AI Text Generation Perfion AI Content Guide
Product names, titles and descriptions written by AI from your own product data, with reusable templates, one-job bulk generation and a person approving every text.
AI text generation in Perfion writes product names, titles, headers and descriptions from the data already in the product record. Create with AI passes a prompt, built with feature tags from the item’s own features, to a language model and shows the reply for editing before it is saved. A prompt that works is saved as a prompt template and can be applied to many items in one job. The built-in AI uses OpenAI, with your own API key and default model set in AI Setup.
At a glance
- Perfion name
- AI Integration, AI Setup, Enrichment Templates and Create with AI
- Where
- Create with AI on String and Text features, and AI administration in the web clientCreate with AIPrompt templatesApply TemplateAvailable once an OpenAI API key is provided
- Used by
- Product executives, product data managers and administratorsTemplates are saved by users with the Can manage AI Integration role
- Works with
- FormulasMain gridTranslations
- Documented AI service and models
- One built-in AI service, two models named in the documentation, and AI in Actions and importsOOpenAI4GPT-4-turbo (default)5GPT-5 (images and files)SSELECT.GPT and SELECT.API
How AI text generation works in Perfion, from product facts to approved text
Your own OpenAI key connects the service. Prompts draw on each item’s features, a good prompt becomes a template, and one job applies it to a whole range, with people reviewing what comes back.
Connect your AI service
A user with the Can manage AI Integration role enters the business’s own OpenAI API key and chooses the default model. The built-in AI in Perfion currently supports OpenAI only.
Build the prompt from product facts
The prompt names the item’s features with feature tags, so each product’s own sweep, power or material is filled in before the prompt is sent.
Generate and refine on the item
Create with AI shows the reply from the language model in a result view, where it can be edited. It becomes the feature value only when it is saved.
Save the prompt as a template
A prompt that gives the right tone and format is saved as a prompt template for one feature on one configuration, ready for every user to apply.
Apply it to many items
Apply Template runs the template as one job for all the selected items, filling in each item’s own tags, and the results are ready to review in the grid.
Writing by hand vs Perfion AI: what changes when text is written from the record
By hand, every text depends on a writer, a spec sheet and the time available. In Perfion, texts are generated from the product’s own features with approved templates, and people review them.
| Task | Writing by hand or in a chatbot | Perfion AI text generation |
|---|---|---|
| Product names and titles | Typed one by one, in whatever style the writer prefers. | Generated from the product’s own features with a saved prompt template. |
| Long descriptions | Written from a spec sheet sent by email, which may already be out of date. | Written from the features held on the item, filled into the prompt by feature tags. |
| Facts inside the text | Copied by hand, or left to a chatbot to guess. | Each tag is replaced by the item’s value, and the full prompt can be checked before it is sent. |
| Tone and format | Different for every writer and every prompt. | The same template, so every product of a kind is described the same way. |
| Product photos and datasheets | Looked at by the writer, if there is time. | Images and files such as PDFs can be part of the prompt, with a model that supports them. |
| A new range of 240 products | Weeks of writing before the range can go live. | One job applies the template to each selected item, each with its own prompt. |
| Other languages | Each market writes or prompts in its own language. | One main language, then Perfion’s translation tools for every market. |
| Who controls the prompts | Everyone, in their own chatbot account. | Templates saved centrally, by users with the Can manage AI Integration role. |
See how texts for your own products would be generated in Perfion.
See it on your product data →Perfion AI capabilities: eight ways to write product text from your data
Each capability shows what it does in Perfion and what it is worth to the business.
Operational Reality (Capabilities)
Create with AI generates a value for a feature from a prompt you write. It works on any explicit, single-value String or Text feature, such as a product name, a title, a header or a description. The window has three parts: a prompt editor, a result view and the list of the item’s features. Generating passes the prompt to the language model and shows the reply in the result view. The reply can be edited there, and it becomes the feature value only when the user saves it.
- Any explicit, single-value String or Text feature
- Prompt, result and feature list in one window
- Edited before it is saved
Executive ROI Value
Product executives start from a solid first draft of a name or a description, on the item itself, instead of a blank field and a separate chatbot.
- First drafts in seconds
- No copying between tools
Operational Reality (Capabilities)
The feature list shows every feature in the item’s configuration: on the Product configuration, all the features of a Product. Choosing one inserts a feature tag into the prompt, and these tags work the same way as tags in Perfion formulas. Before the prompt is sent, each tag is replaced by the item’s own value. The prompt can be viewed exactly as it will be passed to the language model, with every tag filled in, so you see what the model will be told.
