un informe de un cliente real” on the first page. No more than three pages. Preserve all accents, ñ, ¿?, and the € symbol in both formats. Write amounts in Spanish format, for example 21.800,00 €. Check both files after generating them and confirm in chat that the table matches the original CSV row by row. Return both new downloadable files. Also replace the previous interactive chart with one using these six exact rows and two bar series, sales and operating expenses, with the axis starting at zero. The June row must be 21800 and 15700. Do not generate an artistic image of the chart; we need a calculated chart with exact values.

Follow-up 2

The amounts in the second version are now correct. Preserve its six rows and totals exactly: sales 119.800,00 €, operating expenses 88.200,00 € and arithmetic difference 31.600,00 €. Two issues still need fixing in the files, not just in the chat response: 1. Both documents have lost the comparison between May and June that I requested. Add the section “Comparación de junio con mayo” with this content: “Las ventas pasan de 22.500,00 € en mayo a 21.800,00 € en junio, una disminución de 700,00 €. Los gastos operativos pasan de 15.300,00 € a 15.700,00 €, un aumento de 400,00 €. La diferencia aritmética entre ventas y gastos operativos pasa de 7.200,00 € a 6.100,00 €, una reducción de 1.100,00 €. El archivo no permite determinar las causas de estas variaciones”. 2. The PDF is still removing accents, ¿ and €. A “standard encoding” that changes the Spanish text does not meet the requirement. Generate the PDF with an embedded Unicode TrueType font, such as DejaVu Sans or Liberation Sans. You can use ReportLab and register that font. Do not transliterate to ASCII, remove diacritics or replace € with EUR to avoid the problem. If you need to locate a font, look for a TrueType font available in the environment or among the fonts of the installed libraries. Regenerate the DOCX and PDF with the same complete content: title “Informe de actividad: ejemplo ficticio, enero–junio de 2026”; opening note “Ejemplo con datos ficticios; no es un informe de un cliente real”; six-month table; totals; the comparison above; the three questions for the meeting; and the note identifying the missing taxes, depreciation, inventory, receipts and payments and explaining that the difference is not profit, cash flow or a bank balance. Keep all amounts in Spanish format and a maximum of three pages. Before delivering the PDF, extract its text and check that it preserves “aritmética”, “Comparación”, “reunión”, “¿A qué”, “¿Cómo”, “próximo”, “tesorería” and “€” exactly. Also check that both files include the comparison and that June retains 21.800,00 € and 15.700,00 €. Return both new downloadable files. Do not generate or modify the chart in this turn.

Initial request

The attached CSV contains entirely fictional data for a demonstration. Generate a downloadable report in DOCX and PDF, in Spanish, titled “Informe de actividad: ejemplo ficticio, enero–junio de 2026”. Include the original table, total sales and operating expenses, and the arithmetic difference between them; a comparison of June with May; three questions for the client meeting; and a visible note about missing data. Do not call that difference profit, cash flow or a bank balance: the file does not include taxes, depreciation, inventory, receipts or payments. Do not invent reasons for the change. No more than three pages. Include “Ejemplo con datos ficticios; no es un informe de un cliente real” on the first page.

Follow-up 1

The review found an error in both files: you changed the June row. The CSV I attached contains sales of 21800 and operating expenses of 15700 for 2026-06, but your documents show 23000 and 15800. The PDF has also lost accents and opening question marks. Read the attached CSV actividad-cliente-ficticio-2026.csv again and regenerate the DOCX and PDF from its six exact rows. Do not replace data or reconstruct the table from memory. Before generating the files, check that the rows are: mes,ventas_eur,gastos_operativos_eur 2026-01,18000,14000 2026-02,19000,14200 2026-03,17500,14100 2026-04,21000,14900 2026-05,22500,15300 2026-06,21800,15700 Verify these totals: sales 119800, operating expenses 88200 and arithmetic difference 31600. From May to June, sales fall by 700 and expenses rise by 400; sales minus expenses goes from 7200 to 6100. Do not invent a causal explanation. Keep the full title “Informe de actividad: ejemplo ficticio, enero–junio de 2026”, the three questions for the meeting and the visible note about the missing taxes, depreciation, inventory, receipts and payments. The difference is not profit, cash flow or a bank balance. Include “Ejemplo con datos ficticios; no es un informe de un cliente real” on the first page. No more than three pages. Preserve all accents, ñ, ¿?, and the € symbol in both formats. Write amounts in Spanish format, for example 21.800,00 €. Check both files after generating them and confirm in chat that the table matches the original CSV row by row. Return both new downloadable files. Also replace the previous interactive chart with one using these six exact rows and two bar series, sales and operating expenses, with the axis starting at zero. The June row must be 21800 and 15700. Do not generate an artistic image of the chart; we need a calculated chart with exact values.

