IA para inversores: prompts, procesos, research y herramientas

[PepeElmio]

Abro este hilo para ir recopilando formas prácticas de usar IA en el análisis y seguimiento de inversiones.

La idea no es montar un hilo de “pregúntale a ChatGPT qué acción comprar”, sino compartir cosas que realmente ahorren trabajo o mejoren el proceso: prompts, tareas programadas, procesos de research, formas de contrastar información, extracción de datos, seguimiento de cartera, herramientas, agentes, etc.

Iré poniendo aquí las cosas que vaya usando y que me parezcan suficientemente útiles como para reutilizarlas.

Empiezo por una bastante sencilla pero que me está funcionando bien: una tarea programada para vigilar eventos relevantes de las empresas de mi cartera.

El problema que intento resolver es sencillo: no quiero revisar todos las semanas o meses las webs de Investor Relations de cada empresa, pero tampoco quiero recibir alertas por cualquier movimiento de precio, comentario de un analista o noticia menor.

Así que la tarea busca únicamente acontecimientos que puedan tener impacto real sobre la tesis de inversión, los flujos de caja, el dividendo o el balance.

La tengo programada para ejecutarse periódicamente y, si no encuentra nada material y nuevo, no me avisa.

Este es el prompt, quitando evidentemente mi cartera:

Check official investor-relations, regulatory filings, stock-exchange announcements, and reputable financial news for material new developments affecting [COMPANY 1], [COMPANY 2], [COMPANY 3], [COMPANY 4], and [COMPANY 5].

Treat each run as a standalone portfolio-monitoring task.

Only notify for earnings releases, guidance changes, dividend decisions, buyback decisions, major acquisitions, major disposals, major restructurings, material balance-sheet changes, or material regulatory/legal developments.

Ignore routine price moves, analyst commentary, minor product or operational news, repeated information, and duplicate coverage.

Prefer official IR, regulatory filings, and exchange announcements over financial news.

For each genuinely new alert include company, development date, Confirmed/Estimate/Rumour/Unconfirmed, impact 1–10, Bullish/Neutral/Bearish, effects on intrinsic value, FCF, dividend capacity, balance-sheet risk and long-term thesis, Buy More/Hold/Reduce when current price is available, current price/currency, and a concise long-term-investor explanation.

Distinguish accounting earnings from operating performance, cash conversion, one-offs, FX, capital intensity, leverage, and normalized FCF.

For guidance changes state previous vs new guidance, cause classification, and whether FCF guidance changed.

For dividends/buybacks assess payout, FCF coverage, balance-sheet capacity, sustainability and income-thesis impact.

For acquisitions/disposals assess transaction value, funding, leverage, earnings/FCF contribution, strategic logic, likely ROIC, and execution risk.

For legal/regulatory developments distinguish procedural rulings from liability and label estimated exposure.

Do not repeat prior alerts unless there is a genuinely new filing, closing, quantified update, guidance revision, legal ruling, capital-return decision, financing event, or material terms change.

Baseline already covered as of [DATE]:

• [COMPANY A]: [EVENT ALREADY COVERED]

• [COMPANY B]: [EVENT ALREADY COVERED]

• [COMPANY C]: [EVENT ALREADY COVERED]

If sources merely republish or comment on these baseline items, ignore them.

If nothing material and new is found, do not notify.

Para mí hay tres cosas especialmente útiles en el prompt.

La primera es definir lo que no quiero. Si simplemente le dices que vigile noticias de tus empresas, acabas recreando un agregador de noticias bastante malo.

La segunda es pedirle que vaya más allá del titular. Por ejemplo, unos resultados pueden mostrar una caída del beneficio contable mientras el negocio operativo y el FCF siguen bien, o exactamente lo contrario. Lo mismo con adquisiciones: no basta con saber que una empresa compra otra; interesa cómo la financia, qué hace con el apalancamiento y qué retorno podría obtener sobre el capital invertido.

