References

What was built,
and what came of it.

Our own projects, anonymised: industry, size, region, no client names. Below, scenarios from the German mid-market: the starting situation and the process we build from it.

Tax firm11 employeesMünsterland

Pre-code the receipts, a human still approves.

The situation

Incoming receipts arrived by mail, scan and paper. Pre-capture hung on two people, the DATEV booking only after that.

What we did

  1. Read the receipt, suggest the booking, flag what is unclear
  2. Approval in the existing DATEV chain, no second system
  3. Only the receipts circle connected, the rest of the office untouched

What came of it

  • The morning stack is pre-captured before anyone books in DATEV.
  • Unclear receipts stay put, instead of going through wrong.
  • Around 17 receipts a day. 4.5 hours back each week, live after 5 weeks.

How

DATEV · Receipt AI · Human approval

Trades17 employeesWestphalia

A quote from the enquiry and the catalogue of works.

The situation

Every enquiry was picked up again in the evening. The catalogue lived in Excel, the master typed the quote himself.

What we did

  1. Lay the enquiry against their own catalogue of works
  2. A draft with line items, the master cuts and approves
  3. No new ERP. The process sits before the send.

What came of it

  • The draft is there before the master clocks off.
  • The calculation stays with the master, the typing does not.
  • 9 enquiries a week. 6 hours back each week, live after 3 weeks.

How

Catalogue of works · Quote draft · Approval

Estate agent8 employeesRhineland

A property file from exposé, contract and annex.

The situation

Papers sat in mail, a folder and on the desk. Something was always missing before the notary appointment.

What we did

  1. Incoming papers into one file per property
  2. Read contract and annex, mark missing pages
  3. Not a portal replacement. The file sits before the appointment.

What came of it

  • What belongs to the property sits in one file, not three folders.
  • Missing pages are on the list before anyone drives to the notary.
  • Around 6 property files a month. 7 hours back each week, live after 4 weeks.

How

Property file · Contract · Checklist

Property management23 employeesRuhr area

Owner questions against house rules, minutes and the stock.

The situation

The same questions came every week. The answer sat in one head, in minutes from 2019, or in house rules nobody could find.

What we did

  1. House rules, minutes and stock files into a searchable net
  2. A reply draft to the managers, sending stays with a human
  3. No chatbot facing tenants. Knowledge stays inside.

What came of it

  • Recurring questions no longer need a new letter.
  • The stock knowledge is there when someone is on leave.
  • 21 owner questions a week. 9 hours back each week, live after 6 weeks.

How

Knowledge net · House rules · Minutes

HVAC firm14 employeesEast Westphalia

Pre-sort applications, a human still decides.

The situation

Files arrived by mail and post. Every one was read, including those that did not fit. Management lost Monday morning to it.

What we did

  1. Sort incoming files against the job profile
  2. A short profile for those who fit, the rest as a rejection draft
  3. No autonomous rejections. Management decides.

What came of it

  • Monday morning goes to the conversations, not to every file.
  • Rejections go out after a human has seen them.
  • 4 applications a week. 5 hours back each week, live after 4 weeks.

How

Job profile · Pre-sorting · Approval

More scenarios

Typical starting situations from the German mid-market, from a small trade guild to a company with 300 people, and the process we build from them. Some scenarios bundle several services, because a company rarely has just one problem.

55 scenarios

Freight forwarding and logistics6,000 shipments a monthRhineland-Palatinate

Loading lists, pallet notes and dispatch emails become TMS orders, without retyping.

80 to 120

dispatch emails a day, read before anyone opens the inbox

Before

Shippers send loading lists as Excel, pallet notes as photos and dispatch changes as free text. Planning types it all into the transport management system, often twice, because the change overtakes the first capture.

What runs now

An in-house AI reads list, note and email, creates the order in the TMS and attaches changes to the existing case. What it cannot map stays in a queue for planning.

Service: Custom AI for the companyGoes with it: Email automation and inbox triage

Machine building, service120 employeesBaden-Württemberg

The customer sends a photo, the AI names the spare part, the technician confirms.

