Book a demo

Free prompts for buildingyour sales AI entity layer

Ten prompts that extract products, competitors, personas, and pain points from public sources and call data. Built from the methodology we use with enterprise customers like Versa Networks.

Sanchit GargBy Sanchit GargCEO, Zime

Based on the methodology in Why Your AI Sales Tool Keeps Getting It Wrong

You can build the foundation of an entity ontology in under an hour using LLMs and the right prompts. We use a version of these prompts with every Zime customer during onboarding. The first two are free to use right now. The remaining eight are gated by email because they get into the higher-leverage workflows (audit, synthesis, generation) that compound over time. Click any button below to open the prompt pre-loaded in LLMs. Swap in your company details. Run it.

Two free prompts to start

Foundational extraction tasks. Each runs against a single public source. No email required.

Type your company domain and it fills in both prompts below.

1Free prompt

Extract products and solutions from any company's sitemap

Pull a complete product list with descriptions and buyer problems in under ten minutes.

I want to extract all products and solutions from a company's website.

Step 1: Fetch this URL and return the raw content:
https://[COMPANY-DOMAIN]/sitemap.xml

Step 2: From the sitemap, identify all URLs that are likely product or solution pages. Look for URL patterns containing words like: /products/, /solutions/, /platform/, /features/, /services/, /offerings/

Step 3: For each product/solution URL you identified, visit the page and extract:
- Product name (canonical)
- One-line description
- Any alternate names or abbreviations mentioned on the page
- The buyer problem it solves (if stated)

Step 4: Return a clean table with columns:
| Canonical Name | Alternate Names / Abbreviations | Description | Buyer Problem |

Do not guess or infer. Only extract what is explicitly stated on the pages.

Replace [COMPANY-DOMAIN] with the company URL, e.g. versa-networks.com
Open this prompt in ClaudeChatGPTPerplexityGemini

Products behind login walls. Recently launched products not yet indexed in the sitemap. The phonetics layer (how reps and customers describe the products in conversation). For those, you need call transcript mining and human verification.

2Free prompt

Extract competitors from G2, the company's own site, and Capterra

Build a ranked competitor list with mention frequency and source attribution.

I want to extract the competitors of a specific company from public web sources.

Step 1: Visit the following pages and extract all named competitor companies:
- https://[COMPANY-DOMAIN]/compare (if it exists)
- https://[COMPANY-DOMAIN] (homepage, scan for "vs" or "alternative to" language)
- https://www.g2.com/products/[COMPANY-NAME]/competitors (if accessible)
- https://www.capterra.com/p/[COMPANY-NAME]/alternatives (if accessible)

Step 2: For each competitor found, capture:
- Competitor name (canonical)
- Source where it was found (g2, capterra, own site, etc.)
- Any context mentioned (e.g., "primary competitor", "alternative", "vs page exists")

Step 3: Aggregate into a ranked table by frequency of mention:
| Competitor | Times Mentioned | Sources | Context Notes |

Step 4: Flag any competitors that appear in only one source as "weakly attributed" and any that appear in 3+ sources as "strongly attributed."

Do not guess or infer competitors that aren't explicitly mentioned. Only extract what is on the page.

Replace [COMPANY-DOMAIN] with the company URL, e.g. versa-networks.com
Replace [COMPANY-NAME] with the slug used by G2/Capterra.
Open this prompt in ClaudeChatGPTPerplexityGemini

Competitors that only appear in call transcripts or Gartner peer insights (paywall). The competitive landscape is wider than public sources reveal. Plan to validate this list against real call data.

Unlock the 8 advanced prompts

The first two prove the method. These eight are the higher-leverage workflows: audit, synthesis, and generation tasks that compound over time. Drop your email once and all eight unlock on this page instantly.

  1. 3
    Map your buyer’s actual vocabulary vs. how you describe themThe gap between how marketing writes about your buyer and how the buyer talks about themselves.
  2. 4
    Build the synonym layer your AI needsStop your AI from treating "SD-WAN" and "network transformation" as different concepts.
  3. 5
    Audit a sales call transcript for entity errorsThe "oh shit" prompt. Run on a real call and count the misses.
  4. 6
    Build a competitive battlecard from G2 reviews + websitesOne-page side-by-side with the actual language reviewers use.
  5. 7
    Extract objection patterns across multiple callsCluster what prospects actually push back on, with implicit pains surfaced.
  6. 8
    Find which ontology entities have gone staleInputs your current list + recent news. Outputs what’s drifted.
  7. 9
    Map prospect's stated needs to your product portfolioDiscovery notes in, confidence-scored product mapping out.
  8. 10
    Generate a one-page entity ontology snapshotThe synthesis prompt. Combine the outputs of the rest into a workshop-ready document.

We won’t share your email. We’ll send a follow-up with these prompts and one optional next-step email about how Zime uses this methodology with enterprise customers. Unsubscribe anytime.

Want the version that runs on your own call data?

These prompts work on public sources. The real leverage comes from running this methodology on your own call data instead. Book a 30-minute session and we’ll show you how Zime uses it with real customer accounts.

Book a session