How SMEs can successfully get started with AI
Updated on 23 September 2026: a customer example without supporting evidence has been removed, the heading of Scenario 1 no longer states a percentage, and the assumed input values are marked as assumptions. The automations are now listed under their current name, and document analysis runs via Nex in the AI chat. The values in the example calculations are unchanged.
Many SMEs consider AI important, but do not yet use it. Five fully calculated scenarios show how to get started successfully.
What stands out is where the problem usually does not lie: in the technology. Chatbots, knowledge search, document processing and AI telephony are standard building blocks today that can be set up in days. What is missing is the translation of the general insight into a concrete first use case, and that almost always lies where recurring routine work has so far been done by hand.
So why do so many hesitate? The most common reasons we hear in conversations with SMEs:
- “Too complex”: The idea that you first have to hire data scientists and set up projects lasting months
- “Too expensive”: The assumption that AI is only for large corporations with budgets in the millions
- “Data protection”: The concern that business data will end up with US providers
- “Don't know where to start”: No concrete idea of which processes are suitable
All four objections are understandable, and all four can be overcome. This article presents five concrete scenarios that SMEs can implement straight away, with disclosed costs and a transparent calculation. No research project, no data science team. Solutions that are productive in days rather than months.
A note on the ROI figures in this article, before the first of them comes up: the calculation is arithmetically correct and discloses its input values. Of this, only the cost side is supported by evidence. The savings side is an assumption whose derivation is not documented. Anyone who stands behind the figure is arguing with an example calculation, not with a measurement from customers.
Scenario 1: Chatbot for recurring customer enquiries
The classic, and for good reason. Customer enquiries often follow a Pareto pattern: A small number of topics account for the majority of questions. “What are your opening hours?”, “Where can I find the invoice?”, “How do returns work?”, these questions come up hundreds of times, and the answers are in your FAQs, general terms or product documentation.
An AI chatbot that accesses your knowledge base answers exactly these questions, around the clock, in seconds, in the customer's language. Not with rigid if-then rules like first-generation chatbots, but with a genuine understanding of the context.
- Assumption: a customer support employee costs €45,000 per year (salary + non-wage labour costs + workplace), which is around €3,750 per month
- A deflection rate of 80% is also assumed (the share of enquiries that the chatbot resolves without human intervention). No measured value is available for this. Mathematically, this corresponds to 0.8 full-time support employees
- Assumed monthly savings: ~€3,000
- The chatbot itself is free of charge. All you need is a Nexoria Base licence from €29 per user per month. The token costs incurred by the chatbot are covered by the shared token pool system. For 5 users: €145 for the platform
- ROI: 1,969%, and that is not a typo
Besides the cost side, response time matters. An enquiry that lands in the inbox in the evening is traditionally answered the next morning. The chatbot replies in seconds, even at 11 pm, and customers do not have to wait for the next working day.
Implementation: You upload your FAQs, product catalogues and support documentation to the Nexoria knowledge base. The chatbot is embedded on your website as a widget, with one line of JavaScript. Time required until go-live: typically 1-2 working days.
Scenario 2: Knowledge base searches all internal documents
Imagine this: a new employee starts on Monday. They have questions about the travel expense policy, holiday requests, using the CRM system and the data protection training. Normally they ask their team leader, who asks the HR department, which searches a SharePoint folder with 47 subfolders.
In many SMEs, the rules are scattered across Word documents, PDFs, wiki pages and email chains. According to the McKinsey Global Institute report “The social economy: Unlocking value and productivity through social technologies” from July 2012, interaction workers spend 19% of their working time searching for and gathering information. For a 10-person team with an average annual salary of €50,000, that amounts to €95,000 per year, spent on searching for information.
The solution: You upload all internal documents (manuals, policies, process descriptions, training materials) to the Nexoria knowledge base. The AI understands the content semantically and can answer questions in natural language:
- “What is the mileage allowance for business trips?”: The AI finds the relevant paragraph in the 120-page travel policy
- “Who approves holidays of 3 weeks or more?”: The AI knows the approval levels from the HR handbook
- “What steps are required for customer onboarding?”: The AI lists the process from the sales documentation
- Assumed 60% reduction in search time: €57,000 per year for a 10-person team, or €4,750 per month
- Onboarding time for a new employee: assumed to drop from 4 weeks to 2 weeks (halved)
- Costs: 5 Nexoria users on the Base plan = €145/month
- ROI: 3,176%
The biggest advantage is hard to quantify, but enormously important: Consistency. Nex answers from the stored policies and cites the source, regardless of who is asking.
Scenario 3: Document analysis extracts data from invoices
Let us assume that every month the accounts department of an SME receives 200 to 500 incoming invoices . Each invoice has a different format, different fields, different line items. An employee opens each PDF, looks for the invoice number, date, amounts and VAT ID, assigns them, and types the data into the ERP system. Per invoice: 3-5 minutes. For 300 invoices: 15-25 hours per month.