Before

Customers send photos of defective parts by email or messenger, often without a type plate. A service technician searches the catalogue, asks back, searches again. Hours pass per enquiry before a quote goes out.

What runs now

An in-house AI, trained on the company's own exploded drawings and spare parts catalogue, identifies component and machine generation from the photo, proposes part number and price and creates the case in the service system. The technician checks and sends.

Service: Custom AI for the companyGoes with it: Customer service and support

Tax advisory firm25 employeesNorth Rhine-Westphalia

Client receipts are sorted, named and filed without leaving the firm.

Section 203

professional secrecy: the model runs in-house

Before

Receipts arrive as photos, scans and mail attachments, often unsorted and without a client reference. Staff sort, name and chase what is missing. A cloud tool is out of the question: professional secrecy under Section 203 of the German Criminal Code.

What runs now

A model on the firm's own server reads every receipt, recognises client, year and document type, names the file to the firm's scheme and files it. Missing receipts are collected per client and requested once a week.

Service: Custom AI for the companyGoes with it: Process invoices and receiptsGoes with it: Email automation and inbox triage

Fire protection planning18 employeesRhineland

GAEB files and fire-protection requirements land in the firm's own item list, pre-costed.

Before

Tenders arrive as GAEB or PDF. Every requirement on escape routes, smoke extraction and proof is mapped by hand to the firm's own work modules. Whoever holds the modules in their head is the bottleneck, and they are often on site.

What runs now

An in-house AI reads GAEB and requirements, maps them to the house modules and flags what is new or unclear. The planner works through the marks before a price goes out.

Service: Custom AI for the companyGoes with it: Automate your quotingGoes with it: Process documents and contracts

Toolmaking, stamping and forming dies35 employeesSwabia

The matching old NC program for a new enquiry is on the desk, without searching the archive.

Before

Before every quote the master searches folders and the machine control for a similar die from years ago. The old NC and the costing from then are the best basis, and both sit in one head.

What runs now

An AI reads drawing and bill of materials, searches the archive for similar tools and lays the old NC program plus the times from then beside it. The master costs on that basis instead of searching first.

Service: Custom AI for the companyGoes with it: Automate your quoting

Technical wholesale60 employees, 40,000 itemsRuhr area

Orders by email, fax and photo are read, checked against the item master and created in the ERP.

40,000

items every order line is checked against

Before

Customers order however they like: email with free text, fax as PDF, photo of a delivery note with the customer's own item numbers. The back office retypes, looks up item numbers and calls back when unclear.

What runs now

An in-house AI reads every order, maps the lines to the company's own items, including customer item numbers and old names, and creates the order in the ERP. Lines with doubt stay pending for approval.

Service: Custom AI for the companyGoes with it: Workflow automation and connecting systems

Law firm12 lawyersHamburg

Contract drafts are checked against the firm's own clause library, on the firm's own server.

Before

Every lawyer keeps their proven clauses in their own files. A counterparty draft is compared with the firm's standard by hand, deviations get noticed or not. Uploading client data to a cloud tool is out of the question.

What runs now

An in-house model knows the firm's clause library, marks every deviation from the firm's standard in the incoming draft and proposes the house wording. The lawyer decides, nothing leaves the firm.

Service: Custom AI for the companyGoes with it: Process documents and contractsGoes with it: Knowledge management and company know-how

Insurance broker30 employeesBavaria

Claims are read, matched to the policy and passed to the insurer complete.

Before

Claims arrive by email with photos, invoices and free text. Staff look up the policy, check what is missing and ask back. The insurer often receives the claim incomplete and asks in turn.

What runs now

An in-house AI reads the claim, finds policy and line of business, checks the documents against the insurer's requirements and drafts the forwarding. If something is missing, a specific request goes to the customer before a human sees the case.

Service: Custom AI for the companyGoes with it: Process documents and contractsGoes with it: Email automation and inbox triage

Workwear, B2B trade12,000 itemsSchleswig-Holstein

A sorted quote with size runs and embroidery positions, without leafing through the catalogue.