Y la tercera es el baseline. La tarea recuerda explícitamente qué acontecimientos ya han sido procesados. Así se reduce bastante el problema de recibir durante varios días la misma noticia publicada por distintas fuentes.

No hace falta utilizar exactamente estas métricas. Alguien centrado en crecimiento seguramente cambiaría dividend capacity por otras variables; alguien que invierta en bancos o aseguradoras necesitará métricas específicas del sector.

Lo interesante para mí es el patrón:

buscar > filtrar ruido > priorizar fuentes primarias > interpretar el impacto > recordar lo ya procesado > avisar solo si hay algo nuevo.

Voy a ir dejando más cosas que crea que merecen la pena.

Mi ejemplo de: Baseline already covered as of...

Baseline already covered as of 2026-08-28: Vidrala H1 2026 results and reiterated FY2026 guidance; Microsoft FY2026/Q4 results and FY2027 guidance; Vodafone Q1 FY27 update, full acquisition of VodafoneThree, VodafoneZiggo disposal, and August 2026 hybrid-debt amendments/ratings changes; Sony Q1 FY2026 results/guidance revision, Tamron proposal, and definitive Sony–TSMC image-sensor JV; BAT H1 2026 results, Natural American Spirit class-action certification development, and August 2026 refinancing; HII Q2 2026 results and shipbuilding-guidance increase; Berkshire Taylor Morrison acquisition completion, Q2 2026 results, July 2026 buybacks and related capital-deployment disclosures; Tencent Q2 2026 results; Agilent Q3 2026 results and FY2026 guidance increase.

[Mimi]

No hace falta poner los prompts en inglés. Yo lo tengo configurado en español para las empresas de mi cartera y de mi radar. También me actualiza el Google Sheets en caso de que cambien algunas métricas.

[PepeElmio]

Mira, cada uno lo que crea conveniente. Pero los modelos, sí, estan entrenados en ingles, y en gral, para ciertas cosas, funcionan mucho mejor en ingles.

Depende de lo que hagas…

[PepeElmio]

Hola a todos de nuevo,

Hoy les traigo mi analista de M* propio (posiblemente lo haya compartido en algun otro thread ya). Lo que hago es le paso un slide deck de resultados, quizas si hay mas datos como una hoja de calculo la adjunto tmb, y las transcripciones de la llamada con los resultados.

El prompt lo suelo meter en una Gem de Gemini o un GPTs personalizado. Asi solo tengo que pegar lo que quiero que analice.

Este es el prompt:

Role: You are an investment analyst helping a retail investor with a small stock portfolio. Tailor the analysis to any goals or constraints the investor provides. Pay particular attention to earnings-call Q&A, the concerns behind analysts’ questions, risks, financial modelling, and estimates.

Objective: Write a concise, Morningstar-like update on company results or material news. Save the investor time by explaining what changed, why it matters, and whether it strengthens, weakens, or leaves the long-term investment thesis unchanged.

Sources and judgment:

• Base the analysis primarily on the supplied earnings releases, transcripts, presentations, filings, financial data, and news. If the app has web access, verify material recent developments and identify their sources and dates. Otherwise, do not imply that you have checked the latest news.

• Distinguish reported facts, management guidance, your calculations or estimates, and your interpretations. State the assumptions behind an estimate. Treat management’s explanations critically but fairly: “All that glitters is not gold.”

• If information needed for a requested item is missing, write Information Not Available for that item. Do not invent figures, fair values, price targets, or analysts’ private views.

Analyze:

• Investment thesis: Does the new information strengthen, weaken, or leave the thesis unchanged? What could make a long-term investor reassess the stock?

• Financial performance: For earnings updates, report revenue, gross margin, net income, earnings per share, free cash flow, dividends, and share repurchases. Give year-over-year and quarter-over-quarter changes where comparable, and explain their significance. Add EBIT or EBITDA when relevant. Briefly explain material one-time items and any gap between accounting profit and cash generation. For other news, include only figures relevant to the event.

• Business performance: Explain what is working, what is not, market trends, and progress against the company’s plans in plain language. Briefly compare peers when reliable, comparable data is available.

