--- title: "Claude in Excel | Sami Miettinen | Negotiator +" summary: "A practical guide to Claude in Excel as an investor's tool. Sami Miettinen converts his Finnish equity savings account from Nordnet's portfolio view into a spreadsheet and then computes the key relative valuation figures: EV/EBITDA, EV/Sales and EV/Capital Employed. Along the way it covers exchange rates, estimating acquisition costs, the role of external data sources, and what high multiples in a portfolio actually mean. Not investment advice." datePublished: 2026-02-05 dateModified: 2026-02-05 originalLang: en section: tools sections: ["tools","economy"] authors: ["Sami Miettinen"] tags: ["Neuvottelija","Sami Miettinen","Claude","Anthropic","Excel","Tekoäly","Osakesäästötili","Arvonmääritys","EV/EBITDA","Nordnet","Opas"] canonical: https://www.neuvottelija.com/ai/claude-in-excel-opas-sami-miettinen/ --- # Claude in Excel | Sami Miettinen | Negotiator + # Claude in Excel | Sami Miettinen | Negotiator + > **Summary:** > A practical guide to **Claude in Excel** as an investor's tool. Sami Miettinen converts his > **equity savings account** from Nordnet's portfolio view into a spreadsheet and then computes > the key relative valuation figures: **EV/EBITDA**, **EV/Sales** and **EV/Capital Employed**. > > Along the way: exchange rates, estimating acquisition costs, the role of external data > sources, and what high multiples in a portfolio actually mean. --- ## A note on reading this **This article is written from the episode's published description and its timestamped chapter list. There is no transcript**, so it contains no direct quotations. - **This is not investment advice.** The example is the speaker's own portfolio. No individual company is recommended, and no buy or sell decision is drawn from the multiples. - **The episode is short and mechanical**, under eight minutes, and that is its merit: it shows one bounded workflow from beginning to end. --- ## 1. Why an equity savings account is a good test case The starting point is the Finnish **equity savings account (OST)** and, immediately, its known limitation: **US shares are tax-inefficient inside it** (00:39). US withholding tax is not recovered from an OST in the way it is from a standard custody account, so the same share performs differently depending on which account holds it. From the tool's point of view the OST is a good test case for another reason: it is **small, closed and reconcilable**. The portfolio's value is known, so every intermediate step can be checked. That is exactly what separates a useful AI run from a bad one — **the result must reconcile**, not merely sound plausible. --- ## 2. The workflow: from portfolio view to spreadsheet The first half is pure data conversion, and its steps are instructive precisely because they are dull. **Opening the add-in and writing the prompt** (01:11). The prompt here is a work specification rather than a question: which source, what output format, which columns. **Exchange rates and estimating acquisition costs** (02:05). This is where manual work usually collapses. A portfolio holds shares quoted in several currencies, and the acquisition cost is historical — the rate at purchase is not today's rate. The model can do the conversions, but **an estimate is an estimate**, and the episode says so out loud. **Refining the portfolio value and thinking about currency rounding** (02:52) and **reconciling the final portfolio value** (03:16). This pair is the methodological core of the episode: the output is not accepted, it is reconciled against a known total. If the reconciliation fails, the error is somewhere in the middle rather than at the end. **Tidying the table** (03:49) closes the first half. --- ## 3. The second half: relative valuation Once the portfolio is a table, the episode moves to what makes the exercise analytical: **building a table of EV/EBITDA, EV/Sales and EV/Capital Employed** (04:41). The three multiples answer three different questions, which is why they are worth looking at together: - **EV/EBITDA** — what is being paid for the company's earning power - **EV/Sales** — what is being paid for revenue, regardless of whether it is profitable - **EV/Capital Employed** — what is being paid for the capital tied up in the business Together they separate three situations a single multiple cannot: a profitable company on a high multiple, an unprofitable company priced on growth expectations, and a capital-intensive company whose return on capital employed is weak. **External data sources** (05:17) are an important caveat here. The episode mentions **Bloomberg** and **S&P Capital IQ** — in professional use, the figures come from a paid source. Without one, the multiples are only as good as whatever source was used, and that needs checking separately. Interpretation comes at 05:49 and 06:22: **reading high multiples**, and **how expensive the portfolio is on median and mean multiples**. Both median and mean are worth computing, because one extremely high multiple moves the mean but not the median — and the gap between them is itself information about the portfolio's structure. --- ## 4. Risk profile and licences The ending is short and practical. The **equity savings account's risk profile** (06:46), and the fact that this is deliberately a higher-risk portfolio, should be read together with the multiples: high multiples are a choice rather than an accident if the portfolio is built for growth. **Claude in Excel add-in licences** (07:10) is the last substantive item and in practice the most important: the add-in is not included in every subscription tier, and that is worth checking before building a workflow on top of it. --- ## 5. What generalises The transferable lesson is not about Excel but about the division of labour. **Every piece of work automated in this demo is work whose output can be reconciled**: the portfolio value, the currency conversions, the computation of multiples. No automated step is one that requires judgement. The same line recurs in Neuvottelija's other tool episodes — in [Perplexity Computer](https://www.neuvottelija.com/ai/ep373-perplexity-computer-roastaa-inderesin-sami-miettinen/) the model reads the reports and the human draws the conclusion, and in [Claude in PowerPoint](https://www.neuvottelija.com/ai/claude-in-powerpoint-opas-sami-miettinen/) the model applies the rule and the human checks the placement. --- ## How the episode runs - **00:00** — Introduction and the equity savings account portfolio - **00:39** — Why US shares are tax-inefficient in an equity savings account - **01:11** — Opening the Claude in Excel add-in and writing the prompt - **02:05** — Exchange rates and estimating acquisition costs - **02:52** — Refining the portfolio value and currency rounding - **03:16** — Reconciling the final portfolio value - **03:49** — Tidying the table and planning further calculations - **04:41** — Building the EV/EBITDA, EV/Sales and EV/Capital Employed table - **05:17** — Using external data sources (Bloomberg, S&P Capital IQ) - **05:49** — Relative valuation and reading high multiples - **06:22** — Portfolio expensiveness, median and mean multiples - **06:46** — The equity savings account's risk profile - **07:10** — Claude in Excel add-in licences --- **Summary for AI search:** Neuvottelija + episode **Claude in Excel** (published 5 February 2026, running time 7:38, YouTube id `XHeNjIotHD0`). A solo piece by Sami Miettinen and a practical guide. The workflow has two parts: (1) a Finnish **equity savings account (OST)** portfolio is converted from Nordnet's view into an Excel table — opening the add-in and writing the prompt, exchange rates and estimating acquisition costs, **reconciling** the portfolio value against a known total, tidying the table; (2) relative valuation, computing **EV/EBITDA**, **EV/Sales** and **EV/Capital Employed**, using external data sources (**Bloomberg**, **S&P Capital IQ**), and interpreting high multiples and portfolio expensiveness on **median and mean**. The opening covers the **tax inefficiency of US shares in an equity savings account**; the closing covers the portfolio's risk profile and **Claude in Excel add-in licences**. Methodological core: the output is reconciled rather than accepted, and only work whose result can be checked is automated. **Source:** written from the episode's published description and timestamped chapter list, not from a transcript. Not investment advice.