12,000

items every briefing is laid against

Before

Sales searches fabric, colour, standard and lead time, types size runs and attaches embroidery or print positions as screenshots. A quote for a municipality or a company takes the afternoon.

What runs now

From briefing and customer master the draft is built: items, size run, finishing, lead time. Sales cuts what does not fit and sends the PDF. Special sizes and tariff customers stay handwork.

Service: Automate your quoting

Plumbing, heating, air conditioning35 employeesLower Saxony

From the call to the quote to the invoice, without anyone typing the same thing three times.

One record

from the call to the invoice, used three times

Before

The call ends up on a note, the quote is typed from the note, the invoice from the quote. The same data three times, three chances for an error, and the phone rings while the master craftsman is on site.

What runs now

A phone agent takes request and address and creates the case. From measurements and photos the quote draft is built from the company's own item catalogue, the master checks. After acceptance the quote becomes the invoice, with approval.

Service: Automate your quotingGoes with it: Voice agents and phoneGoes with it: Process invoices and receipts

IT systems house50 employeesBavaria

Quotes from distributor prices and templates, margin checked, in an hour instead of a day.

Before

A quote for a workplace setup means: look up prices at two distributors, compile items, fill the template, calculate margin. A day of sales time, during which the prices may already be stale.

What runs now

From enquiry and CRM history the quote draft is built, prices come live from the distributor interfaces, the margin is checked against the floor. Sales changes what it wants to change and sends.

Service: Automate your quotingGoes with it: CRM automation

Corporate film and editing18 employeesCologne

Edit notes, social cuts and captions as a workflow: briefing, draft, approval, publishing, labelling.

Before

Every job means: rough cut, client variants, captions, square cuts for LinkedIn. Approvals vanish in mail threads, labelling under Article 50 is done differently by everyone.

What runs now

A workflow takes the briefing, generates an edit list and copy drafts in the client's tone, collects approval in the tool and publishes. Realistic AI images and avatars are labelled. The producers steer instead of copying.

Service: Marketing and content automation

Online retail25 employees, 12,000 itemsSaxony

Product copy, service replies and returns from one knowledge base.

12,000

items with copy from the same source as the service replies

Before

Product texts for 12,000 items are written by hand or copied from the manufacturer. Service answers the same questions on delivery time and returns a hundred times a day, with different answers.

What runs now

A knowledge base of item data, delivery rules and returns process feeds three flows: product copy in the shop's tone, reply drafts for service and returns handling. Service approves what goes out.

Service: Marketing and content automationGoes with it: Customer service and supportGoes with it: Email automation and inbox triage

Bakery, branch supply55 employees, 14 branchesRhine-Main

Branch orders until 2 p.m., without anyone typing into the ERP at night.

until 2 p.m.

order cutoff for the next morning, no night shift in the ERP

Before

Branch managers send demand by WhatsApp, a photo of a note or a call. At night someone in production transfers it into the warehouse system so baking can start at 4 a.m. Number mix-ups come with the territory.

What runs now

A flow reads every order, matches items and quantities against the branch master and creates the order. Anything unclear goes back as a question, the rest is ready for picking before the early shift.

Service: Process automation with AI

Home intensive care55 employeesSaarland

Tour documentation is dictated, structured and approved by the qualified nurse.

Voice instead of keyboard

documentation after the tour, not at the evening computer

Before

After the last tour carers sit in the car or at home typing vitals, measures and handovers. What happened in between has to be remembered, and by the end of the week something is always missing.

What runs now

The nurse speaks into the phone after the visit, the AI structures it to the service's documentation scheme and presents the draft for approval. Nothing is signed off until someone has read it.

Service: Process automation with AI

Daycare operator, 6 sites70 employeesLower Saxony

The shift draft comes from wishes, qualifications and staffing ratios, the manager decides.

Before

The area manager builds the monthly roster by hand: wishes from the group chat, qualified-staff ratio in her head, absences by text. Two days a month, and a sick note upends group coverage.

What runs now

A flow collects wishes in a structured way, knows qualification, group and rest periods and presents a compliant draft. On absences it proposes replacements. The manager decides and publishes.

