These are ideas, not guaranteed features. This article lists enrichments you could build with an AI integration. Not every idea is possible in every AI integration: supported data types, field names, and how results are returned differ between integrations. Before you build one, check that integration's documentation (Background, Mapping Documentation, and Known Limitations) to confirm what it supports.
Overview
AI integrations send an iPaaS.com record to an AI model with an instruction, then bring the result back: a new product description, a cleaned-up customer name, a risk rating on an order, or a plain-language explanation of an error. This article explains how these "recipes" are built and lists ideas grouped by data type.
How a recipe works
Each recipe is mapping configuration, not code. The same integration can run all of them side by side. A recipe is usually made of:
An outbound (FROM iPaaS.com) mapping collection on one data type. Its filter decides which records are worth sending to the AI.
Mappings that tell the AI what to do:
Instruction: what you want the AI to do, usually a static value.
Context (optional): extra text such as your brand voice, an allowed category list, or a formula that lists the order lines.
Fields the AI may change: limits the result to specific fields so nothing else is overwritten.
Where the result goes: back onto the same record, into an iPaaS.com Message that a chat integration (such as Slack or Microsoft Teams) can post, or both.
An inbound (TO iPaaS.com) mapping collection that writes the changed fields and custom fields back to the record.
The exact field names for the instruction, context, and result settings vary by integration. See the integration's Mapping Documentation.
Before you build
Create custom fields first. Many recipes write to custom fields (for example
SEO TitleorAI Segment). Create them in iPaaS.com before you turn the recipe on. Ask for them by name in the instruction.Enrich each record once. Have the recipe write an
Enriched Atcustom field, and filter the outbound collection on that field being empty. Otherwise every edit to the record sends it to the AI again.Protect the fields you are not changing. Some inbound mappings clear fields that are not mapped on every write. Check the integration's Mapping Documentation and preserve the fields the recipe should not touch.
Start small. AI providers charge per use, and results vary. Test a recipe on a handful of records and review the output before running it across your catalog or customer list.
Pick the model for the job. Where the wording is customer-facing (SEO copy, translations, store and category pages, sales emails), a more capable model is often worth the extra cost if the integration lets you choose one.
Keep the AI honest. Tell it not to invent facts that are not in the data, and to leave correct values alone.
Reading the recipe tables
Fields it may change: the standard fields the AI is allowed to update. "None" means it only writes custom fields; "N/A" means the result is a Message, not a record update.
Custom fields written: custom fields you create in iPaaS.com for the result.
Result goes to: Record, Message, or both.
Products and catalog
Recipe | Data type | Example instruction | Fields it may change | Custom fields written | Result goes to |
Product description writer | Product | "Write a 60–120 word customer-facing description from the name, attributes and context. No claims that are not in the data. Plain text." | description | Enriched At, AI Summary | Record |
SEO title and meta description | Product | "Write an SEO title (max 60 characters) and meta description (max 155 characters). Put them in custom fields SEO Title and SEO Description." | None | SEO Title, SEO Description | Record |
Title normalizer | Product | "Rewrite the name as Brand + Product + Key attribute, title case, no ALL CAPS, no internal codes." | name | Enriched At | Record |
ERP shorthand expander | Product | "The description is ERP shorthand (for example 'BLK TEE CTN M'). Expand it into plain words." | description, name | Enriched At | Record |
Attribute extractor | Product | "From the name and description, extract Material, Color, Size and Gender into custom fields of those names. Leave a field out if not stated." | None | Material, Color, Size, Gender | Record |
Category suggester | Product | "Suggest the single best category path from this list: {context}. Put it in custom field Suggested Category." | None | Suggested Category | Record |
Translation | Product | "Translate name and description into French Canadian. Put them in custom fields Name FR and Description FR." | None | Name FR, Description FR | Record |
Compliance check | Product | "Flag wording that makes health, safety or environmental claims (for example 'cures', or 'eco-friendly' without proof). Set custom field Compliance Flag to OK or REVIEW with a reason." | None | Compliance Flag | Record, plus a Message when REVIEW |
Tax class suggester | Product | "Suggest a tax class (clothing, food, general, digital) in custom field Suggested Tax Class. Do not change the tax class." | None | Suggested Tax Class | Record |
Variant title writer | Product Variant | "Write a short customer-facing variant label from the options in context (for example 'Navy / Large')." | None | Variant Title | Record |
Category page copy | Product Category | "Write a two-sentence category page intro for an online store." | description | Enriched At | Record |
Kit or bundle description | Kit | "Describe what the bundle includes and who it is for, from the components in context." | description | Enriched At | Record |
Low-stock explainer | Product Inventory | "Available quantity is below the threshold in context. Write one sentence for the buyer: item, location, quantity, suggested action." | N/A | N/A | Message |
Cost anomaly check | Product Inventory | "Compare cost to the typical cost in context. If it differs by more than 25%, explain in one sentence; otherwise answer 'OK'." | N/A | N/A | Message |
Related-product sanity check | Related Product | "Is this pairing sensible for a shopper (context: both product names)? Answer KEEP or REVIEW with a reason." | N/A | N/A | Message |
Category assignment check | Product Category Assignment | "Does this product (context) belong in this category? Answer YES or NO with a reason." | N/A | N/A | Message |