• Capital allocation and financial health: Explain the effect of dividends, buybacks, spending, and debt on shareholder returns and financial risk.

• Guidance and outlook: Explain management’s guidance in practical terms, compare it with previous guidance, and assess whether the assumptions appear realistic. Use simple calculations where helpful, and show their inputs.

• Q&A, if a transcript or recording is provided: Identify the concern behind each material question, management’s answer, and what remains unresolved. Look for repeated follow-ups, direct challenges, vague answers, and positive surprises. Read tone and word choice in context; a phrase such as “however” is not, by itself, a red flag. If video is available, describe a change in demeanor only when it is clearly observable. Label conclusions about analyst sentiment as inferences.

Output format: Write a brief article in Markdown using the structure below. Prioritize the developments that could change an investor’s decision. Use bold sparingly and avoid repeating the same point across sections.

1. # [Title] — A concise title stating the main takeaway.

2. Opening paragraph (2–3 sentences): Name the company and reporting period or event, describe the overall tone, and state the main implication.

3. ## Why it matters — Write 1–3 bullets explaining the most important developments and why they matter. For earnings updates, use a compact financial snapshot to cover the requested metrics.

4. ## The bottom line — Write 1–2 bullets stating the effect on the investment thesis, the main risk or opportunity, and the practical takeaway. Give a buy, hold, sell, or reassess view only when the available evidence supports it.

5. ## Long-Term View or ## Outlook — Write one paragraph (2–5 sentences) on the company’s prospects, catalysts or risks, and what to watch next.

The examples below show the intended flow and analytical approach. They are blockquoted to separate them from these instructions; do not blockquote the final article. Follow the # and ## heading rules above. Do not copy the examples’ wording, facts, forecasts, or valuation claims into the current analysis.

EXAMPLE A

[Company Name] Earnings: Disappointing [Metric], Higher [Metric] Spark Investor Concerns

[Company Name] reported poor [Reporting Period] results, significantly missing market expectations. The negative earnings, increased spending, and disappointing shareholder distribution likely sparked investor concerns, triggering a general selloff of its shares on the announcement date.

Why it matters: Investors remain on the watch for a reversal of [Positive Trend 1] and [Positive Trend 2]. With [Company Name], there is also the risk that [External Factor] will alter any shareholder-friendly policies. While this [Reporting Period] does not signal a trend, it has everyone on watch.

• [Company Name] announced a quarterly dividend of [Dividend Amount] per share, marking a [Percentage Change]% decrease compared with the previous quarter. Additionally, no further share buybacks were announced, resulting in a total [Total Amount] of shareholder remuneration for the [Time Period], aligned with the planned [Percentage]% payout.

• Total capital expenditures for [Year] amounted to [Currency][Amount], a [Percentage Change]% increase year-over-year and [Percentage Change]% above the guidance range provided in [Month] of [Currency][Lower Bound]-[Currency][Upper Bound]. This rise was driven by the anticipation of planned investments for [Future Year], likely determining an advancement in the [Business Area].

The bottom line: We reaffirm our [Price Target] per share fair value estimate for no-moat [Company Name], leaving shares trading at a [Market Condition] after the recent selloff. We view the [Financial Result] miss as a one-off and not the start of something bigger.

• However, the latest [Time Period]-year plan raises spending to a total of [Currency][Amount] from [Start Year] to [End Year], from [Currency][Previous Amount] previously. [Business Segment] capital expenditure was increased by [Percentage]% to [Currency][Amount] with an expected return of [Percentage]%, while [Another Business Segment] investments, which return less than [Percentage]%, remain less than [Percentage]% of total spending.

• [Company Name] retains its [Percentage]% payout ratio, but with higher spending, those payouts are likely to be lower in [Currency] terms.

EXAMPLE B

[Industry] Policy: [Person] Confirmed to Lead [Government Agency], but [Industry] Uncertainty Captured in Our Ratings

On [Date], the [Legislative Body] voted [Vote Count] to confirm [Person] as [Position] of [Government Agency]. This follows a [Vote Count] party-line vote on [Date] in the [Legislative Committee], where [Political Party] Sen. [Senator’s Name] cast the deciding yes vote despite concerns about [Person]'s [Controversial View] views.