- Every feature in the item’s configuration
- Tags work like tags in formulas
- The filled-in prompt can be checked first
Executive ROI Value
Texts are written from the facts held in Perfion, not from memory or an old spec sheet, so the description matches the product it sits on.
- Facts taken from the record
- Fewer errors in published text
Operational Reality (Capabilities)
A prompt can include images and other files, such as PDFs, alongside its text. Not every language model supports them. Boyum’s documentation notes that GPT-5 supports images and files, while some older models, such as GPT-4-turbo, do not. So the model in use decides whether a prompt can draw on a product photo or a datasheet.
- Images and files such as PDFs
- Supported by GPT-5
- Not by older models such as GPT-4-turbo
Executive ROI Value
Descriptions can reflect what a photo or a datasheet shows, which helps when a product’s look or its technical document says more than its features.
- Richer descriptions
- Datasheets put to work
Operational Reality (Capabilities)
Getting a language model to reply in the right tone, in the right format and with the right content is often trial and error. Once a prompt works, it can be saved as a named prompt template and reused on other products, and templates can include images and files too. Each template belongs to one feature on one configuration: a template saved for Subheader on Product is only ever offered on the Subheader feature of Products. Saving templates needs the Can manage AI Integration role; other users choose from the templates already saved.
- Saved with a name, reused on other items
- One feature on one configuration
- Saved by role holders, used by everyone
Executive ROI Value
The work of finding a good prompt is done once, and every product of that kind gets text in the same structure and the same tone.
- One brand voice
- No prompt writing for each item
Operational Reality (Capabilities)
When one or more items are selected on a feature that has templates, Apply Template takes the place of Create with AI. The chosen template runs as a job on the application server, which applies it to each selected item individually. If the prompt contains feature tags, they are filled in for each item, so each item gets its own prompt and its own result. If it contains none, the same prompt and the same result go to every item. When the job is done, the new values are ready to review and compare in the grid.
- Runs as a job on the application server
- Tags filled in for each item
- Results reviewed in the grid
Executive ROI Value
A new range of hundreds of products gets its names and descriptions in one run, each written from its own data, so the range is ready to sell sooner.
- Faster range launches
- Writers’ time spent on review
Operational Reality (Capabilities)
Users with the Can manage AI Integration role manage prompt templates centrally as Enrichment Templates. The table shows which templates are available on which features and base features, such as Description on Product. A new template is given a name, the base feature it is available on, the feature it fills, and the template text itself. An example product can be chosen, so the feature tags in the template show real values while it is being written.
- All templates, by feature and base feature
- Name, base feature, feature and template
- An example product for the tags
Executive ROI Value
The business sees and controls every prompt in use, instead of each person keeping their own version in a personal chatbot account.
- Central control of prompts
- Templates kept current in one place
Operational Reality (Capabilities)
Users with the Can manage AI Integration role manage the AI service, its API key and the default model in AI Setup. The built-in AI in Perfion currently supports OpenAI only. The key is provided by the AI service, and its licence is handled between you and that service. The default model for OpenAI is GPT-4-turbo. Create with AI becomes available once a key has been provided.
- OpenAI for the built-in AI
- Your own key, licensed with OpenAI
- Default model: GPT-4-turbo
Executive ROI Value
The business owns its AI account and its usage, and decides which model writes its product text.
- Account under your control
- The model chosen by you
Operational Reality (Capabilities)
AI in Perfion goes beyond Create with AI. SELECT.GPT is an Action command that prompts a language model without any coding, and SELECT.API can be used to integrate with any AI API. The import wizard can also analyse an import file and match it with features in Perfion. All of this arrived with Perfion 2025-R3, together with template-based generation of product names, titles, headers and descriptions.
- SELECT.GPT in Actions, without coding
- SELECT.API for any AI API
- Import files matched with features
Executive ROI Value
AI can be part of automated jobs and imports, not only something people use one field at a time.
- AI in automated jobs
- Quicker imports
5 Core Concepts of AI Text You Can Trust
The first is about prompting. The next two are about templates and scale. The fourth is about review, and the last is about languages.
-
Concept 01 · Facts in, facts out
Build the Prompt from Product Facts
The product data problem it solves
A copywriter pastes a product name into a chatbot and asks for a description. The reply reads well, but it gives the fan a 75 W motor when the product uses 28 W, and the mistake goes live on two marketplaces.
