Use cases

What we would buildfor your team.

Agents and apps for the work that still runs on email, Excel and memory. On Salesforce, on your own stack, or both.

Typical scenarios from building materials and manufacturing. Open any row to see how it would work. Each one is a starting point for a 30-minute conversation.

Example scenario

Purchase-order capture

  1. Input:PO as PDF or email
  2. Agent:Reads lines and matches articles
  3. Agent:Checks prices, quantities, dates
  4. Person:Clerk handles exceptions
  5. Result:ERP order created
  • Website & GEO6
  • Sales & marketing4
  • Customer service4
  • Operations & supply chain4
  • Product & back office4
Every use case below is laid out this way.
Salesforce partner since 2018
22 of 22 use casesSalesforce partner since 2018

GEO product website

Website & GEO
Generates a page for every product and system from the PIM, with structured data, full specs and answers to installer questions.StandaloneOpen detailsClose details

Generates a page for every product and system from the PIM, with structured data, full specs and answers to installer questions.

Today

Product data sits in the PIM, but the site shows PDFs and thin category pages that AI assistants cannot quote.

How it works
  1. Input:Product changes in the PIM
  2. Agent:Generates the page with specs and structured data
  3. Agent:Adds FAQ answers and links to systems
  4. Person:Marketing approves
  5. Result:Page live for search and AI

Blue steps are agent work. Outlined steps are decisions a person keeps.

Runs on

Standalone

Connects to

PIM, ERP, DMS, Salesforce

Talk through this use case →

AI visibility monitor

Website & GEO
Puts your buyers' questions to the major AI assistants every week, records who is cited and proposes the content that closes the gap.StandaloneOpen detailsClose details

Puts your buyers' questions to the major AI assistants every week, records who is cited and proposes the content that closes the gap.

Today

Nobody knows what ChatGPT says when an architect asks about your product category.

How it works
  1. Input:Buyer questions per market
  2. Agent:Asks the AI assistants every week
  3. Agent:Records who is cited and what is said
  4. Agent:Proposes content fixes
  5. Person:Marketing decides what to publish

Blue steps are agent work. Outlined steps are decisions a person keeps.

Runs on

Standalone

Connects to

Website, CMS, dashboard or Salesforce

Talk through this use case →

Content freshness agent

Website & GEO
Checks every page against the PIM and current standards, drafts the correction and sends it for approval.StandaloneOpen detailsClose details

Checks every page against the PIM and current standards, drafts the correction and sends it for approval.

Today

Pages still name a withdrawn standard or a discontinued product, and AI assistants repeat it.

How it works
  1. Input:Pages, PIM and standards
  2. Agent:Detects outdated facts
  3. Agent:Drafts the correction
  4. Person:Page owner approves
  5. Result:Page updated

Blue steps are agent work. Outlined steps are decisions a person keeps.

Runs on

Standalone

Connects to

PIM, CMS, standards register

Talk through this use case →

Market localisation agent

Website & GEO
Builds each market version from one source and adapts units, standards, approvals and terminology.StandaloneOpen detailsClose details

Builds each market version from one source and adapts units, standards, approvals and terminology.

Today

The German site is complete. The other markets are partial translations.

How it works
  1. Input:Source page
  2. Agent:Adapts units, standards and terms
  3. Agent:Builds the market version
  4. Person:Local team reviews
  5. Result:Published and kept in sync

Blue steps are agent work. Outlined steps are decisions a person keeps.

Runs on

Standalone

Connects to

CMS, PIM, terminology base

Talk through this use case →

Technical product advisor

Website & GEO
Answers from your datasheets and guides, shows the source, and hands real prospects to sales with the conversation attached.StandaloneOpen detailsClose details

Answers from your datasheets and guides, shows the source, and hands real prospects to sales with the conversation attached.

Today

A planner's question at 9 pm gets a contact form and an answer two days later.

How it works
  1. Input:Visitor asks a technical question
  2. Agent:Searches datasheets and guides
  3. Agent:Answers and shows the source
  4. Person:Sales takes real prospects
  5. Result:Lead in the CRM with the chat

Blue steps are agent work. Outlined steps are decisions a person keeps.

Runs on

Standalone

Connects to

DMS, PIM, Salesforce or CRM

Talk through this use case →

Account-based landing pages

Website & GEO
Builds a page per account from public information for sales to review, and tracks visits back to the account.BothOpen detailsClose details

Builds a page per account from public information for sales to review, and tracks visits back to the account.

