CRM That Imports Business Cards Automatically

You came home from the conference with 23 business cards. They're sitting on your desk. It's been two weeks. You still haven't entered them into anything.

This is normal. It's also a problem, because those 23 people are slowly becoming strangers again.

DenchClaw can process a business card photo into a CRM contact entry in about 10 seconds. Here's how it works.

The Business Card Pile Problem#

Business cards are a fundamentally awkward format. They're physical objects that need to become digital records, and the translation layer — you, manually typing names and emails — is the bottleneck.

Various apps have tried to solve this over the years. CamCard, Contacts+ scanner, even Apple's built-in card scanner. They work okay for personal contacts. But they don't connect to your CRM, they don't let you add context notes, and they definitely don't help you follow up.

What you actually need isn't just extraction — it's extraction + enrichment + filing + follow-up trigger. That's what DenchClaw does.

How the Image Analysis Works#

DenchClaw uses AI vision to analyze business card photos and extract structured contact data. The flow:

1. You send a photo of the business card (via Telegram, WhatsApp, Discord, or iMessage)
2. The agent analyzes the image and identifies fields: name, title, company, email, phone, website, address
3. It creates a new contact entry in DuckDB with those fields populated
4. It asks if you want to add any context notes ("Where did you meet? Any notes?")
5. Done — contact is in your CRM

The whole process takes under 30 seconds including the time to send the photo.

What Gets Extracted#

The vision model is good at pulling:

• Full name (handles both Western and non-Western name formats)
• Job title / role
• Company name
• Email address (one or multiple)
• Phone number (mobile, office, direct — labeled when the card distinguishes)
• Website / LinkedIn URL if printed on the card
• Physical address if present
• Company logo / branding (stored as context, not a separate field)

Cards with unusual layouts, logos over text, or embossed printing can trip it up occasionally. Standard printed cards: essentially 100% accuracy. Fancy foil-embossed cards from that VC at the conference: maybe 90%. Good enough.

The Full Workflow: Photo to CRM Entry#

Here's the complete flow in practice:

At the conference (or right after):

Send a photo of the business card to your DenchClaw Telegram bot with a quick note:

[photo of card] "Picked this up at YC Demo Day. She's building a logistics startup, seemed interested in our API."

DenchClaw extracts the contact, creates the entry, and appends your note to the entry document. The contact now has:

• Structured fields (name, title, company, email, phone)
• A context note about where/how you met
• A timestamp
• A "Source: Business Card" tag

Later, when you want to follow up, all that context is there.

Deduplication: What Happens When the Person Is Already in Your CRM#

DenchClaw checks for existing contacts before creating a new entry. It looks for matches on:

• Email address (exact match)
• Name + Company (fuzzy match)
• Phone number

If it finds a likely match, it'll tell you:

"This looks like it might be Marcus Rodriguez at Acme Corp — you already have an entry for him. Want to update his record or create a new one?"

You choose. Usually you'll want to update (add the new card as a data source, update the title if it's changed) rather than create a duplicate.

Batch Import: Processing a Stack After an Event#

After a conference, you don't want to send 20 photos one at a time. DenchClaw handles batch imports.

You can either:

Option 1 — Send multiple photos in one message: Most messaging apps allow multi-photo sends. Send them all at once. DenchClaw will process each card and create an entry for each one.

Option 2 — Specify batch mode:

"I'm going to send you 15 business cards from NeurIPS. Process them all as new contacts, tag them 'NeurIPS 2026', and ask me for notes on each one after you're done extracting."

DenchClaw will queue through them, then prompt for context notes at the end.

Option 3 — Photo folder import: If you've already taken photos and they're in a folder, you can point DenchClaw at the folder: "Import all business card photos from ~/Desktop/cards/ as contacts."

The Mobile Workflow#

The most useful scenario: you meet someone, swap cards, and before you've even left the building you've got them in your CRM.

On iOS or Android:

1. Open Telegram (or WhatsApp, or iMessage)
2. Take a photo of the card directly in the chat
3. Add a quick voice note for context
4. Send

DenchClaw does the rest. By the time you're on the train home, that contact is fully filed with your notes attached. The 48-hour follow-up window is still open.

This beats every other workflow I've tried for in-the-moment capture. Notes apps lose things. Email chains get buried. Dedicated card scanners require a separate app with its own database. A message to your CRM bot is immediate and frictionless.

Tips for Getting Clean Extractions#

A few things that make a big difference in extraction quality:

Lighting: Natural light or even indoor overhead light works fine. Avoid flash glare on glossy cards — it obscures text.

Angle: Straight on, not angled. The card should fill most of the frame.

Focus: Make sure the text is sharp. Blurry photos produce unreliable extractions.

Background: High contrast helps. White card on a dark desk or dark card on a light surface. Not card-on-white-tablecloth.

Folded cards or double-sided cards: Send both sides as separate photos in the same message. DenchClaw will merge the data.

Non-Latin scripts: The vision model handles Japanese, Chinese, Korean, Arabic, Hebrew, and Cyrillic reasonably well. If you're regularly working with contacts in a specific language, mention it in your request for better accuracy.

Beyond Extraction: What to Do Next#

Getting the card into the CRM is step one. Step two is the follow-up.

After batch importing from a conference, you can say:

"Draft personalized follow-up emails for everyone I imported from NeurIPS 2026. Reference my notes for each one."

That's where the context notes you added during import pay off. Each email will reference something specific — because you captured it right after meeting them, when it was still fresh.

The business card import workflow isn't just about data entry. It's about making sure the relationships those cards represent actually go somewhere.

FAQ#

What image formats are supported?

JPEG and PNG work great. Most phone camera photos are JPEG. Screenshots, PDFs of digital business cards, and scanned images also work.

Can I import digital business cards (vCard / QR code)?

For QR codes, scan the code with your phone first, then send the extracted contact data to DenchClaw. vCard (.vcf) files can be sent directly and DenchClaw will parse and import them.

How accurate is the field extraction?

For standard printed cards in good lighting: extremely accurate. Complex layouts or low-contrast designs occasionally miss a field. When in doubt, DenchClaw will surface what it extracted and ask you to confirm before saving.

Does it work with cards in other languages?

Yes. The vision model handles most major scripts. For non-Latin cards, mention the language in your message for best results.

What if the card has multiple email addresses?

DenchClaw will extract all of them and label them if the card distinguishes (e.g., "work" vs "personal"). You can then choose which to set as primary.

Ready to try DenchClaw? Install in one command: npx denchclaw. Full setup guide →