AI integration for SMEs · Routine accelerator
AI doesn't replace expertise - it multiplies it. Our philosophy: start with the process, choose the tool by ROI, write the SOP, train the team, measure savings. Not marketing tools, not the latest shiny model - the tool that solves the concrete task. ChatGPT, Claude and Gemini for cognitive tasks. Copilot for office workflows. Make and n8n for automation. Custom GPTs for domain expertise. All with Enterprise privacy plans where needed. The goal: 30-50% less time on routine work, free time for real value. Typical clients: law firms handling 200+ contracts monthly, accounting practices with 100+ clients, marketing agencies, SMEs with 10-150 employees, professional firms (doctors, architects, engineers) without internal IT, ministries and local governments in digital transformation. Industries covered: legal, accounting, marketing, construction, tourism, finance, retail, HR, education, healthcare. Geographically the Adria region (Montenegro, Serbia, BiH, Croatia), EU clients via remote audit, DACH companies via German processing, GCC via Arabic processing. Standards: GDPR compliance, DPA signing with each AI vendor, ISO 27001 principles for data access, EU AI Act readiness.
What you get
- 30-50% less time on routine tasks within 90 days
- SOPs the team actually uses and doesn't discard within a week
- Tools chosen by ROI, not hype
- No vendor lock-in - models and platforms swappable
- Privacy structure - Enterprise plan where data is sensitive
- GDPR compliance in every process - DPA signed, data categories mapped
- Team training that doesn't get discarded - practical exercises, not PowerPoint
- Real savings measured pre/post with your number, not our estimate
What we do
AI audit & process mapping
1-2 days workshops with the team. Mapping all processes, identifying spots where AI multiplies (cognitive work, routine, documents, support, translation, classification, content generation). Prioritization via ROI matrix (impact × ease of implementation). Output: written report 20-40 pages with proposals, savings range, rollout plan, tool list with licensing options.
Tool & platform selection
ChatGPT (OpenAI) Enterprise/Team/Plus, Claude (Anthropic) for Work/Pro, Gemini (Google) Workspace, Copilot (Microsoft) Business/Enterprise for cognitive tasks. Make, n8n, Zapier for automation. Enterprise vs Team vs individual plans by GDPR/DPA readiness. No vendor lock-in - always an alternative. Specialized tools (Notion AI, Otter.ai, Fireflies, Granola, Stenograph) for vertical needs.
Custom GPTs & AI assistants
Domain expertise packaged into assistants - sales, customer support, contract analysis, finance, HR onboarding. Model training on internal documentation (SOPs, product catalogs, contract templates, FAQ databases). Not replacement, but first layer freeing experts for hard cases. Typically 4-6 weeks from kick-off to production assistant.
Workflow automation
Make and n8n scenarios - email→CRM, forms→spreadsheet, documents→folder, alert→Slack/Teams, invoices→accounting, lead→qualification→sales. Integrations with existing tools (Notion, Airtable, Google Workspace, Office 365, HubSpot, Salesforce, Pipedrive, Calendly, Stripe). No-code/low-code approach where possible.
Team training
Practical 1-on-1 and group workshops. Prompt engineering basics and advanced techniques (chain-of-thought, few-shot, role prompting). Tool use in daily work with real examples from client industry. Ethics and safety (what can be sent to AI, what can't, GDPR data categories, how to recognize AI hallucination). Typically 2-3 days per cohort with follow-up sessions.
SOPs & internal documentation
Written standards - how a tool is used for a specific task. Prompt templates (by department, by use case). Errors to avoid (privacy leak, hallucination, vendor lock-in). Decision procedure for when AI yes, when no. SOP versioning - technology changes faster than documentation, so we have quarterly review cycles.
AI for marketing & content
Generating articles, social media posts, images (DALL-E 3, Midjourney, Stable Diffusion), video scripts, podcast previews. Localization to 9 languages (LAT/CIR Serbian, English, Russian, Chinese, Arabic, Turkish, German, Hebrew). Brand voice consistency through fine-tuning and style guide implementation. Result measurement via analytics (engagement, conversion, time-saved).
AI for sales & customer support
Lead qualification (BANT - Budget Authority Need Timeline), follow-up sequences (cold email, LinkedIn, WhatsApp), FAQ assistant on website, chatbot integrated with CRM. Reduced response time from hours to minutes. Escalation to experts only when needed (AI recognizes the boundary of its confidence).
AI for finance & analytics
Report automation (PnL summary, cash flow projection, sector analysis), anomaly detection in transactions, scenario projections (baseline/upside/downside). Forensic analysis of larger datasets through AI tools (pattern recognition, outlier detection, correlation analysis). Model and source verification - AI proposes, expert verifies.
Privacy & security
Enterprise privacy plans (ChatGPT Enterprise, Claude for Work, Copilot Business). DPA signing with each vendor, GDPR compliance, data categories allowed in tools (public, internal, confidential, sensitive personal, regulatory-protected). Local LLM (Llama 3, Mistral, Qwen) where essential for data that mustn't leave EU. EU AI Act compliance scoping and risk assessment.
