AI automation
AI that answers, reads and writes for you, with people in control.
AI automation from MakeAI, our AI practice: assistants on your website and WhatsApp, voice agents, documents read automatically, content at scale and AI in your products.
Already live: EduResurse (UNICEF), Notiqo, FlexMag, ORDO, Hipnosya, Maneos.
Talk to an AI assistant. Mira works for a fictional store. Write to her in Romanian, Russian or English.
What the demo shows
- Sources in viewCatalogue, stock, policies: the sources it used light up under the chat.
- In the customer’s languageRomanian, Russian or English, recognized from the message.
- It books and reservesFree slots from the calendar, a confirmation, a reservation in the CRM.
- It hands over to a personWith a summary of the conversation, so the customer doesn’t have to repeat anything.
Real assistants, built by us
- EduResurseeduresurse.gov.md · the search assistantCase study EduResurse
- ORDO Systemsordo.md · the AI consultant on the home pageCase study ORDO Systems
What we automate with AI. Ten services in four families. Pick one to see what it does, on demo data, and where it already runs.
- Message
- Understanding
- Answer from sources
- Qualification
- Handover
Multichannel AI chat
An assistant we build on your data, on your website, WhatsApp, Telegram, Viber or Messenger. It answers from your own documents and hands over to a person with full context.
Who it is for
- Online stores and retail
- Appointment-based services
- B2B companies
What we measure
- Time to first response
- Assisted resolution rate
- Leads accepted by sales
- Correct handovers
Already running at
Hipnosya
ORDO Systems
EduResurse
- Call
- Identification
- Action
- Confirmation
- Transfer
AI voice agents
Calls answered by a voice agent clearly identified as AI: appointments, routing, common questions, transfer to an operator.
Who it is for
- Clinics and repair shops
- Call centers
- Delivery and logistics
What we measure
- Calls answered
- Valid appointments
- Correct transfers
- Average duration
The voice demo is in preparation: we will publish it once it has been fully tested in all three languages.
Ready-to-start package: After-hours call answering- Ideal profile
- Collection
- Scoring
- Verification
- CRM
Lead generation
Public sources, fit criteria, verification and entry into your CRM — so your sales team talks only to the people who matter.
Who it is for
- B2B companies
- Agencies and integrators
- Exporters
What we measure
- Accepted leads
- Contact rate
- Meetings
- Cost per accepted lead
A demo batch: collection from public sources, scoring against criteria, verification and entry into the CRM — on synthetic data, with no results attributed to a client.
- Event
- Rule
- AI, only if needed
- Action
- Log and alert
Workflow automation
Your apps talk to each other: an event triggers the next steps automatically, and exceptions go to a person.
Who it is for
- Operations and back office
- Marketing and sales
- Management
What we measure
- Reaction time
- Hands-off operations
- Errors
- Hours saved
Already running at
Ready-to-start package: Event-triggered ads- Document
- Extraction
- Validation
- Exceptions
- System
Document and invoice processing
Invoices, contracts and requests: data extracted, checked against rules and sent to your systems; exceptions go to a person.
Who it is for
- Accounting and procurement
- Consulting and professional services
- Legal and administration
What we measure
- Minutes per document
- Field accuracy
- Exception rate
- Processing time
Already running at
EduResurseThe municipal consulting platform (in Romanian)
- Question
- Access
- Search
- Answer
- Feedback
Internal knowledge assistants
Your team asks in its own words and gets answers from company documents, with a link to the source.
Who it is for
- Support and sales teams
- Procedure-heavy companies
- Onboarding
What we measure
- Useful answers with a source
- Time to information
- Unanswered questions
- Usage
Already running at
Ready-to-start package: Internal policy assistant- Catalog
- Style guide
- Generation
- Translation
- Review
AI content at scale
Catalog descriptions, SEO copy and translations produced at scale, following your style rules and reviewed before publishing.