Follow-up 2

The amounts in the second version are now correct. Preserve its six rows and totals exactly: sales 119.800,00 €, operating expenses 88.200,00 € and arithmetic difference 31.600,00 €. Two issues still need fixing in the files, not just in the chat response: 1. Both documents have lost the comparison between May and June that I requested. Add the section “Comparación de junio con mayo” with this content: “Las ventas pasan de 22.500,00 € en mayo a 21.800,00 € en junio, una disminución de 700,00 €. Los gastos operativos pasan de 15.300,00 € a 15.700,00 €, un aumento de 400,00 €. La diferencia aritmética entre ventas y gastos operativos pasa de 7.200,00 € a 6.100,00 €, una reducción de 1.100,00 €. El archivo no permite determinar las causas de estas variaciones”. 2. The PDF is still removing accents, ¿ and €. A “standard encoding” that changes the Spanish text does not meet the requirement. Generate the PDF with an embedded Unicode TrueType font, such as DejaVu Sans or Liberation Sans. You can use ReportLab and register that font. Do not transliterate to ASCII, remove diacritics or replace € with EUR to avoid the problem. If you need to locate a font, look for a TrueType font available in the environment or among the fonts of the installed libraries. Regenerate the DOCX and PDF with the same complete content: title “Informe de actividad: ejemplo ficticio, enero–junio de 2026”; opening note “Ejemplo con datos ficticios; no es un informe de un cliente real”; six-month table; totals; the comparison above; the three questions for the meeting; and the note identifying the missing taxes, depreciation, inventory, receipts and payments and explaining that the difference is not profit, cash flow or a bank balance. Keep all amounts in Spanish format and a maximum of three pages. Before delivering the PDF, extract its text and check that it preserves “aritmética”, “Comparación”, “reunión”, “¿A qué”, “¿Cómo”, “próximo”, “tesorería” and “€” exactly. Also check that both files include the comparison and that June retains 21.800,00 € and 15.700,00 €. Return both new downloadable files. Do not generate or modify the chart in this turn.

Follow-up 3

You delivered only the DOCX. Now deliver the outstanding PDF of the same complete report, preserving exactly the content and amounts in the DOCX you just generated. Use an embedded Unicode TrueType font available in the environment, preserve accents, ¿ and € exactly, and check the extracted text before delivering the PDF. Do not modify the DOCX or generate the chart again.