The concept explained
A language model writes from what it is given. Asked about a product name alone, it fills the gaps with general knowledge, and general knowledge is not your product. A prompt built from the item’s own features gives the model the facts it should use. In Perfion, feature tags place each feature into the prompt and are replaced by the item’s values before the prompt is sent, so the model is told the sweep, the motor and the power of this fan, not of fans in general.
Everyday analogy
Briefing a new salesperson. Hand over the product sheet and they describe the product correctly; send them out with only its name and they improvise.
In Perfion
The feature list in Create with AI holds every feature in the item’s configuration, and adding one inserts its feature tag, which works like the tags in formulas. The prompt can be viewed with every tag filled in before it is sent.
Feature tags in the template are replaced by the item’s own values, so the model writes from this product’s facts. -
Concept 02 · Trial and error, once
Turn a Good Prompt into a Template
The product data problem it solves
Five people write their own prompts for product descriptions. One asks for bullet points, one for a paragraph, and one forgets the warranty. The catalogue reads as if five different companies wrote it.
The concept explained
A prompt that gives the right tone, format and content usually takes several attempts. That effort should be made once and kept. A prompt template saves the prompt that works, so every product of that kind is described the same way. Because each template belongs to one feature on one configuration, a description template is offered only on descriptions, and only for the products it was written for. Keeping the right to save templates with a few people keeps the brand voice in safe hands.
Everyday analogy
A company letterhead and standard letter format. Everyone writes their own letters, and every letter still looks like it came from the same company.
In Perfion
Prompt templates are saved with a name and are unique per configuration and per feature. Only users with the Can manage AI Integration role can save templates; other users choose from the templates already saved.
Several attempts, one saved template for Description on Product, offered to every user who works on product descriptions. -
Concept 03 · Many items, one job
Generate in Bulk, Tailored to Each Item
The product data problem it solves
A new range of 240 fans needs descriptions before summer. Writing them one by one takes weeks, and pasting one general paragraph onto all 240 makes every listing the same.
The concept explained
Bulk generation is only useful if each item still gets its own text. In Perfion, a template applied to many selected items runs as one job, and the job fills in the template’s feature tags for each item separately. Each fan’s prompt carries its own sweep, motor and power, so each description is about that fan. A template without feature tags does the opposite: the same prompt, and the same result, on every item.
Everyday analogy
A mail merge. One letter, hundreds of names and addresses, and every letter addressed to the right person.
In Perfion
When several items are selected on a feature with templates, Apply Template is offered. The job runs on the application server and applies the template to each item individually; the new values then show in the grid for review.
One template, one job: each item’s tags are filled in separately, so 240 fans get 240 different descriptions. -
Concept 04 · A draft, not a decision
Let the AI Draft and People Approve
The product data problem it solves
Generated descriptions go straight from the product to the web shop. One of them calls a pedestal fan suitable for outdoor use, which the product team never claimed.
The concept explained
A language model writes fluent text quickly, and fluent text is easy to trust. The business still answers for every claim it publishes. Perfion leaves room for review: in Create with AI the reply can be edited before it is saved, and after a bulk run the new values sit in the grid, ready to be checked and compared. Treating every generated text as a draft keeps the speed and keeps the responsibility where it belongs.
Everyday analogy
A junior writer’s first draft. Quick and often good, but the editor reads it before it goes to print.
From BPAS · prodRecord newsletterprodRecordKeeping the record straight.One product data practice every month, written in plain English for product data teams.In Perfion
The result view holds what will be saved, and the text can be edited first. After an Apply Template job, the generated values show in the grid so users can verify the results and compare the changes.
The AI writes a draft; a product executive corrects and approves it before it is saved and published. -
Concept 05 · Write once, translate with Perfion
Prompt in One Language, Translate with Perfion
The product data problem it solves
Each market team writes its own prompts in its own language. The German description mentions the remote, the Arabic one does not, and nobody can say which text is right.
The concept explained
Language models generally reply in the language they are prompted in: a prompt in English gives English, and a prompt in German, or a line such as “Reply in German”, gives German. Boyum recommends always prompting in the same main language and using Perfion’s translation tools for the other languages. One template in one language keeps the content the same everywhere, and translation, with your glossaries and reviewers, carries it to each market.
Everyday analogy
A head office press release. Written once, then translated for each country, so every market announces the same news.
BPAS method · toolDeploy EnterpriseDesign before configurationThe data model is designed and signed off before anything is configured: the Design phase of toolDeploy Enterprise.In Perfion
Prompts can be written in any language, or can ask for the reply in a given language. Boyum recommends prompting in one main language and handling localisation with Perfion’s translation tools, covered in guide 07.