Who it is for
- Online stores
- Distributors and manufacturers
- Multilingual companies
What we measure
- Time per catalog
- Correction rate
- Cost per product
- SEO impact
Already running at
FlexMagThe article on FlexMag’s descriptions (in Romanian)
- Brief
- Scripts
- Variants
- Review
- Scheduling
AI-assisted UGC and video production
From brief to script, video variants, review and scheduling — with approval before anything is published.
Who it is for
- Consumer brands
- E-commerce
- Marketing agencies
What we measure
- Time per asset
- Approval rate
- Cost per asset
- Performance
We don’t have a public case study for this service yet. We show a clearly labeled demo on our own products.
- Discovery
- Prototype
- Testing
- Product
- Maintenance
AI in products and platforms
AI features in existing apps or new platforms: assessment, generation, search, assisted decisions — designed and maintained by the same team.
Who it is for
- Companies with existing platforms
- Startups and new products
- Institutions and organizations
What we measure
- Usage
- Evaluated accuracy
- Time saved
- Stability
Already running at
- Initial assessment
- Exercises on your tasks
- A prototype on one process
- A 30-day plan
- Final assessment
AI training
Hands-on programs for management, teams and institutions, built around your real tasks.
Who it is for
- Leadership
- Sales and marketing
- Operations
- Public institutions
What we measure
- Processes applied after the course
- Adoption after 30 days
The trainers work on projects such as FlexMag, Notiqo and the municipal consulting platform.
AI in production. Our projects, on their real screens. For each: what the AI does, what the rules do and what a person decides.


A national library that fills in and translates itself
For UNICEF and the Ministry of Education: the assistant searches in natural language, and every uploaded PDF is read by AI, which fills in the resource record and translates it.
How it works
- AI understands the question and searches by meaning; fills in the record from the PDF
- AI translates the record between Romanian and Russian
- Rules levels, grades, subjects and versions
- Person a moderator approves every record before publishing
EduResurseeduresurse.gov.md · Claude APICase study EduResurse


Every student gets feedback on every mistake
Tests for the BAC and for grades 9 and 4: closed questions are marked by rules, open answers are assessed by the AI against the mark scheme, and the teacher decides the grade.
How it works
- AI assesses open answers, item by item, with explanations
- Rules mark closed questions instantly, with partial credit
- Person the teacher checks and can change any grade
Notiqonotiqo.io · Claude APICase study Notiqo


Sya, the assistant that sells in three languages
In the Hipnosya store, Sya answers in the customer’s language, recommends from the catalogue and, when a person is needed, sends the request to the panel.
How it works
- AI answers in the customer’s language and recommends from the catalogue
- Person the team picks up, in the panel, the requests that need a person
- A store in Romanian, Russian and English
Hipnosyahipnosya.comCase study Hipnosya


An AI consultant on the home page
Visitors describe their building and get the right services and products from the ORDO catalogue, plus a code for a specialist.
How it works
- AI understands the description of the building and picks from the catalogue
- Person the code reaches an ORDO specialist
- A B2B site in three languages
ORDO Systemsordo.mdCase study ORDO Systems


Product descriptions written with AI, in Romanian and Russian
The catalogue is connected to an AI model: every product gets a description and a meta description, translated into Russian and published in batches on the product page.
How it works
- AI writes the description and meta description, with ChatGPT
- AI translates into Russian, with Amazon Translate
- Rules instructions by category, product selection, publishing in batches
FlexMagflexmag.mdCase study FlexMag


An app that sees what you eat
Maneos scans a meal from a photo, works out the calories and macros and builds workout plans with an AI coach, on iOS and Android.
How it works
- AI recognizes the meal in a photo: calories and macros
- AI generated workouts and an AI coach
- The brand, the app and the site, by the same team
Maneosmaneos.io · iOS and AndroidCase study Maneos
The ads start on their own when hail comes
A flow watches the weather alerts and, in under 2 minutes, starts the campaigns in the counties hit. Here clear rules were safer than an AI model, so we didn’t use one.