Follow-up 4

I reviewed the downloaded PDF: it now preserves accents and ¿, the six rows, the totals and the comparison. The remaining issue is that ALL € symbols have turned into ?. Preserve what is already correct and regenerate only the PDF. The embedded TrueType font I suggested was one technical way to preserve the characters. For this Spanish report, use this specific solution compatible with fpdf2, which is already installed: from fpdf import FPDF pdf = FPDF() pdf.core_fonts_encoding = "cp1252" pdf.set_font("Helvetica", size=11) Set core_fonts_encoding BEFORE writing any text. Pass the original Python strings with €, accents and ¿ directly to the fpdf2 methods. Remove any conversion to latin-1 or ASCII with errors="replace" or errors="ignore", along with any manual substitution of € with ?, EUR or another character. Do not install packages or rely on ReportLab, matplotlib or a font you have not found. The cp1252 encoding preserves every character in this report, including €. Use the complete approved content, with an aligned three-column table, Mes / Ventas (€) / Gastos operativos (€), and these six exact rows: 2026-01,18000,14000 2026-02,19000,14200 2026-03,17500,14100 2026-04,21000,14900 2026-05,22500,15300 2026-06,21800,15700 Totals: sales 119.800,00 €, operating expenses 88.200,00 €, arithmetic difference 31.600,00 €. Keep Spanish formatting for every amount. Title: “Informe de actividad: ejemplo ficticio, enero–junio de 2026”. Opening note: “Ejemplo con datos ficticios; no es un informe de un cliente real”. Section “Comparación de junio con mayo”: “Las ventas pasan de 22.500,00 € en mayo a 21.800,00 € en junio, una disminución de 700,00 €. Los gastos operativos pasan de 15.300,00 € a 15.700,00 €, un aumento de 400,00 €. La diferencia aritmética entre ventas y gastos operativos pasa de 7.200,00 € a 6.100,00 €, una reducción de 1.100,00 €. El archivo no permite determinar las causas de estas variaciones”. Section “Preguntas para la reunión”: 1. ¿A qué factores atribuye la variación observada en las ventas entre mayo y junio? 2. ¿Existen cambios operativos planificados para el próximo trimestre? 3. ¿Cómo planea ajustar la estructura de gastos operativos ante las fluctuaciones de actividad? Final note: “Este informe excluye impuestos, amortizaciones, existencias, cobros y pagos, por lo que la diferencia aritmética no representa beneficio, tesorería ni saldo bancario”. Before creating the PDF, check that the original strings preserve “aritmética”, “Comparación”, “reunión”, “¿A qué”, “¿Cómo”, “próximo”, “tesorería” and “€” exactly, and that encoding and decoding them with cp1252 returns them unchanged. Print the verified table and totals. Then reopen the PDF as bytes, check that it starts with %PDF and print its size and page count. Do not claim to have extracted text from the PDF if no reader is available in the environment: I will review the downloaded file independently. Save only the new PDF in ./output, no more than three pages. Do not modify the approved DOCX or generate another chart.

Initial request

Create a downloadable four-slide PowerPoint presentation in Spanish for a weekly meeting of a five-person team. Title: “Actualización del proyecto de catálogo · ejemplo ficticio”. Use only these invented facts: objective, prepare a catalogue of 12 products; status, 8 product entries reviewed and 4 awaiting photography; next internal delivery, Friday 18 September 2026; person responsible for follow-up, “Persona responsable” (placeholder). Slide 1: objective and status. Slide 2: 8 of 12 product entries reviewed, also showing the 4 pending entries and without inventing overall progress percentages. Slide 3: next steps, review specifications and prepare the four photographs. Slide 4: meeting decisions, who reviews each entry and whether the delivery date stands; leave both answers as “por confirmar”. Write “Demostración con datos ficticios” in every footer. Clear design, white background, dark blue, one green accent and large type. No logos, external photographs, personal data or customer results. No more than six lines per slide. Generate an editable PPTX file, open it and check all four slides before delivering it.

Follow-up 1

Correct the previous PPTX and deliver the editable four-slide file again. Restore the exact title “Actualización del proyecto de catálogo · ejemplo ficticio”. Remove the duplicate bullets from the eleven points: use one native bullet per paragraph, without typing the • character in the text. Add the requested green accent, for example on “8 de 12” on slide 2. Write both answers on slide 4 exactly as “por confirmar”. Preserve the data and all four “Demostración con datos ficticios” footers. Keep the white background, dark blue and large type. Do not exceed six visible lines per slide, including the footer. If the full title takes two lines, combine the status on the first slide into “Estado: 8 revisadas y 4 pendientes de fotografía” to make room. Open and visually check all four slides before delivering the new version.