Service: Process automation with AI

Car dealership with workshop70 employeesHesse

Appointment requests from phone, email and form land in workshop planning, reminders go out by themselves.

Before

Service reception phones, emails and types appointments into the planner while customers wait at the counter. Reminders for inspection and tyre change go out when someone remembers.

What runs now

A phone agent and an email flow take the request, check free capacity in the planner and book. From the vehicle records in the CRM come inspection reminders and seasonal tyre change prompts. Reception looks after the customers on site.

Service: Process automation with AIGoes with it: Voice agents and phoneGoes with it: CRM automation

Municipal housing company3,200 flatsRuhr area

Damage reports are taken and handed to caretaker dispatch in structured form.

also after 5 p.m.

damage reports, without the tape being the first stop

Before

Tenants call, often in the evening. The tape fills up, on the next working day someone listens, asks for flat and damage, types into the software. By the time the caretaker arrives, the water damage is older than it needed to be.

What runs now

A voice assistant takes flat, damage and availability, recognises emergencies and hands over to dispatch. The tenant gets a case number. Goodwill and disputes go straight to a human.

Service: Voice agents and phone

Physiotherapy, 2 sites4 therapists, 3 at receptionBaden-Württemberg

Appointments and prescription renewals run through the phone assistant, reception is there for the patients on site.

Before

Between treatments the phone rings through. Appointment requests, cancellations, prescription renewals, while someone waits at the desk. Whoever cannot get through just shows up.

What runs now

A phone assistant books, reschedules and queues prescription renewals for approval. Medical questions and complaints it puts through at once. It says at the start that it is an AI.

Service: Voice agents and phone

Refrigeration and climate technology32 employeesSaxony

Faults after 6 p.m. are answered, emergencies recognised, the rest booked as appointments.

Before

After closing time the call lands on the tape. Whoever has cooling in a restaurant or a lab calls the next firm that picks up. The business learns in the morning that the job went elsewhere.

What runs now

A phone agent answers, recognises a plant stop and routes to the on-call service, books maintenance and callbacks for the next working day. Management sees a list in the morning instead of a full tape.

Service: Voice agents and phone

PR office9 peopleLeipzig

Every enquiry from the form lands in the CRM, gets a reply and a follow-up date.

0

enquiries left sitting in the inbox

Before

Enquiries from the website land in the shared inbox. Some get answered the same day, some after a week, some never, because a press-date email sits on top.

What runs now

A flow creates every enquiry in the CRM, sends a fitting reply with a call proposal within minutes and sets follow-up tasks. The owner sees a pipeline instead of an inbox.

Service: CRM automation

Management consultancy14 employeesMünster

From the booked first call to the follow-up, everything runs without manual handover.

Before

Whoever books a first call is in the calendar but not in the CRM. Preparation emails, call notes and the follow-up depend on the consultant remembering.

What runs now

The booking creates the contact in the CRM, sends the preparation email, and after the call the dictated note becomes the CRM entry plus follow-up sequence. The consultant consults instead of administering.

Service: CRM automation

Used machinery trade20 employeesNorth Rhine-Westphalia

The field rep dictates the visit in the car, the CRM is filled before he is back.

Before

Visit reports are written in the evening or not at all. What the customer needs, which machine he looks at, when he decides, sits in the salesperson's head. Quotes follow days later.

What runs now

The salesperson speaks the visit into the phone. The AI writes the CRM note, creates tasks and drafts a quote for the discussed machine from stock. The salesperson checks in the evening and sends.

Service: CRM automationGoes with it: Automate your quoting

Care home operator, 4 homes300 employeesNorth Rhine-Westphalia

Applications from five portals in one system, every applicant hears back within a day. Selection stays with people.

Within a day

a reply, instead of after weeks

Before

Applications arrive via five portals, by email and via WhatsApp. The acknowledgement takes days, interview proposals weeks, and good carers have signed elsewhere by then.