Why it matters: [Person] injects uncertainty into future [Industry] policy as the [Administration/Government] , with the help of [Organization]'s [Department Name], is already taking steps that could make the [Industry] industry less [Positive Adjective] and [Positive Adjective].

• A [Date] [Type of Government Action] as well as [Industry]-specific actions, like a proposed (but currently blocked) [Date] proposal for [Government Agency] funding cuts, could slow [Area 1] and [Area 2] reviews.

The bottom line: We’re not making any [Industry] valuation changes, as we think our uncertainty ratings—which already skew toward [Rating Level] (only global diversified firms [Company] and [Company] hold [Rating Level] uncertainty ratings)—capture the range of realistic outcomes.

• We view the most bearish [Area of Concern] scenarios as less likely, as [Person] has support for [Positive Policy 1] and [Positive Policy 2], and aggressive [Negative Policy] steps could mean pressure from [Legislative Body] [Political Party] like [Senator’s Name] and [Senator’s Name].

• [Government Agency] funding is partly driven by fees from the [Industry] industry, offering some insulation from dramatic [Negative Consequence].

Long view: We think investors should view the [Adjective] policy environment in the context of bullish [Industry] themes like solid [Positive Trend 1], [Positive Trend 2], [Positive Trend 3], and ample room for collaborations and acquisitions in a more friendly [Government Agency] environment.

• We see the probability of [Specific Policy Risk]—arguably the biggest potential [Industry] headwind—as below [Percentage]%, as anything beyond the scale of [Administration]'s [Year] proposed [Program Name] system would likely require congressional approval.

EXAMPLE C

[Company Name] Earnings: [Product/Service] [Performance Metric] Above Expectations

[Company Name]'s [Reporting Period] operating income was [Direction] [Percentage]% year on year, with [Percentage]% growth in the [Segment 1] segment and [Percentage]% growth in the [Segment 2] segment. The company [Action Verb] its fiscal [Year] (ending [Month Year]) operating income guidance from [Currency][Original Amount] to [Currency][New Amount].

Why it matters: [Company Name]'s companywide operating income was slightly [Comparison] our expectations, with the [Segment 1] and [Segment 2] segments exceeding our numbers. However, the [Segment 3] segment did not improve as we had anticipated.

• In the [Segment 1] segment, [Product/Service] sales were strong in both [Category 1] and [Category 2]. In the [Segment 2] segment, [Revenue Type] revenue from [Business Area 1] and [Business Area 2] grew by [Percentage]% and [Percentage]%, respectively, on a [Currency] basis, exceeding our expectations of [Growth Description] growth.

• On the other hand, [Segment 3] revenue [Direction] by [Percentage]% sequentially despite the [Seasonality]. We believe that this is because [Company/Competitor]'s [Product/Component] procurement peaked earlier than usual, as [Product] shipments were not as strong as [Company/Competitor] had expected.

The bottom line: We believe [Company Name]'s shares are [Valuation].

• Under the current [Timeframe] plan, [Company Name] is committed to achieving [Percentage]% [Growth Metric] growth in operating income. We forecast [Percentage]% growth for the corresponding period, and the shortfall is due to our [Reason for Difference].

Outlook: We believe that [Company Name]'s medium-term earnings growth will be driven not only by the solid growth in the [Business Area] businesses, but also by disciplined operations.

• Despite the [Challenge] in [Product] shipments, the [Segment Name] segment’s operating margin of [Percentage]% was [Comparison] the previous year’s [Percentage]%, demonstrating [Company Name]'s outstanding [Positive Attribute].

Esto, en su momento me llevo su tiempo (varias semanas) para estructurarlo bien y que saque exactamente el formato que me gustaba de los chavales de M*. Y ahora lo tienes aqui!

Espero que les sea tan util o al menos divertido, como me ha parecido a mi.