Customers and companies
Recipe | Data type | Example instruction | Fields it may change | Custom fields written | Result goes to |
Name and company cleanup | Customer | "Fix capitalization and obvious typos in first name, last name and company. Never change the email address or customer number. Leave correct values alone." | first_name, last_name, company | Enriched At | Record |
Segment classifier | Customer | "Classify the customer as Retail, Wholesale, Nonprofit or Government from the data and context (order history summary). Put it in custom field AI Segment." | None | AI Segment | Record |
B2B vs. B2C detector | Customer | "If the company or email domain indicates a business, set custom field Account Type to B2B, otherwise B2C." | None | Account Type | Record |
Duplicate-risk note | Customer | "Context lists customers with similar names. If this looks like a duplicate of one of them, say which and why in custom field Possible Duplicate." | None | Possible Duplicate | Record and Message |
Comment summarizer | Customer | "The comment field holds years of staff notes. Summarize them in two sentences in custom field Notes Summary. Do not change the comment." | None | Notes Summary | Record |
Address standardizer | Customer Address | "Check the address for obvious errors (city/postal code mismatch, missing unit, wrong case). Report a corrected address or 'OK'." | N/A | N/A | Message |
Company industry classifier | Customer Company | "Classify the industry (NAICS sector name) from the name, URL and description into custom field Industry." | None | Industry | Record |
Company profile writer | Customer Company | "Write a one-paragraph account summary for sales from the data. Do not invent facts." | description | Enriched At | Record |
Customer category descriptions | Customer Category | "Write a short description of who belongs in this group and what they get." | description | Enriched At | Record |
Orders and transactions
Recipe | Data type | Example instruction | Fields it may change | Custom fields written | Result goes to |
Order summary | Transaction | "Summarize the order in one sentence for the warehouse: what, how many, anything unusual. Put it in custom field AI Summary." | None | AI Summary | Record |
Fraud or risk flag | Transaction | "From the totals, addresses and lines in context, rate the fraud risk LOW, MEDIUM or HIGH with one reason. Put it in custom field AI Risk." | None | AI Risk | Record, plus a Message when HIGH |
Special handling detector | Transaction | "Read the order notes in context. If they ask for special handling (gift wrap, delivery window, fragile), list it in custom field Handling Instructions." | None | Handling Instructions | Record |
Gift message check | Transaction | "Check the gift message in context for typos or anything inappropriate for printing. Answer OK or a corrected message in custom field Gift Message Checked." | None | Gift Message Checked | Record |
High-value order alert | Transaction | Filter: total > 1000. "Write a one-line alert for the sales team with the customer, total and top items." | N/A | N/A | Message |
Return reason classifier | Transaction (Return) | "Classify the return reason in context as Damaged, Wrong Item, Not as Described, Changed Mind, or Other. Put it in custom field Return Reason." | None | Return Reason | Record |
B2B quote follow-up draft | Transaction (Quote) | "Draft a three-sentence follow-up email for this quote. Put it in custom field Follow-up Draft." | None | Follow-up Draft | Record |
Shipping method copy | Shipping Method | "Write a customer-facing description with the typical delivery time from context." | description | Enriched At | Record |
Payment method label | Payment Method | "Write a short customer-facing label and description." | description | Enriched At | Record |
Gift cards, locations, and people
Recipe | Data type | Example instruction | Fields it may change | Custom fields written | Result goes to |
Gift message polish | Gift Card | "Polish the gift card description into a warm one-line message for the recipient." | description | Enriched At | Record |
Expiring balance nudge | Gift Card | Filter: expiring in 30 days and balance > 0. "Write a friendly reminder for the recipient." | N/A | N/A | Message |
Store page copy | Location | "Write a two-sentence store description for a store locator page." | description | Enriched At | Record |
Region description | Location Group | "Describe the region these locations serve." | description | Enriched At | Record |
Employee name cleanup | Employee | "Fix the capitalization of first name and last name." | first_name, last_name | Enriched At | Record |
Timesheet anomaly check | Timesheet | "Flag shifts longer than 12 hours or notes that mention an incident. Answer OK or a one-line flag." | N/A | N/A | Message |
Messages
Recipe | Data type | Example instruction | Fields it may change | Custom fields written | Result goes to |
Error explainer | Message (errors) | "This is an iPaaS.com transfer error. Explain in plain words what probably went wrong and the first thing to check." | N/A | N/A | Message (to your ops channel) |
Translate alerts | Message | "Translate the text into Spanish." | N/A | N/A | Message |
Digest line | Message | "Rewrite this message as one short line for a chat digest." | N/A | N/A | Message |
Combining with other integrations
AI recipes are most useful in the middle of a flow between two other systems:
ERP → AI → storefront (product recipes): the ERP creates a sparse item, the AI writes the copy, and the storefront (for example BigCommerce or Shopify) receives a publish-ready product. Map the description and SEO custom fields in the storefront's outbound collection.
Order or error → AI → chat (fraud flag, high-value alert, error explainer): the AI's Message is posted by a chat integration such as Slack or Microsoft Teams, so the team gets a readable alert instead of a raw record. See Notification Use Cases: Sending iPaaS.com Events to Chat Integrations.
Storefront customer → AI → CRM (customer recipes): the CRM (for example HubSpot or Salesforce) receives a cleaned, segmented customer.
Gift card → AI → marketing (expiring balance nudge): the Message drives a reminder campaign in a marketing platform through your own flow.