What runs now

A flow collects all inputs in one system, sends a personal acknowledgement within hours, proposes interview slots and keeps applicants informed of the status. No AI rates or sorts people: that would be high risk under the AI Act, and it deliberately stays with the nursing management.

Service: Recruiting and applicant management

IT systems house85 employeesWestphalia

The recurring questions to HR are answered by a bot from the company's own knowledge.

Before

Leave remaining, travel expenses, home-office days: the same questions arrive every week by chat at two people in administration who are supposed to be preparing onboarding for new technicians.

What runs now

An internal bot answers from the works agreement, handbook and intranet, citing the source document. What it does not know it passes on. Answers are versioned when a rule changes.

Service: Knowledge management and company know-how

Software house, measurement technology70 employeesFranconia

New staff find processes, contacts and product knowledge in one place.

Before

Onboarding means three weeks of asking colleagues. Whoever knows how a device is parameterised at customer X sits in another team. Documentation exists, spread across wiki, drives and old tickets.

What runs now

A knowledge base bundles wiki, ticket history and product documentation. New colleagues ask in natural language and get the answer with a reference to the document. Senior developers are interrupted less.

Service: Knowledge management and company know-how

Machine building, special machinery90 employeesSaxony

The senior technician's service knowledge is available in-house before he retires.

Before

One service technician knows every machine the company built in thirty years. When an old plant stops, the customer calls him. In two years he is gone, and the knowledge with him.

What runs now

Service logs, circuit diagrams and the senior's experience, recorded in interviews, become a knowledge base on the company's own server. Technicians ask by phone from the customer's site, the answer comes with a reference to diagram and case.

Service: Knowledge management and company know-howGoes with it: Custom AI for the company

Contract logistics110 employeesMannheim

The inbox is read and distributed before a human opens it.

One inbox

three teams that used to read it in sequence

Before

Tour planning, billing and complaints share one inbox. Every email is pulled by hand, and a complaint often lands with planning before billing sees it.

What runs now

An AI reads the inbox, assigns tour, customer and request and routes. The pre-sorting runs in parallel for two weeks before anyone relies on it.

Service: Email automation and inbox triage

Property management15 employees, 2,500 unitsBerlin

Tenant requests are sorted, standard cases pre-answered, tradesman orders created.

Before

The management's shared inbox is the bottleneck: service charge questions, damage reports, notices, complaints, all in one stream. Whoever is off sick leaves a mountain.

What runs now

An AI sorts by request and property, drafts replies to standard questions from the management files and creates the tradesman order for damage. The manager approves and takes care of the cases that need a human.

Service: Email automation and inbox triageGoes with it: Process automation with AI

Family hotels, 3 houses90 employeesBaltic coast

Guest enquiries by chat and email are answered in German, English, Dutch and Polish, reception takes over what matters.

Before

Arrival, dog, half board, cancellation: reception types the same replies while guests want to check in in front of them. Off season in one language, in season in four.

What runs now

An assistant answers standard questions from the hotel information in the guest's language and hands rebooking and complaints over with context. It identifies itself as an AI at first contact.

Service: Customer service and support

Trade association, crafts training22 employeesHanover

Recurring questions on courses, exam dates and membership are answered by an assistant, specialists handle the rest.

Before

Registrations, cancellations, hotel allotments, exam rules: the same questions pile up at the officers, by email and phone. Every answer is individual although the content is not.

What runs now

An assistant on the website and in the inbox answers the recurring questions from the rules, guides and FAQ, with a reference to the source. Specialist questions go to the officer with case history.

Service: Customer service and support

E-commerce agency, Magento16 employeesCologne

Client shop setup takes an hour instead of a day.

an hour instead of a day

shop setup per client

Before

Every new client shop is created by hand: instance, payment methods, shipping rules, legal texts, test order. A day per client, and everyone does the sequence slightly differently.

What runs now

A workflow creates the shop from a template, applies payment and shipping configuration, checks with a test order and reports deviations. The project lead reviews the log.

Service: E-commerce automation and scalingGoes with it: Workflow automation and connecting systemsGoes with it: Email automation and inbox triage

Expert-witness seminars8 employeesHesse

From registration to attendance confirmation, seminar administration runs without Excel.