Initial request

Analyze the attached CSV to prepare a short management note on AI use by businesses in six European countries between 2024 and 2025. Respond in Spanish. These are aggregated public Eurostat data, not customer data or results from an Ilisai test. They come from the Figure 2 sheet in https://ec.europa.eu/eurostat/statistics-explained/images/b/bf/2025_Enterprises_using_AI_tables_and_graphs.xlsx, extracted in December 2025, dataset isoc_eb_ai. Population: enterprises with at least ten employees and self-employed persons, in the business sectors covered by Eurostat’s ICT survey (NACE Rev. 2, C–J, L–N and group 95.1). The variable measures the percentage of enterprises using at least one of the AI technologies surveyed. These are six selected countries, not the whole EU. First check that there are twelve observations, six countries and one row per country and year; detect duplicates, missing values and values outside 0–100. Stop if any check fails. Keep the original values without rounding them before calculation. For each country, calculate the difference between 2025 and 2024 in percentage points. Show a table with both years and that difference, to two decimal places. Sort countries by their 2025 percentage, highest first. Do not calculate an EU average from this selection. Show an interactive grouped bar chart inside the chat: six countries, two series (2024 and 2025), a vertical axis starting at zero, country labels in Spanish, and the unit “% de empresas”. In the CSV, 42.03 means 42.03%, not 4,203%. The chart must show those same values. Write a short note distinguishing observed results from questions these data cannot answer. Do not attribute changes to any cause or turn them into an estimate of productivity or profitability. State that Eurostat added image, video and audio generation as a category in the 2025 questionnaire: this comparison does not isolate the effect of that change. Generate a downloadable XLSX called eurostat-ia-empresas-2024-2025.xlsx, with three sheets: “Datos” (the twelve original rows), “Comparacion” (a table by country and formulas for the 2025 minus 2024 difference) and “Notas” (summary, units, population, source and limits of the comparison). Include the source link and do not invent figures, causes or forecasts.

Follow-up 1

The chart now shows all twelve correct values. Correct the XLSX and complete the response: 1. In “Datos”, preserve exactly the four columns and twelve rows of the original CSV: country_code, country, year, ai_using_enterprises_pct. country_code was missing. 2. In “Comparacion”, keep the six countries in their current order and their 2024 and 2025 values. Column D must contain Excel formulas: D2 =C2-B2, and the same pattern through D7. Do not save only the numeric results. Show two decimal places and make the units clear: B/C are percentages expressed from 0 to 100; D is percentage points. Adjust widths and text so that headings, values and notes are readable. 3. In “Notas”, correct the population to “empresas con al menos diez personas ocupadas, incluidos asalariados y trabajadores por cuenta propia, en NACE Rev. 2 C–J, L–N y grupo 95.1”. Keep the source and add “Figure 2, extracción de diciembre de 2025”. Replace the generic reference to generative AI with: “Eurostat añadió la generación de imágenes, vídeo y audio como categoría en el cuestionario de 2025. Esta comparación no aísla el efecto de ese cambio”. Add that these are six selected countries, not an EU average, and that the data do not measure causes, productivity or profitability. Deliver the corrected XLSX. After the file, include a table in your response with country, 2024, 2025 and difference in percentage points, followed by three sentences: what is observed, what cannot be concluded and the questionnaire limitation. Do not end with just the file. You do not need to repeat the chart.

Follow-up 2

The data and all six formulas are now correct. Finish only the presentation of the “Notas” sheet and deliver the final XLSX: - Leave “Datos” and “Comparacion” unchanged, including their values and formulas. - In “Notas”, use an approximate width of 30 for column A and 90 for B. Enable text wrapping in B3:B9 and set heights that display all the text: for example, 60 points for rows 3, 5, 7 and 9, and 45 for 4, 6 and 8. The text and link must be readable without horizontal scrolling. - Add A9 “Observación:” and B9 “Los seis países registran aumentos. Dinamarca tiene la mayor proporción en 2025 (42,03 %) y el mayor incremento de esta selección (14,45 puntos porcentuales).” Keep the population, source and limitation notes as corrected. Do not generate the chart again or repeat the analysis: deliver the final file with this adjustment.

Follow-up 3

The final XLSX preserves the correct data and formulas, and “Notas” is now readable. Only one visual adjustment remains: “Datos” has reverted to the default widths and two headings are cut off. Open that same final file and change only the widths of the four columns in “Datos”: A=20, B=20, C=12, D=32. Do not rebuild the workbook or change the cells, formulas, other sheets or their formatting. Preserve exactly the four original headings: country_code, country, year, ai_using_enterprises_pct. Keep the text wrapping, heights and observation in “Notas”. Deliver the XLSX with the name “eurostat-ia-empresas-publicable.xlsx”.