With Nex, the Nexoria AI chat, it works differently:
- Upload: The invoice PDF is uploaded (individually or as a batch)
- Extraction: The AI automatically recognises: invoice number, date, sender, line items, net/gross amounts, VAT rate, VAT ID, bank details
- Template creation: You define a template once, with the fields you want. The AI fills it in automatically for every subsequent invoice
- Export: The extracted data is output as a structured Excel spreadsheet or directly as a journal entry
- Manual time required: 20 hours/month at an hourly rate of €30 = €600/month
- Time required with Nex: 2 hours/month (checks + exceptions) = €60/month
- Assumed savings: €540/month
- Costs: 2 Nexoria users = €58/month
- ROI: 831%
Scenario 4: VoiceBot takes calls outside business hours
One point that many business owners underestimate: a significant share of all calls come in outside business hours. In the evening, at the weekend, during the lunch break. Every unanswered call is potentially a lost customer. If you cannot be reached outside business hours, you often lose the caller to the next provider who picks up the phone.
The classic solution (an answering machine saying “Please leave a message”) has an assumed response rate of under 20% in this example calculation. Many callers hang up and call the next provider.
A Nexoria VoiceBot changes the game fundamentally:
- Natural conversation: The VoiceBot understands free-form speech, no “press 1” menu. The caller describes their request in their own words
- Backed by the knowledge base: The VoiceBot accesses your company documentation and answers questions about products, prices, availability
- Appointment booking: The VoiceBot can enter appointments in your calendar and give the caller a confirmation
- Taking messages: If the request cannot be resolved automatically, the VoiceBot records a structured message: name, callback number, request
- Forwarding: In urgent cases, the VoiceBot can forward the call to an emergency number
- Assumption: 15 missed calls per week outside business hours for a 20-person company
- Of these, an assumed 30% are potential new customers = ~4.5 leads per week, or ~18 leads per month
- With an assumed conversion rate of 20% and an assumed average order value of €1,000: ~€3,600 in potential revenue per month, which is lost without a VoiceBot
- VoiceBot costs: from €249/month (Starter plan). The VoiceBot requires a Nexoria subscription, and customers with Nexoria Pro receive a 10% discount on all VoiceBots. The example calculation uses only the list price of the VoiceBot
- ROI: 1,346%
Particularly relevant for medical practices, trade businesses, law firms and brokers, industries in which availability directly determines whether orders are won. A plumber who takes a call from a customer with a burst pipe on a Saturday evening via the VoiceBot and immediately books an emergency appointment wins that customer for years.
Scenario 5: Automating recurring workflows
Think about your onboarding process for new employees. It often comprises a dozen or more individual steps involving several departments:
- HR draws up the employment contract and sends it for signature
- IT sets up the email account, VPN access and laptop
- The team leader creates the induction plan
- HR schedules the data protection training
- Facility management prepares the workstation
- The mentor is informed and introduces themselves
- IT sets up access to CRM, ERP and internal tools
- HR collects all documents (social security, tax ID, bank details)
In many companies, this is handled through email chains, Excel lists and verbal agreements. The result: tasks are forgotten, deadlines pass, and on the first day the new employee sits in front of a laptop that has not been set up.
With Automations in Nexoria, this becomes an automated workflow:
- Trigger: HR adds a new employee. The automation starts
- Tasks are assigned in parallel: IT receives a task with the deadline “3 days before start”, as does facility management
- Automatic reminders: The person responsible is reminded 48 hours before the deadline. If the deadline is missed, the task is escalated
- Status tracking: A dashboard shows in real time which steps are complete and where things are stuck
- Dependencies: The “Set up CRM access” step only starts once “Email account created” has been completed
- Assumed manual coordination effort per onboarding: ~8 hours (spread across HR, IT, team management)
- With 5 new hires per quarter: 40 hours/quarter = ~13 hours/month
- Weighted hourly rate (HR/IT/team management): ~€40
- Assumed monthly savings: ~€520
- Costs: automations are included in the core product. For this, the example calculation applies a proportionate cost share of ~€60/month . This value is stipulated and cannot be derived from the price list
- ROI: 767%
Onboarding is just one example. The same logic applies to proposal approvals, complaint processes, regular maintenance cycles or compliance checks. Wherever several people have to complete tasks in a fixed sequence, automations take over the manual coordination.
ROI summary
The following overview summarises all five scenarios, using the assumptions above for a typical SME with 20 to 50 employees:
- Scenario 1: Chatbot (customer support): Monthly costs: €145 | Monthly savings: €3,000 | ROI: 1,969%
- Scenario 2: Knowledge base (internal search): Monthly costs: €145 | Monthly savings: €4,750 | ROI: 3,176%
- Scenario 3: Documents in the Nex AI chat (invoices): Monthly costs: €58 | Monthly savings: €540 | ROI: 831%
- Scenario 4: VoiceBot (availability): Monthly costs: €249 | Monthly revenue potential: €3,600 | ROI: 1,346%
- Scenario 5: Automations (recurring workflows): Monthly costs: €60 | Monthly savings: €520 | ROI: 767%
Overall picture: A company that implements all five scenarios invests around €657 per month and generates savings and revenue potential of more than €12,400 per month. That is an overall ROI of more than 1.788 %.
Important: these are assumptions, not measured values. The cost side follows the price list, with the exception of the stipulated cost share in Scenario 5. The savings side is estimated and needs to be filled in for your own company. Secondary effects such as fewer errors, faster response times and higher customer satisfaction are not included.