Today

Every target account gets the same generic page.

How it works
  1. Input:Target account list
  2. Agent:Reads public company information
  3. Agent:Builds a tailored page
  4. Person:Sales reviews
  5. Result:Visits tracked to the account

Blue steps are agent work. Outlined steps are decisions a person keeps.

Runs on

Both

Connects to

Salesforce or CRM, marketing automation

Talk through this use case →

Reorder and churn agent

Sales & marketing
Learns each customer's ordering rhythm, spots the break early and alerts the rep with the history and a reason to call.BothOpen detailsClose details

Learns each customer's ordering rhythm, spots the break early and alerts the rep with the history and a reason to call.

Today

A regular customer quietly stops ordering and nobody notices until the quarter is over.

How it works
  1. Input:Order history from the ERP
  2. Agent:Learns the ordering rhythm
  3. Agent:Spots the break early
  4. Agent:Alerts the rep with history
  5. Person:Rep calls the customer

Blue steps are agent work. Outlined steps are decisions a person keeps.

Runs on

Both

Connects to

ERP, Salesforce or CRM

Talk through this use case →

Distributor sell-out agent

Sales & marketing
Reads every format, maps products and regions to your master data, and flags drops by dealer, region or product line.BothOpen detailsClose details

Reads every format, maps products and regions to your master data, and flags drops by dealer, region or product line.

Today

Dealers send sell-out reports as Excel, PDF and email text. Nobody consolidates them.

How it works
  1. Input:Dealer reports in any format
  2. Agent:Reads and maps to master data
  3. Agent:Flags drops by region and product
  4. Person:Sales manager reviews
  5. Result:One consolidated report

Blue steps are agent work. Outlined steps are decisions a person keeps.

Runs on

Both

Connects to

Email, ERP, PIM, BI

Talk through this use case →

Competitor watch

Sales & marketing
Monitors competitor sites, price lists and trade press, and sends a short weekly brief to the people who need it.StandaloneOpen detailsClose details

Monitors competitor sites, price lists and trade press, and sends a short weekly brief to the people who need it.

Today

Launches and price changes reach sales by accident, weeks late.

How it works
  1. Input:Competitor sites, prices and press
  2. Agent:Detects changes
  3. Agent:Summarises what matters
  4. Result:Weekly brief to the team

Blue steps are agent work. Outlined steps are decisions a person keeps.

Runs on

Standalone

Connects to

Public sources, Teams or email, CRM

Talk through this use case →

Trade-fair follow-up

Sales & marketing
AI turns scans into clean, enriched contacts with an owner and a personal follow-up draft for review.BothOpen detailsClose details

AI turns scans into clean, enriched contacts with an owner and a personal follow-up draft for review.

Today

Badge scans and business cards sit in a box, and follow-up starts a week late.

How it works
  1. Input:Badge scans and business cards
  2. Agent:Cleans and enriches contacts
  3. Agent:Drafts a personal follow-up
  4. Person:Rep reviews and sends
  5. Result:Contact in the CRM

Blue steps are agent work. Outlined steps are decisions a person keeps.

Runs on

Both

Connects to

Scanner app, Salesforce or CRM

Talk through this use case →

Installer support agent

Customer service
Answers by phone, chat or WhatsApp from your manuals and approvals, cites the source and hands hard cases to an expert.BothOpen detailsClose details

Answers by phone, chat or WhatsApp from your manuals and approvals, cites the source and hands hard cases to an expert.

Today

Installers call with the same technical questions, and the answer depends on who picks up.

How it works
  1. Input:Question by phone, chat or WhatsApp
  2. Agent:Finds the answer in manuals
  3. Agent:Replies and cites the source
  4. Person:Expert takes hard cases
  5. Result:Case logged

Blue steps are agent work. Outlined steps are decisions a person keeps.

Runs on

Both

Connects to

DMS, PIM, telephony, ticketing

Talk through this use case →

Order-status agent

Customer service
Answers order and delivery questions from ERP and carrier data, and warns customers about delays before they ask.StandaloneOpen detailsClose details

Answers order and delivery questions from ERP and carrier data, and warns customers about delays before they ask.

Today

Where is my delivery? takes a service rep three systems and ten minutes.