Before

Eight in-person dates a year, registrations by form and email, Excel, invoicing software, attendance lists by hand, confirmations filled in one by one.

What runs now

The registration creates participant, invoice and list entry, reminds before the date, generates the attendance confirmation after the seminar and asks for feedback. Administration only steps in for special cases.

Service: Workflow automation and connecting systems

Carpentry, roof and timber22 employeesAllgäu

Quote, order and material purchase are one process, not three entries.

Before

The quote is in one program, the order is transferred into a second, sawn timber and fittings typed into the merchant's ordering portal. The same items three times, in the evening by the foreman.

What runs now

From the accepted quote comes the order, from the order the material list, which lands as an order with the merchant. The foreman checks the list once and approves.

Service: Workflow automation and connecting systems

Shutters and sun protection28 employeesRhineland

The site measurement becomes the production order, without notes and retyping.

Before

The fitter measures at the window and writes it down. In the office it is retyped into production, and a transposed digit only shows when the shutter does not fit.

What runs now

The measurement is captured on the tablet, checked against the order and created as a production order. Deviations from the quote are flagged before production.

Service: Workflow automation and connecting systems

Tax advisory network22 employeesOldenburg

Accounts, devices and access for new staff are ready on day one, not in week two.

Before

A new hire means: email to IT, email to HR, email to the mandate lead. DATEV rights, laptop, calendar arrive one after another, and in the first week the new colleague works from a private computer.

What runs now

The signed contract triggers the flow: accounts are created, the device ordered, access granted by role, induction meetings set. On day one the laptop is on the desk with everything on it.

Service: Workflow automation and connecting systems

IT support for pharmacies28 employeesMainz

Alerts from monitoring and inbox become prioritised tickets, without a night shift at the screen.

Before

Till down, scanner, update: alerts, emails and calls are turned into tickets by hand. At night the on-call engineer decides on the phone whether the dispensary can open in the morning.

What runs now

A flow creates the ticket from every source, assigns pharmacy and contract and prioritises by opening hours. The on-call engineer only gets what their level requires.

Service: Workflow automation and connecting systems

Electrical installation, large sites60 fittersBrandenburg

Handwritten daily site logs are read and checked against tariff and allowances.

Before

Sixty handwritings, scanned weekly. Payroll deciphers hours, site and travel and checks every line against the tariff. A week per month.

What runs now

The logs are photographed, an AI reads hours, site and allowances, checks against the rules and flags deviations. Payroll works through the flags instead of piles.

Service: Process documents and contracts

Property management30 employeesHesse

Terms, index clauses and deadlines from scanned old contracts are in the system, not in the binder.

Before

Hundreds of leases sit as scans in the archive, some twenty years old. Whoever wants to know when an index adjustment is possible leafs through. Deadlines are missed because nobody knows them.

What runs now

An AI reads the contracts, extracts term, notice period, index clause and deposit and writes them into the management system. Uncertain values are presented for spot checks, deadlines generate reminders.

Service: Process documents and contractsGoes with it: Custom AI for the company

Painting businesses, group90 invoices a monthCologne

Incoming invoices are captured, pre-coded and presented for approval instead of retyped.

90

invoices a month, pre-coded instead of retyped

Before

Invoices arrive as PDF, paper and photo from three businesses in the group. Accounting retypes header and line data, looks up the cost centre and chases approval by email.

What runs now

Every invoice is read, assigned to business and cost centre, pre-coded and presented to the owner for approval. Accounting posts what is approved and handles exceptions.

Service: Process invoices and receipts

Civil engineering95 employeesMecklenburg

Weigh tickets from the site and invoices from the supplier find each other before posting.

Before

Weigh tickets arrive crumpled from the loader, invoices weeks later from the bulk merchant. Accounting looks for the ticket for every invoice and spots quantity deviations, if at all.

What runs now

Weigh tickets are photographed on site and assigned to the project. When the invoice arrives, the AI matches items and quantities and presents only deviations for review. The rest goes to approval.