AI-assisted contract review (legal tech)
Pre-screening contracts for law firms and in-house legal teams. Custom GPT trained on client contract library (template contracts, past reviews, most common errors). Automatic identification of problematic clauses (auto-renewal, indemnification asymmetry, unlimited liability, broad IP assignments), negotiation recommendations, redlining draft. Cooperation with specialized legal tech tools (LawGeex, Kira Systems, Spellbook) where justified. Typically 60-75% reduction in time on first contract analysis.
Predictive cash-flow modeling
AI-enhanced cash flow projection models with scenario analysis (baseline/upside/downside) and Monte Carlo simulation. Integration with ERP/accounting systems (Pantheon, Synergy, SAP B1, Microsoft Dynamics, Calculus) for real-time data feed. Anomaly detection in current transactions, projection vs actual variance tracking, owner alerts when cash position crosses threshold. Typically for SMEs 1-15M EUR revenue with complex cash flow cycles.
Customer churn analytics for SMEs
Predicting which customers will leave based on CRM data patterns. Risk score per customer (high/medium/low churn risk), root cause analysis (price, service quality, contact frequency, competitor pressure), retention strategy by segment. Cooperation with CRM systems (HubSpot, Salesforce, Pipedrive). Typically for B2B SaaS, professional services, retail/e-commerce firms - where churn cost is 5-25x acquisition cost.
Document OCR + classification
Automatic processing of incoming documents (invoices, contracts, reports, email attachments) - OCR for PDF/scan, classification by type and content, extraction of key data (amounts, dates, parties, invoice number), routing to appropriate folder or system. Cooperation with Microsoft Azure Document Intelligence, Google Document AI, ABBYY FineReader. Typically for accounting practices with 100+ clients, law firms with high document volume, local governments in public administration digital transformation.
AI personalization for e-commerce
Personalized product recommendations, dynamic pricing, customer segmentation, abandoned cart recovery with AI text. Integration with Shopify, WooCommerce, Magento. A/B testing of different personalization strategies with Statistical Significance thresholds. Typically for e-commerce firms 500K-10M EUR annual GMV where conversion rate optimization brings direct ROI. Brand voice consistency through fine-tuned LLMs.
How we work
- Audit - 1-2 days with the team (remote or on-site). Process mapping through structured interviews with key personnel. Opportunity identification via ROI matrix. Prioritization based on impact + ease. Written report with concrete proposals and savings range within 7 days.
- Pilot - 30 days. Implementation of 2-3 highest-priority tasks. Time measurement before and after with exact numbers (not estimates). Prompt and SOP adjustment through iterations. Pilot exit: you know real savings, not estimates. Decision whether to roll out or not.
- Rollout - 60-90 days. Extension to the rest of the team. Group (cohort) and 1-on-1 training for senior users. SOP writing for all identified use cases. Usage tracking via analytics and intervention where team doesn't adopt (resistance management, additional workshops, mentorship).
- Optimization - Continuous via retainer or quarterly review. Monthly review of usage and savings metrics. New use cases as the team opens to possibilities. Tool replacement where better arrives on market. No lock-in to one vendor - strategy is 2-3 tools in combination.
Pricing
Price is set after we discuss the matter and scope the work. AI audit from 1,500 EUR. 30-day pilot from 4,500 EUR. Complete rollout (audit + pilot + training + SOPs) from 8,500 EUR for firms up to 20 employees. Retainer for continuous optimization from 850 EUR monthly. No hourly billing, no hidden costs.
Why Schule
ROI measured, not estimated
Pre-implementation measurement of time on routine tasks. Post-implementation measurement of the same tasks. Your number, not our estimate. Typically 30-50% savings in 90 days - but that's a range, not a promise. Measurement is part of the standard mandate.
No vendor lock-in
SOPs written generically with modular prompt structure. Prompts portable between ChatGPT, Claude, Gemini. Make/n8n scenarios can be migrated between automation platforms. Strategy: 2-3 tools in combination, not lock-in to one ecosystem.
GDPR compliance from day 1
DPA signing with each AI vendor. Mapping data categories (public, internal, confidential, sensitive personal, regulatory-protected). Enterprise plans where data is sensitive. Local LLM (Llama 3, Mistral) for data that mustn't leave EU. EU AI Act ready.
Industry-specific expertise
We don't approach every industry the same. Law firms get different treatment than marketing agencies, accounting practices differ from engineering firms. We have template SOPs for 12 industries with over 200 use cases mapped.
Training that sticks
Practical workshops with real examples from client industry, not generic PowerPoint. We measure adoption via analytics (what % of employees uses the tool regularly). Typically 85%+ adoption rate in 90 days, which is 3-4x the industry average.
9 languages operational
Content localization to 9 languages (LAT/CIR Serbian, English, Russian, Chinese, Arabic, Turkish, German, Hebrew). Brand voice consistency through fine-tuning and style guide. 30% of our clients have multi-language needs - our advantage is covering it in-house.
AI routine accelerator for our other services
We're primary users of AI tools in our own expertise, capital, EU funds, and China services. Forensic analysis of hundreds of thousands of transactions, business plan variant generation, EU documentation translation, briefing for Chinese partners - all with AI. We learn on our own work before learning on yours.