One English template writes the master text; Perfion’s translation tools carry it to German, French and Arabic.
10 Best Practices for AI Text Generation in Perfion
Each practice says what good AI-generated content looks like and what it costs to ignore it. Setting them up for your team is what BPAS consultants do.
-
01Complete the product facts before generating text
The rule
Texts are generated only for items whose key features are filled in, because a prompt can only use what the record holds.
Risk if ignored
Prompts are sent with blanks where facts should be, and the AI writes vague or invented text around the gaps.
Get it right with BPAS
Not sure this is in place for your team? Talk to a BPAS consultant.
Talk to BPASRelated:Concept 01Guide 06: Validation
-
02Put product facts into prompts as feature tags
The rule
Every product fact in a template comes from a feature tag, so the template stays right when the data changes.
Risk if ignored
Facts typed into a template go out of date, and a bulk run gives every item the same text.
Get it right with BPAS
Not sure this is in place for your team? Talk to a BPAS consultant.
Talk to BPAS -
03Prove each template on example products first
The rule
Each template is proven on a few varied example products, with the filled-in prompt read, before it is applied to many items.
Risk if ignored
A flaw in one template is copied into hundreds of descriptions in a single job.
Get it right with BPAS
Not sure this is in place for your team? Talk to a BPAS consultant.
Talk to BPASRelated:Capability: Enrichment template managementCapability: Prompts built from product data
-
04Keep templates with a small, named group
The rule
Only a few named people hold the right to save and change templates; everyone else uses the approved ones.
Risk if ignored
Templates multiply, each with its own tone, and nobody knows which one is approved.
Get it right with BPAS
Not sure this is in place for your team? Talk to a BPAS consultant.
Talk to BPAS -
05Name every template for what it writes
The rule
Each template name says what it writes and where the text is used, such as “Marketplace description, ceiling fans”.
Risk if ignored
Users pick the wrong template from the list, and a text meant for one purpose is used for another.
Get it right with BPAS
Not sure this is in place for your team? Talk to a BPAS consultant.
Talk to BPASRelated:Capability: Reusable prompt templatesVocabulary: Prompt template
-
06Review generated text before it is published
The rule
Every generated text is read and approved by someone who knows the product before it reaches a channel.
Risk if ignored
An invented claim or a wrong figure is published in every channel, under the brand’s name.
Get it right with BPAS
Not sure this is in place for your team? Talk to a BPAS consultant.
Talk to BPASRelated:Concept 04FAQ 06
-
07Keep the AI account and key owned by the business
The rule
The OpenAI account and its API key belong to the company, with a named owner and an agreed budget for usage.
Risk if ignored
The key sits in one person’s private account, and AI generation stops when that person leaves.
Get it right with BPAS
Not sure this is in place for your team? Talk to a BPAS consultant.
Talk to BPASRelated:Capability: AI setup with your own keyVocabulary: API key
-
08Choose the default model on purpose
The rule
The default model is chosen deliberately, and supports images and files wherever templates rely on them.
Risk if ignored
Templates that depend on a product photo run on a model that cannot read it, such as GPT-4-turbo.
Get it right with BPAS
Not sure this is in place for your team? Talk to a BPAS consultant.
Talk to BPASRelated:Capability: Images and files in promptsVocabulary: Default model
-
09Prompt in one main language and translate the rest
The rule
Templates write in one main language, and the other languages come from Perfion’s translation tools and fluent reviewers.
Risk if ignored
Each language is generated separately, and customers in different markets read different facts about the same product.
Get it right with BPAS
Not sure this is in place for your team? Talk to a BPAS consultant.
Talk to BPAS -
10Keep confidential data out of prompts
The rule
Prompts use only product facts meant for customers, never cost prices, margins or unreleased plans.
Risk if ignored
Confidential figures are sent to an outside AI service and can surface in published text.
Get it right with BPAS
Not sure this is in place for your team? Talk to a BPAS consultant.
Talk to BPASRelated:Vocabulary: PromptCapability: Prompts built from product data
AI content self-audit: how many are in place today?
Tick each of the ten practices above that your product team already follows. Your score updates as you go.
Tick the practices that are in place.
Nothing you tick is stored or sent.
Illustrative scenario: an Indian fan and appliance maker, with every description written by hand
A fan and small appliance maker has about 1,400 products: ceiling, pedestal and wall fans, exhaust fans and air coolers. It sells through dealers across India, on two marketplaces and on its own web shop, and launches two new ranges every year before summer. Orders, stock and invoicing are in the ERP. Names and descriptions are written by two copywriters and an agency, from spec sheets sent by email.