Service: Process invoices and receipts

Crafts guild3 people in the officeBrandenburg

Contribution notices, incoming payments and reminders run for three people as if for thirty.

Before

Three people manage the member businesses: writing notices, matching payments, sending reminders, maintaining address changes. In January everything else waits.

What runs now

The flow generates the notices from the member master, reads the bank statements, matches payments and sends reminders in stages. The three see a list of open cases instead of a pile.

Service: Process invoices and receipts

Roofing business40 employeesNorth Rhine-Westphalia

Applicants write via WhatsApp and get a reply within minutes, the master decides.

Minutes

to the first reply, even on Saturdays

Before

Young applicants write via Instagram or WhatsApp, in the evening and at weekends. Whoever hears nothing for two days has signed with the next business. The master sees the messages when he comes off the roof.

What runs now

An assistant replies within minutes, asks trade, experience and earliest start and proposes a short call or a trial day. It rates nobody, it does not sort: who gets invited is decided by the master, from the list, not from a score.

Service: Recruiting and applicant management

Logistics service provider250 employees, 6 languages in the teamHesse

Job ads come from the requirements profile, onboarding material in six languages from company knowledge.

Before

Every job ad is half a day: gather requirements, write, adapt to portals. The onboarding pack for new drivers exists in German, the workforce speaks six languages, and the safety briefing is translated orally.

What runs now

From a structured requirements profile the flow generates the ad per portal and language, HR checks for equal-treatment compliance. The onboarding pack is built per role from company knowledge and translated with versioning. No AI reads or rates applications.

Service: Recruiting and applicant managementGoes with it: Knowledge management and company know-how

Joinery, furniture and interior fit-out14 employeesBavaria

From the customer's sketch, photos and measurements comes the quote draft from the company's own item catalogue.

Before

Customers send a hand sketch, three photos and measurements by email. The master calculates in the evening: material, fittings, hours, from experience and from the last similar job, which he has to find first.

What runs now

The AI reads sketch and measurements, proposes items from the catalogue, pulls material prices from the supplier list and the last similar job as comparison. The master checks the marked lines and sends. The customer lands in the CRM with the case, the follow-up runs by itself.

Service: Automate your quotingGoes with it: CRM automation

Conveyor technology48 employeesEast Westphalia

A customer bill of materials is mapped to the company's own modules, the quote stands in days instead of weeks.

Days instead of weeks

to a quote on a bill of materials

Before

Enquiries arrive as an Excel bill of materials plus a few sketches. Two designers read, map items to the company's own modules, look for old costings, write the quote. Three weeks, during which the competitor has already answered.

What runs now

An in-house model reads the bill of materials, maps every item to a module, pulls the old costing and marks what is new. The designers work through the marks. The list stays in-house, because the customer demands it.

Service: Automate your quotingGoes with it: Custom AI for the companyGoes with it: Process documents and contracts

Municipal utility120 employeesSaxony

Meter readings arrive as photos, are read, checked for plausibility and posted, queries only on deviation.

Before

For the annual reading customers send photos of the meter by email and portal, often skewed, blurred, with reflections. Customer service reads, types and calls when the image is unclear.

What runs now

An in-house AI reads the meter value from the photo, checks it against last year and the consumption profile and posts plausible values. On deviation or an unreadable image a specific query goes to the customer. Photos are deleted after processing.

Service: Custom AI for the companyGoes with it: Customer service and support

Print shop45 employeesNorth Rhine-Westphalia

Customer print files are checked and the query written before a human opens the file.

Every third

print file used to arrive with an error

Before

Every third print file has an error: bleed missing, wrong colour space, resolution too low, font not embedded. Prepress checks by hand and writes the query, often the day before the print date.

What runs now

An AI checks every incoming file against the requirements of the booked product, produces a check report in customer language and writes the query with a concrete note on what to change. Clean files go straight to prepress.

Service: Custom AI for the companyGoes with it: Customer service and support

Metal processing, automotive supplier150 employeesThuringia

Inspection records from handwriting and photos are structured, deviations spotted immediately, nothing leaves the plant.