How it works
- Rules alert → hail size → budget → 6 campaigns
- Rules a 48-hour pause between activations
- Person the operator is notified in Telegram and can step in at any time
Dent Wolfdentwolf.com · n8n, Google Ads, MetaCase study Dent Wolf
Funding dossiers: a draft in under a minute, validated by a consultant
For a consulting firm that works with town halls: eligibility checked by AI and the dossier generated in the official template, which the consultant validates before export.
How it works
- AI eligibility, the dossier’s sections, new programmes
- Rules budgets consistent across sections, official DOCX and XLSX templates
- Person the consultant edits and validates every dossier
Municipal consultingClaude API · data in MoldovaRead the article (in Romanian) Municipal consulting
Ready-to-start AI packages. Seven processes ready to start, with the duration and what you get fixed in advance.
-
Conversations3 weeks
AI front desk on WhatsApp
Clinics, beauty salons, repair shops and private schools
Answers questions, books appointments and sends reminders, including evenings and weekends.
- Setup
- Integration
- Test and launch
- Assistant on WhatsApp and your website
- Calendar booking
- Automatic reminders
- Handover to a person
-
Operations4 weeks
Invoices straight into 1C
Accounting teams, distributors and companies with hundreds of invoices a month
Invoices from email are read, checked against rules and recorded. A person approves only the exceptions.
- Collection
- Extraction
- 1C integration
- Launch
- Automatic collection
- Field extraction
- Rule-based checks
- Recording in 1C
I want this package: Invoices straight into 1CThe method from the municipal consulting platform
-
Content3 weeks
Auto-written product catalog
Online stores, distributors and importers
Titles, descriptions and attributes in Romanian and Russian, generated from the supplier’s product sheet and checked before publishing.
- Tone and rules
- Test batch
- Publishing
- Tone guide and rules
- Generation in Romanian and Russian
- Automated checks
- Batch approval
I want this package: Auto-written product catalogThe method from FlexMag
-
Operations4 weeks
AI digital library
Schools and universities, libraries, publishers and public institutions
AI reads every uploaded PDF: the record fills itself in and translates itself, and people find the right document by asking in their own words.
- Record structure
- Reading and translation
- Assistant and moderation
- Archive and launch
- Reading the documents
- Records filled in automatically
- Romanian ⇄ Russian translation
- Moderation before publishing
I want this package: AI digital libraryThe method from EduResurse
-
Operations4 weeks
Internal policy assistant
Large teams, retail chains and institutions
Employees ask in their own words and get an answer that cites the paragraph in the policy.
- Inventory and access
- Indexing
- Testing
- Launch
- Document inventory
- Assistant in your tools
- Source references
- Access rights
I want this package: Internal policy assistantThe method from EduResurse
-
Conversations3 weeks
After-hours call answering
Appointment-based businesses, repair shops, clinics and real estate agencies
A voice agent answers when your team can’t, takes down the request and passes it on in the morning, ready to act on.
- Scripts and voice
- Integration
- Test and launch
- Voice agent in Romanian and Russian
- Request capture
- Booking or task
- Summary for your team
-
Growth2 weeks
Event-triggered ads
Local and seasonal marketing: repair shops, hospitality, retail, tourism
Campaigns switch on and off by themselves when a signal comes in: weather, stock, a calendar date or an alert.
- Signal and rules
- Integration and test
- Signal monitoring
- Campaign rules
- Automatic start and stop
- Team notifications
I want this package: Event-triggered adsThe method from Dent Wolf
-
Custom
Your process isn’t here?
We look at it together in a 20–30 minute call and tell you honestly whether AI makes sense and where to start.
You get the price and the monthly costs (AI models, messages, hosting) separately, in the offer, after the first call.
What could you automate? Tick what happens in your company: we show you which AI service fits and start the conversation from there.
Ten situations, one answer
The list of services updates with every tick.
Situations ticked: 0/10
The recommendation shows up here.
What would help
- AI chat on your website and WhatsApp, answering from your data.
- A voice agent that takes the calls your team can’t answer.
- Leads that match your criteria, checked and added to the CRM.