The Operational Challenge
Each new range waits weeks for its descriptions. Copywriters work from spec sheets that are sometimes out of date, so wattages and sweep sizes in the text do not always match the product. Some staff paste product names into a chatbot in their browser, each with their own prompt, and every listing reads differently.
Last summer, a range of 180 BLDC fans went live on marketplaces with one-line descriptions, and 40 listings had to be corrected after customers pointed out the wrong power figure.
- Weeks of writing for each new range
- Facts in the text that do not match the product
- A different tone in every listing
- Prompts kept in personal chatbot accounts
The Financial Transformation
In Perfion, the product data manager builds templates for names, short headers and descriptions on the Product configuration, with feature tags for sweep, motor, power, speeds and remote. Each template is proven on example products and saved, and only two people can change templates. The company’s own OpenAI account and API key are set up, with a default model chosen to suit the templates.
A new range is selected in the grid and the description template is applied in one job, so each fan’s text is written from its own features. Product executives review and approve every text, and the English master goes to the other languages through Perfion’s translation tools.
- New ranges described in days, not weeks
- Facts taken from the product record
- One tone across every listing
- Templates owned by the business
| Cost line | Before (per month) | After (per month) | Difference (per year) |
|---|---|---|---|
| Writing names and descriptions (after: including AI service usage) | ₹ 1,50,000 | ₹ 55,000 | ₹ 11,40,000 |
| Correcting wrong facts in published listings | ₹ 50,000 | ₹ 10,000 | ₹ 4,80,000 |
| Rewriting texts to one tone and format | ₹ 30,000 | ₹ 5,000 | ₹ 3,00,000 |
| Total | ₹ 2,30,000 | ₹ 70,000 | ₹ 19,20,000 |
See product descriptions generated from your own records in Perfion.
Book an AI content review →prodRecord Live
Enterprise product data insights from BPAS
Keeping the record straight, live every month
prodRecord Live is a monthly webinar series on product information management, held on the second Wednesday of every month from January 2027. Each session takes one product data topic and works through it in plain English with time for your questions.
- Jan
- Feb
- Mar
- Apr
- May
- Jun
- Jul
- Aug
- Sep
- Oct
- Nov
- Dec
Who benefits from text written from the record: one record, three answers
-
Promoter & CXOs
How fast can a new range be ready to sell?
Names and descriptions for a whole range are generated in one job from the product record, so the range reaches dealers and marketplaces sooner, with text the business has approved.
-
Product Data Manager
Can AI text stay accurate and on-brand?
Templates take their facts from feature tags, are saved centrally, and only a named group can change them, so every product of a kind is described the same way.
-
Product Executive
Can I get a good first draft without leaving Perfion?
Create with AI writes a draft on the item from its own data, which I edit and save, and approved templates mean I never start a prompt from scratch.
AI Text Generation Vocabulary and Definitions
From the model that writes, to the prompts, templates and settings that keep its text accurate. Each term shows what it means, why it matters, and a household comparison.
-
Language model
AI · engineLarge language model
Operational definition
An AI model that writes text in reply to a prompt, such as OpenAI’s GPT models, Gemini or Copilot.
Why it matters in Perfion
Perfion’s built-in AI uses OpenAI. Not every model reads images and files: GPT-5 does, GPT-4-turbo does not.
Everyday analogy
A fast, well-read writer who knows nothing about your products until you tell them.
See it in action:Concept 01 under Core Concepts →
-
Prompt
AI · instructionText, tags, images, files
Operational definition
Everything passed to the language model: the instruction with its feature tags filled in, and any images or files.
Why it matters in Perfion
The model writes only from what the prompt gives it, so the prompt decides the quality of the text.
Everyday analogy
The brief handed to a copywriter.
See it in action:Prompts built from product data under Capabilities →
-
Create with AI
AI · one itemOne feature, one item
Operational definition
Perfion’s window for generating a value for an explicit, single-value String or Text feature, with a prompt editor, a result view and a feature list.
Why it matters in Perfion
The reply can be edited, and it becomes the feature value only when the user saves it.
Everyday analogy
A draft written straight onto the product label, waiting for your signature.
See it in action:Create with AI under Capabilities →
-
Feature tag
Prompts · dataWorks like formula tags
Operational definition
A placeholder for a feature in a prompt, such as [Sweep], replaced by the item’s own value before the prompt is sent.