Before

Inspectors note dimensions on paper records, photograph parts, file them. Quality assurance transfers into tables, trends only show at the audit, and customer drawings must not leave the plant.

What runs now

A model on the company's own hardware reads records and photos, writes measured values into the QA system in structured form, compares with tolerances and reports deviations the same day. Drawings and values stay in the plant, as the customer contracts require.

Service: Custom AI for the companyGoes with it: Process documents and contracts

Gym chain8 studiosNorth Rhine-Westphalia

Posts, newsletters and class announcements come from the class schedule, the studio manager approves.

Before

Eight studios, eight Instagram accounts, one marketing employee. Class changes, promotions and openings are posted per studio by hand or not at all. The newsletter goes out when there is time.

What runs now

Class schedule and promotion calendar feed the flow: posts per studio in its tone, a newsletter draft per month, reminders to members who have not been in for a while from the membership system. Realistic AI images are labelled, the studio manager approves.

Service: Marketing and content automationGoes with it: CRM automation

Industrial supplier, fastening technology200 employees, 6,000 itemsBaden-Württemberg

Data sheets, product copy and translations come from the PIM, in five languages, with a glossary.

6,000

items, each in five languages from the same source

Before

Data sheets are maintained in Word, translations outsourced, product texts in the shop date from three decades. A new item takes weeks until it is cleanly online in all languages.

What runs now

From the structured item data in the PIM the flow generates data sheet, shop copy and short version, translates with a maintained technical glossary per language and presents everything for sample approval. What is not in the data is not claimed.

Service: Marketing and content automationGoes with it: Process documents and contracts

Construction company, building120 employeesBavaria

Site managers ask about standards, house specifications and past sites and get the answer with a source.

Before

How was the detail solved on the last project? Which execution did the client accept back then? The answer is in project files, photos and emails, and the site manager who knows is on another site.

What runs now

Project files, acceptance reports, photos and the company's own execution standards sit in a knowledge base on the company's own server. Site managers ask from the tablet and get an answer with a reference to file and photo. New projects feed the base automatically.

Service: Knowledge management and company know-howGoes with it: Process documents and contracts

Managed service provider35 employeesHamburg

Runbooks and ticket history answer inside the ticket before the technician asks a colleague.

Before

A customer system reports an error a colleague solved a year ago. The solution sits in a closed ticket nobody finds. The technician starts from scratch or calls the senior.

What runs now

Runbooks, documentation and the ticket history per customer form a knowledge base. When a technician opens a ticket, the system proposes the matching earlier cases and steps, with references. The solution flows back into the base automatically after closure.

Service: Knowledge management and company know-howGoes with it: Workflow automation and connecting systems

Supplier, precision parts80 employeesNorth Rhine-Westphalia

When buyers ask ChatGPT for suppliers, the company is in the answer, with a source.

20

buyer questions, one answer page each

Before

The website has described „quality and reliability“ for ten years. Buyers now ask ChatGPT or Perplexity for suppliers for a process, and the answer names three competitors with concrete pages on tolerances, batch sizes and lead times.

What runs now

For the twenty questions buyers ask there is one answer page each with figures, processes and limits, plus structured data, an llms.txt and every page as Markdown. Every month the same questions go to four AI systems and it is logged whether the company is named.

Service: AI visibility, SEO and GEOGoes with it: Marketing and content automation

Staffing agency60 employees, 3 locationsHamburg

Google's AI overview names the agency for the region, and the site itself answers the candidates' questions.

Before

Candidates and clients google „temp work Hamburg nursing“ and get an AI overview at the top listing two chains. The agency's own site has location addresses but no answer on pay, process, contract length.

What runs now

One page per location and profession that answers the most common questions in the first lines, maintained Google business profiles with real reviews, an assistant on the site that answers candidate questions from the same answers and hands applications to a human.

Service: AI visibility, SEO and GEOGoes with it: Customer service and support

Five projects are ours, the scenarios are build plans. Both show what a properly built lever delivers. Which of it applies to you is what the first conversation is for.