Implisense

View saved version →

---
name: company-researcher
description: Builds a structured profile of one German company — master data, business model, management, financials and recent register changes — from the Implisense tools. Use when the user asks about a single German company by name, website, VAT ID, LEI or register number.
---

# Company Researcher

Turn a question about one German company ("Tell me about TRUMPF", "Research the company behind check24.de") into a structured, sourced profile. Answer in the user's language.

## Workflow

1. Call `company_profile` with the identifier the user gave (name, website, VAT ID, LEI or register court + type + number). Pass `include=people,financials,announcements` for a complete picture.
2. If the result is ambiguous, `meta.candidates` lists the matches. Ask the user which one they mean (city or register number), then call again. Never pick one yourself.
3. If `meta.confidence` is `low`, say that the name matched only fuzzily.
4. If the user asks what changed recently, call `recent_changes` with the company `id` and a `since` date.
5. If the user asks for competitors or comparable firms, call `similar_companies` with `seed_id`.
6. If the user asks whom to contact in a specific function (sales, purchasing, IT, …), follow the `contact-finder` skill.

## Profile structure

- **Master data**: name, legal form, registered seat (city, state), founding date, size class, industries, register number, VAT ID.
- **Business model**: two or three sentences from `summary` — what the company does, for whom, what stands out.
- **Management**: managing directors, board members and authorized signatories with their roles.
- **Financials**: latest available year with its year stated — revenue or balance-sheet total, net income, equity and equity ratio. If several years are available, a short trend. A brief qualitative reading: profitable? solidly capitalised?
- **Recent changes**: relevant register announcements, if requested.
- **Assessment and open questions**: one or two sentences on what to check next.

## Rules

- Use only what the tools return. If a figure is missing, say so — never estimate or invent numbers.
- Always state the financial year next to a figure.
- No forecasts; interpret historical data only.
- For several companies at once use the `portfolio-analyst` skill; for prospect lists use `lead-qualifier`; for contact persons use `contact-finder`.

View saved version →

---
name: contact-finder
description: Finds decision makers at a German company for given roles (management, sales, purchasing, IT, HR, marketing, finance, sustainability) with sources, using the Implisense real-time contact search. Use when the user asks who to contact at a company or for a contact person in a specific function.
---

# Contact Finder

Find the right business contacts at one German company, with the sources that prove each person. Answer in the user's language.

## Workflow

1. **Resolve the company.** If you don't have its Implisense ID yet, call `company_profile` (or `search_companies` if the user describes the company rather than naming it). If the name is ambiguous, ask which company is meant before going on.
2. **Pick the roles.** Map the user's wording to the role codes `find_contacts` accepts: `MANAGEMENT`, `SALES`, `MARKETING_COMMUNICATIONS`, `PROCUREMENT`, `HUMAN_RESOURCES`, `IT`, `FINANCE`, `SUSTAINABILITY`. Leave `roles` empty for the standard profile.
3. **Run the search.** Call `find_contacts` with `company_id` and `roles`. It researches public sources in real time, costs 3 Implicents per company (refunded if it fails) and takes about a minute. If the user asks for several companies, say what it will cost and confirm first.
4. **Collect the result.** If `find_contacts` returns a `job_id` with status `running`, wait about 30 seconds and call `get_contact_results` with that `job_id` (free). Repeat until the status is no longer `running`. Results are deleted one hour after the search finished.

## Output

Per person: name, role, and the published business channels (email, phone, LinkedIn) — followed by the source URLs. Then:

- Mark entries with `unconfirmed: true` as unconfirmed and suggest checking them before reaching out.
- A phone number with `phoneType: switchboard` is the company's main line, not a direct dial — say so.
- If no one was found for a role, say that plainly and suggest the management from `company_profile` as a fallback.

## Rules

- Business contacts only. Never search for or present private addresses, private phone numbers or other personal details.
- Use only what the tools return; never guess email addresses from name patterns or invent people.
- For the registered management (managing directors, board, authorized signatories) `company_profile` is enough and costs less — use `find_contacts` when the user needs a specific function or direct contact details.

View saved version →

---
name: lead-qualifier
description: Finds and prioritises German companies that match an ideal customer profile (size, industry, region, age, financial criteria) using the Implisense search, lookalike and profile tools. Use when the user wants a target list, prospects, or to qualify a list of companies against criteria.
---

# Lead Qualifier

Turn an ideal customer profile (ICP) into a prioritised list of German companies with reasons. Answer in the user's language.

## Workflow

1. **Clarify the ICP** if it is not given: size class, industries, region, company age, financial criteria (e.g. profitable, minimum equity ratio) and anything else that matters. Suggest an example ICP if the user is unsure.
2. **Find candidates**:
   - From criteria: translate the ICP into `search_companies` filters — `industry` (WZ 2008 codes or prefixes), `size` (MICRO, SMALL, MEDIUM, LARGE), `location` (e.g. `state:DE-BY`), `min_age`/`max_age`, plus `query` for keywords. Start with the default limit of 20 and page with `offset` only if needed.
   - From example customers: resolve them with `company_profile`, then call `similar_companies` with their IDs as `seed_ids` to find lookalikes.
   - From a list the user brings: resolve each company with `company_profile`.
3. **Check the best candidates** with `company_profile` (`include=financials`) where the ICP has financial criteria.
4. **Contacts** only if the user asks: follow the `contact-finder` skill for the companies they pick. Each search costs 3 Implicents and takes about a minute, so confirm before running it for more than one company.

## Output

A list sorted into High, Medium and Low fit, with one or two sentences per company on which ICP criteria it meets or misses. Collect clearly unsuitable companies briefly at the end instead of explaining each one.

## Rules

- Qualitative fit (High / Medium / Low), no invented scores.
- Use only what the tools return; never invent contacts, buying signals or figures.
- For a risk check of a list use `portfolio-analyst`; for one company in depth use `company-researcher`.

View saved version →

---
name: portfolio-analyst
description: Checks a list of German companies — customers, suppliers, partners or investments — for risk flags such as insolvency or liquidation and groups them into a portfolio overview. Use when the user gives several companies and asks about risks, health or warning signs.
---

# Portfolio Analyst

Turn a list of German companies into a risk overview with a short assessment per company. Answer in the user's language.

## Workflow

1. Collect the companies from the user: names, Implisense IDs or a pasted list. If the list is empty, ask for it.
2. Call `portfolio_risk` with the list (up to 100 entries per call; split larger lists into batches).
3. Entries with `match: null` could not be resolved or were ambiguous. List them separately and ask the user for a city, website or register number — never guess which company was meant.
4. If the user wants more depth on a single company, call `company_profile` with `include=financials` for it.

## Grouping

Use the fields `portfolio_risk` returns:

- 🔴 **Risk**: `warning_exists` is true, `active` is false, or a `liquidation_date` is set.
- 🟡 **Watch**: no flag, but the latest net income is negative or no financials are available.
- 🟢 **No flags**: none of the above.

## Output

- A table: company, group, the flag or key figure behind it (with year).
- One sentence per company in the Risk and Watch groups.
- An overall reading: distribution across groups and anything notable.

## Rules

- This is a screening based on register flags and the latest reported figures, not a credit rating. Say so once in the answer.
- Use only what the tools return; never invent figures or flags.