How it works
  1. Input:Customer asks about a delivery
  2. Agent:Reads ERP and carrier data
  3. Agent:Answers instantly
  4. Person:Service approves delay warnings
  5. Result:Delay warnings sent early

Blue steps are agent work. Outlined steps are decisions a person keeps.

Runs on

Standalone

Connects to

ERP, carrier systems, email, phone

Talk through this use case →

Spare-part identification

Customer service
Identifies the part from a photo, names the current replacement and prepares the order.StandaloneOpen detailsClose details

Identifies the part from a photo, names the current replacement and prepares the order.

Today

A photo of a 15-year-old part needs an expert and a catalogue archive.

How it works
  1. Input:Customer sends a photo
  2. Agent:Identifies the part
  3. Agent:Names the current replacement
  4. Person:Service confirms
  5. Result:Order prepared

Blue steps are agent work. Outlined steps are decisions a person keeps.

Runs on

Standalone

Connects to

PIM, product archive, ERP, webshop

Talk through this use case →

Complaint intake

Customer service
Collects everything up front, classifies the case, checks warranty and batch, and routes a complete file.BothOpen detailsClose details

Collects everything up front, classifies the case, checks warranty and batch, and routes a complete file.

Today

Complaints arrive with missing photos, order numbers and batch data.

How it works
  1. Input:Complaint arrives
  2. Agent:Asks for missing photos and data
  3. Agent:Checks warranty and batch
  4. Agent:Classifies and routes
  5. Person:Quality or service handles it

Blue steps are agent work. Outlined steps are decisions a person keeps.

Runs on

Both

Connects to

ERP, QMS, ticketing or Salesforce Service

Talk through this use case →

Purchase-order capture

Operations & supply chain
Reads each PO, matches articles, checks prices and dates, and creates the ERP order. Unusual cases go to a person first.StandaloneOpen detailsClose details

Reads each PO, matches articles, checks prices and dates, and creates the ERP order. Unusual cases go to a person first.

Today

Customer POs arrive as PDFs and emails and are typed into the ERP by hand.

How it works
  1. Input:PO as PDF or email
  2. Agent:Reads lines and matches articles
  3. Agent:Checks prices, quantities, dates
  4. Person:Clerk handles exceptions
  5. Result:ERP order created

Blue steps are agent work. Outlined steps are decisions a person keeps.

Runs on

Standalone

Connects to

Email, ERP, PIM

Talk through this use case →

Supplier confirmation matching

Operations & supply chain
Compares every confirmation with the PO and flags deviations in price, quantity or date the day they arrive.StandaloneOpen detailsClose details

Compares every confirmation with the PO and flags deviations in price, quantity or date the day they arrive.

Today

Confirmations are filed, not checked, and deviations surface at goods receipt.

How it works
  1. Input:Supplier confirmation arrives
  2. Agent:Compares it with the PO
  3. Agent:Flags price, quantity, date deviations
  4. Person:Purchasing decides
  5. Result:ERP updated

Blue steps are agent work. Outlined steps are decisions a person keeps.

Runs on

Standalone

Connects to

Email, ERP purchasing

Talk through this use case →

Certificate and compliance watch

Operations & supply chain
Tracks every certificate per product and market, warns before expiry and answers customer document requests.BothOpen detailsClose details

Tracks every certificate per product and market, warns before expiry and answers customer document requests.

Today

Declarations and approvals expire unnoticed, and nobody can find the current version.

How it works
  1. Input:Certificates per product and market
  2. Agent:Tracks expiry and gaps
  3. Agent:Warns the owner early
  4. Person:Quality approves document replies
  5. Result:Customers get the current version

Blue steps are agent work. Outlined steps are decisions a person keeps.

Runs on

Both

Connects to

DMS, PIM, approvals register

Talk through this use case →

Production planning assistant

Operations & supply chain
Proposes a revised plan in minutes, shows which orders move and why, and lets the planner decide.StandaloneOpen detailsClose details

Proposes a revised plan in minutes, shows which orders move and why, and lets the planner decide.

Today

An urgent order or a machine breakdown sends planners back to Excel.

How it works
  1. Input:Urgent order or breakdown
  2. Agent:Re-plans in minutes
  3. Agent:Shows what moves and why
  4. Person:Planner decides
  5. Result:Plan updated in ERP or MES

Blue steps are agent work. Outlined steps are decisions a person keeps.