Questions about AI integration
- The team fears AI will replace their jobs.
- Fair question. Our position: AI replaces routine cognitive work, not expertise. A team using AI works better. A team not using it loses market. Training includes this conversation - with concrete examples of what AI does well (drafts, summaries, classification) and what it doesn't (judgment, empathy, accountability).
- What about confidential data?
- Enterprise plans (ChatGPT Enterprise, Claude for Work) have DPAs, don't train models on your data, can be hosted in the EU region. Local LLM (Llama 3, Mistral, Qwen) for the most sensitive data that mustn't leave the network. Strict SOPs on which data categories go where. GDPR compliance documentation included.
- ROI - how big realistically?
- Typically 30-50% less time on routine tasks within 90 days per pilot. We measure before and after - your number, not our estimate. Fastest ROI we see in documentation processes (contracts, reports), translation, classification, email communication. Slower ROI in creative work and strategic thinking (where AI serves as sparring partner, not replacement).
- What about vendor lock-in?
- None. SOPs written generically with modular prompt structure. Prompts portable between ChatGPT, Claude, Gemini, Copilot. Make/n8n scenarios can be migrated between automation platforms. Strategy: 2-3 tools in combination, not lock-in to one ecosystem.
- Do we need an internal AI team?
- Smaller firms (up to 20 employees) - no. External retainer covers it. Medium firms (20-100) - internal champion useful, our retainer mentors them. Larger firms (100-500) - small internal team (1-2 people) with us as senior advisor mandate. Large firms (500+) - internal AI team with us as quarterly audit and strategic direction.
- What's the difference between ChatGPT, Claude, Gemini and Copilot?
- ChatGPT (OpenAI) - strongest in coding, images (DALL-E 3), general tasks. Claude (Anthropic) - best for long documents, analysis, ethics. Gemini (Google) - integrated in Google Workspace, strong in multimodal and search. Copilot (Microsoft) - best for Office (Word, Excel, PowerPoint) and enterprise workflows. We typically recommend combinations of 2-3, not one tool for all.
- Is AI safe for a law firm/accounting practice?
- With Enterprise plan and DPA - yes. Without Enterprise plan, no. Privacy is a technological-contractual question, not a marketing argument. We draft DPA documents, map data categories, set up SOPs. Professional liability doesn't allow client data to go into public ChatGPT without DPA.
- How does AI help in financial forensics?
- Pattern detection across hundreds of thousands of transactions in a minute - phantom suppliers, circular flows, timing anomalies in payments, correlations between accounts that mustn't be correlated. Manually 100+ hours, with AI 1-2h. AI doesn't draw conclusions - it generates hypotheses the expert verifies. We use this in our own expert work practice.
- How do you approach the EU AI Act (from 2024+)?
- EU AI Act entered into force in 2024 with risk tiers: minimal, limited, high-risk, prohibited. We do AI system risk assessment per Article 6 criteria, map obligations (technical documentation, data governance, human oversight, transparency, accuracy) and implement compliance framework. Special attention to high-risk systems (HR, credit, education scoring, law enforcement) where compliance has deadlines from February 2025. Client has paper trail for EU regulatory audit.
- What should a client prepare before our conversation?
- Brief description: what you do, employee count, key processes where you see most routine (5-10 concrete examples). Existing tools (ERP, CRM, document system, automation platforms). Industry and data types you handle (public, internal, confidential, sensitive personal, regulatory-protected). Goal - what savings would you consider success (time, quality, competitive). Willingness for 1-2 day workshop with key personnel for audit.
- Delivery format for AI mandates?
- Audit report: PDF 20-40 pages with process map, ROI matrix, tool proposal and SOPs. Pilot documentation: technical setup documentation, prompts library, measurement dashboard, training materials. Rollout deliverables: SOP per use case (1-3 pages each), template prompts, n8n/Make scenarios in exportable form, training video recordings. All in editable format (Google Workspace, Notion) so client can modify after mandate close.
- Do you accept rush AI implementations?
- Yes. Rush audit in 5 days instead of standard 2 weeks with 30-50% premium. Rush pilot in 14 days instead of 30 (shortened measurement, faster feedback loop). Largest rush we did: complete AI rollout (audit + pilot + training + SOPs) for marketing agency that took on new client with urgent deadline - 11 days from kick-off to functional implementation. We do rush only if team capacity at that moment allows without degrading other mandates' quality.
- Confidentiality practice for AI mandates?
- NDA before every workshop. All prompts and SOPs kept on encrypted servers with client-only access. Client data enters only Enterprise plans with DPA - never public ChatGPT/Claude/Gemini accounts. Mapping data categories (public, internal, confidential, sensitive personal, regulatory-protected) with explicit rules on what goes where. Local LLM (Llama 3, Mistral, Qwen) for most confidential data that mustn't leave client network.
Your team doing routine work?
A paragraph about what you do and where you see the biggest routine. We come back in 24 hours with a concrete proposal.
Other verticals:
Schule d.o.o. · Kralja Nikole 12, 81000 Podgorica, Crna Gora · office@schule.co.me · +382 67 338 777