Why it matters in Perfion
Tags let one template write a different, accurate text for every item.
Everyday analogy
The blanks in a mail merge letter.
See it in action:Concept 01 under Core Concepts →
-
Prompt template
Templates · reuseManaged as Enrichment Templates
Operational definition
A saved prompt that can be reused on other items, with its tags, images and files.
Why it matters in Perfion
Each template belongs to one feature on one configuration. Only users with the Can manage AI Integration role can save one.
Everyday analogy
A standard letter format every department uses.
See it in action:Concept 02 under Core Concepts →
-
Apply Template
Templates · many itemsA job on the application server
Operational definition
Runs a saved prompt template on several selected items at once, as a job on the application server.
Why it matters in Perfion
Feature tags are filled in for each item. Without tags, every item gets the same result.
Everyday analogy
Printing a mail merge: one template, one run, every letter different.
See it in action:Concept 03 under Core Concepts →
-
AI Setup
Administration · AIService, key and model
Operational definition
Where the AI service, its API key and the default model are managed.
Why it matters in Perfion
The built-in AI currently supports OpenAI only. Create with AI is available once a key has been provided.
Everyday analogy
The main switch and meter for a new electricity connection.
See it in action:AI setup under Capabilities →
-
API key
Setup · accessProvided by the AI service
Operational definition
The key provided by the AI service that lets Perfion use it.
Why it matters in Perfion
Its licence is handled between you and the AI service, so the account should belong to the business.
Everyday analogy
The key to a shared office: whoever holds it decides when the door opens.
See it in action:Practice 07 under Best Practices →
-
Default model
Setup · modelGPT-4-turbo for OpenAI
Operational definition
The language model Perfion’s built-in AI uses by default. For OpenAI, the default is GPT-4-turbo.
Why it matters in Perfion
The model decides what a prompt can include: GPT-5 supports images and files; GPT-4-turbo does not.
Everyday analogy
The default printer: everything goes there unless you choose another.
See it in action:Practice 08 under Best Practices →
-
Can manage AI Integration
Security · roleA user role
Operational definition
The role that gives a user access to AI administration: AI Setup and Enrichment Templates.
Why it matters in Perfion
Only users with this role can save and manage prompt templates; other users apply the templates already saved.
Everyday analogy
The signing authority on a bank account: many can see it, few can sign.
See it in action:Concept 02 under Core Concepts →
AI text generation: questions product teams ask
Short answers, each with a link to the part of this guide that explains it.
-
The built-in AI in Perfion, including Create with AI and prompt templates, currently supports OpenAI only. Beyond that, SELECT.API can be used to integrate with any AI API, and SELECT.GPT prompts a language model from an Action without coding.
More on AI setup -
Yes. Perfion needs an API key provided by the AI service, and the licence for that key is handled between you and the AI service. Create with AI becomes available once a key has been provided. The default OpenAI model is GPT-4-turbo.
More on Practice 07 -
Create with AI works on any explicit, single-value String or Text feature, which covers product names, titles, headers and descriptions. Prompt templates are saved for one feature on one configuration, such as Description on Product.
More on Create with AI -
Language models generally reply in the language of the prompt, so a prompt can be written in that language or ask for the reply in it, for example “Reply in German”. Boyum recommends always prompting in one main language and using Perfion’s translation tools for the other languages.
More on Concept 05 -
Yes. A saved template can be applied to many selected items. It runs as a job on the application server and is applied to each item individually, with its feature tags filled in for each item, so each gets its own text. A template without tags gives every item the same result.
More on Concept 03 -
On a single item, no: the reply can be edited and becomes the feature value only when the user saves it. A bulk job writes its results into the selected items, which are ready to review in the grid once the job is done, so a review step belongs in the process before anything is published.
More on Concept 04 -
Yes. AI text generation is part of Perfion itself, so this guide applies whichever ERP your product data comes from. What changes from one ERP to another is the integration that brings ERP data into Perfion.
More on ERP integration
Still have a question?
Ask about your own products, categories and channels. A BPAS consultant will answer in plain English.
Ask BPASSee your product descriptions written from your data in Perfion
See names and descriptions for your own products generated from their features, saved as templates, applied to a whole range and reviewed by your team. Choose the path that fits the ERP you run.
The world's first Boyum Partner Center of Excellence.Hosted in the cloud and available for evaluation and proof-of-concept work across the Boyum Solutions portfolio.See it to believe it — get convinced by working on it.Explore the PACE sandbox→