Runs on

Standalone

Connects to

ERP, MES, materials planning

Talk through this use case →

PIM enrichment

Product & back office
Fills missing attributes from technical documents, checks consistency and sends proposals to product management.StandaloneOpen detailsClose details

Fills missing attributes from technical documents, checks consistency and sends proposals to product management.

Today

Product attributes are missing or inconsistent across languages.

How it works
  1. Input:Technical documents
  2. Agent:Extracts missing attributes
  3. Agent:Checks languages and variants
  4. Person:Product management approves
  5. Result:PIM updated

Blue steps are agent work. Outlined steps are decisions a person keeps.

Runs on

Standalone

Connects to

PIM, DMS

Talk through this use case →

Standards and regulation watch

Product & back office
Monitors standards, matches each change to the affected products and alerts the owner.StandaloneOpen detailsClose details

Monitors standards, matches each change to the affected products and alerts the owner.

Today

Changes to DIN, EN and building rules are noticed late, often by a customer.

How it works
  1. Input:Standards and regulation sources
  2. Agent:Detects changes
  3. Agent:Matches affected products
  4. Result:Owner alerted with the documents

Blue steps are agent work. Outlined steps are decisions a person keeps.

Runs on

Standalone

Connects to

Standards sources, PIM, DMS

Talk through this use case →

Receivables agent

Product & back office
Prioritises open items by risk and relationship, drafts reminders in the right tone and flags cases for a personal call.BothOpen detailsClose details

Prioritises open items by risk and relationship, drafts reminders in the right tone and flags cases for a personal call.

Today

Dunning treats a 20-year key account like a one-off buyer.

How it works
  1. Input:Open items from the ERP
  2. Agent:Scores risk and relationship
  3. Agent:Drafts the reminder in the right tone
  4. Person:Finance sends or calls
  5. Result:Status tracked

Blue steps are agent work. Outlined steps are decisions a person keeps.

Runs on

Both

Connects to

ERP, Salesforce or CRM

Talk through this use case →

Company knowledge assistant

Product & back office
Answers employee questions from approved sources, with a link to the document and respect for access rights.StandaloneOpen detailsClose details

Answers employee questions from approved sources, with a link to the document and respect for access rights.

Today

Answers live in SharePoint, old emails and the heads of a few colleagues.

How it works
  1. Input:Employee asks a question
  2. Agent:Searches approved sources
  3. Agent:Answers with a link to the document
  4. Result:Access rights respected

Blue steps are agent work. Outlined steps are decisions a person keeps.

Runs on

Standalone

Connects to

SharePoint, Teams, DMS, ERP

Talk through this use case →

GEO. GEO means that AI assistants such as ChatGPT or Perplexity find your products, understand them correctly and name them in their answers.

Questions

AI agents and custom applications

What can an AI agent do in a manufacturing or building-materials company?

It takes over work that now runs on email, Excel and memory: reading purchase orders, answering installer questions from your manuals, checking supplier confirmations, tracking certificates. The agent reads, matches and drafts. A person decides where it matters. The library above lists typical scenarios for sales, service, operations and the back office.

Should an AI agent run in Salesforce or as a standalone application?

Where the work happens decides it. If your team already works in Salesforce and the data sits there, we build inside your org. If your users have no Salesforce licences, or the process sits far from the CRM, a standalone application fits better, connected to ERP, PIM, DMS and, where needed, Salesforce. Many scenarios work either way. We settle it in the first conversation.

Does an AI agent act without human approval?

A person decides where it matters. The AI proposes, and your team approves, changes or rejects. When an agent writes to a system of record or contacts someone outside the company, a person confirms first, and every answer shows its source. A few agents only read and inform, such as a weekly brief on competitors.

How do I know which of my processes suits an AI agent?

Look for work that arrives in many formats, gets typed in by hand or depends on one colleague's memory, and follows rules someone can explain. Order entry, technical questions and certificate tracking are typical. In a free 30-minute conversation, one of our AI architects can tell you whether your case fits.

Which systems can an AI agent connect to?

The scenarios above connect to systems that building-materials and manufacturing companies already run: ERP, PIM, DMS, email, ticketing, MES, SharePoint, and Salesforce or another CRM. The agent works with the data where it sits. Which connections your case needs is one of the first things we clarify in the 30-minute conversation.

Your use case

Not on this list? Those are the interesting ones.

30 minutes, free of charge, with one of our AI architects. We work out whether it is a case for Salesforce or for your own application.

Talk through your use case30 minutes, free of charge