- Documents read automatically, with exceptions sent to a person.
- Flows that connect the apps, with a log and alerts.
- An assistant that answers from the documents, with the source in view.
- A catalogue described automatically, checked before publishing.
- Ads that start on their own when the signal comes, with the budget under control.
- AI features in the product, from prototype to maintenance.
- Training on your tasks, with a 30-day plan.
The matching services show up here as you tick.
Discuss the result with usYour answers stay in your browser; the button only fills in the form’s message.
How we build an AI flow. Seven questions we answer before writing code.
Which repetitive work or decision you want to change. We note how it is done today, so we can measure what “better” means.
From our projects
Municipal consulting: a 15–30 page dossier took 2–3 weeks of manual work.
What data exists, in which format, and who may use it. We decide what reaches the model and what stays with you.
From our projects
EduResurse: the assistant answers only from the platform’s library, linking to the original resource.
Where the AI enters the process: what it receives, what it suggests and where the result goes.
From our projects
FlexMag: descriptions are generated with ChatGPT, translated with Amazon Translate and go into the store.
How we check the result: rules, criteria and tests on real examples, before launch.
From our projects
Municipal consulting: 155 of 156 test cases passed before handover, according to our article.
What the system doesn’t do, and what happens when it isn’t sure: it doesn’t guess, it hands the decision to a person.
From our projects
Dent Wolf: clear rules were safer than an AI model, so the flow runs without AI.
The model’s cost per request, hosting and maintenance, estimated before launch and tracked afterwards.
From our projects
Municipal consulting: one eligibility check costs under €0.15 in model usage.
Who sees, corrects and approves the result, and how every decision leaves a trace.
From our projects
Notiqo: the teacher can change any grade suggested by the AI.
How we start. Three ways in. You move to the next step only once you’ve seen the results.
-
01
Analysis and design
1–2 weeksFor when you want to know exactly what’s worth automating before you invest.
- Process and systems map
- KPIs and success criteria
- Solution architecture
- Estimated implementation and monthly costs
-
Recommended first step
02
Focused pilot
a few weeksOne process, on your data, with KPIs agreed in advance. We continue only if the pilot meets them.
- Everything included in the analysis
- A working solution for one process
- Testing on real cases, with a person checking
- Results report against the KPIs
-
03
Integrated implementation
in phasesMultiple workflows, ERP and CRM integrations, monitoring and maintenance.
- Integration with your existing systems
- Multiple processes and channels
- Monitoring, alerts and a monthly report
- Maintenance and continuous improvement
AI training for teams. On your real tasks, at your office or online. The team leaves with a 30-day plan, not just impressions.
- 01 · Management
AI for decision-making and prioritization
Where AI adds value in your organization, how to choose the first processes and how to govern them.
- Opportunity map by department
- Prioritization criteria and KPIs
- Risks, personal data, internal policy
For: executives, department heads
- 02 · Teams
AI in day-to-day work
Practical use on your own tasks and, above all, checking the results.
- Exercises on the team’s own documents and cases
- Good prompts, reusable templates
- How to check what AI produces
For: sales, marketing, operations, support
- 03 · Implementation
Workshop: from process to pilot
We map a real workflow and build an AI-assisted prototype together.
- Mapping a chosen process
- Working prototype on anonymized data
- Backlog and KPIs for the pilot
For: mixed business + IT teams
What you keep after the training
- Tailored agendaBuilt on the team’s initial assessment.
- Materials and exercisesBased on your cases, reusable after the course.
- Use case listPrioritized by impact and effort.
- 30-day planWhat each participant applies after the session.
- Initial and final assessmentSo you can see progress, not just attendance.
People stay in control. From the start, we design what the system does on its own and what a person decides.
- Answers only from approved sourcesYour catalogue, prices and documents. When it can’t find something, it says so and doesn’t make it up.
- A person decides where it mattersComplaints, large orders, price exceptions: they reach a person, with the full context.
- The voice agent says it is AITransparency is the rule: callers can ask for a person at any time.
- A log for every actionEvery automation has an audit log, error alerts and a named owner.
- Your data stays with youMinimal access, only to the data the process needs. If your rules require it, everything runs on your infrastructure.
- AI only where it makes senseIf clear rules solve the problem, we use them: they are safer and cheaper. Dent Wolf runs without any AI model.
Models, channels and systems
- Claude API
- ChatGPT API
- Amazon Translate
- n8n
- Google Ads
- Meta
- Twilio
- Telegram
- Google Sheets
- National Weather Service API
- AMO CRM
- 1C
- Laravel Queue
- DOCX and XLSX
- WhatsApp Business
- Viber
- Messenger
- Instagram Direct
- Bitrix24
- HubSpot
- SAP
- Google Workspace
- Microsoft 365
- SharePoint
- Notion
- Slack
- Teams
- SIP telephony
- Google Calendar
- Make
- Zapier
Questions about AI automation. Cost, data, errors, languages and control.
Which service should we start with?
Start with the process where the most time goes into repetitive work and the result is easy to measure. If you’re not sure, that’s exactly what the 20–30 minute conversation is for: we’ll tell you honestly where AI makes sense and where it doesn’t.
How much does an AI automation cost?
It depends on the data, the volume and how much human control is needed. Implementation and monthly costs are always listed separately. Monthly costs (model usage, messages or minutes, hosting) depend on volume: we estimate them before launch and report on them monthly. Here is what changes the effort:
What changes the effort
Relative effort: small · A well-defined first version
- Data quality. Clean, structured data means less preparation.
- Request volume. A high volume needs queues, limits and a carefully estimated running cost.
- Validation and human control. Rules, thresholds and review screens.
- Integrations. Where the data comes from and where the result goes: CRM, store, documents.
- Personal data. Anonymization, access and logging.
- Several languages. Each language is tested separately.
Illustrative, no prices. You get the real estimate in writing after the first call.
How long does it take?
A focused pilot usually takes a few weeks; a full platform is planned in phases. The exact timeline is in the proposal, after the analysis.
What if the assistant doesn’t know the answer?
It says it doesn’t know and offers a handover to a person. It never makes up prices, timelines or policies.
Which languages does it work in?
Romanian, Russian and English, with automatic detection of the customer’s language.
What happens to our data?
We use only the data the process needs, with least-privilege access and providers you approve. Where needed, data stays on your infrastructure. Retention rules are set in the contract.
Who owns the solution?
Ownership of the code, configuration, data and AI provider accounts is set out explicitly in the contract. We recommend keeping critical access in your company’s name.
Do you always use AI?
No. If clear rules solve the problem, they are safer and cheaper. Dent Wolf, for example, is automation without AI.
Can we combine several services?
Yes, and most of the time that’s how it works: the chat hands requests over to an automated workflow, and documents go straight into the ERP. The packages in Solutions are exactly this kind of combination, already tested.
Does Google penalize AI-generated text?
Google evaluates how useful content is, not how it was produced. That’s why we start from real product data and avoid generic text.
Which models do you use?
We choose the model by task, quality, cost and data requirements. In our projects we have used Claude for essay assessment (Notiqo), the EduResurse search and the municipal platform’s dossiers, and ChatGPT for FlexMag’s descriptions.
What if the AI gets it wrong?
We design the flow for it: validation rules, confidence thresholds and a person who sees and corrects the result. At Notiqo, the teacher can change any grade suggested by the AI; at EduResurse, a moderator approves every record.
Let’s talk about your automation. Tell us which repetitive work you want to change and what data you have; we reply within one working day, and the first call takes 20–30 minutes.
What happens next
- You send the requestA few lines about what you want to build.
- We get in touchWe clarify the details and book the call.
- The first callGoal, feasibility, next step.
Prefer to reach us directly?
Thank you! We’ve received your request.
We’ll get in touch to clarify the details and book the first call.
Local demo: the request was not sent anywhere.
Dent Wolf Inc
1XFreight
Notiqo
Maneos