5 Processes You Can Automate Today

When small-business owners think about "automation," the first association is usually something expensive, technically complex, or reserved for Fortune 500 companies. In reality, 2026 is the first year in which every tradesperson, accountant, or agency can set up an AI workflow for under €300 and reclaim 5-10 hours per week.

Here are 5 processes you can automate this week — with specific tools, step-by-step setup, and ROI math for each.

Why 2026 is the turning point for small businesses

Until two years ago, automation meant custom software, an integrator agency, and a minimum €5,000 project. Two shifts changed the math:

1. AI models are available via API. Claude, GPT-4, Gemini. What used to require a team of three engineers (text classification, data extraction, summarization) now runs through a single HTTP call. 2. No-code orchestration tools have matured. [n8n](https://n8n.io), [Make.com](https://www.make.com) and Zapier run production workflows without you owning a server.

The result: the average small business loses 15-25 hours per week on processes that can be automated — while the cost to set it up has dropped to the level of a monthly office-coffee bill.

Rule of thumb: if a task has clear rules and repeats at least 5× per week, it is a candidate for automation. These 5 processes meet that bar in almost every business.

Process 1 — Email triage and categorization

The problem. The average small business owner receives 50-200 emails per day. About 30% require no human input — spam-like offers, automated digests, notifications that only need archiving. Another 40% are routine inquiries (price, hours, availability) that can receive a standard reply. That leaves ~30% that actually requires your focus. Manual sorting burns 45-90 minutes per day.

Tooling. Gmail or Outlook + n8n (self-hosted or n8n.cloud) + [Claude API](https://docs.anthropic.com) (Sonnet for classification).

Setup in 5 steps:

1. Create Gmail API access (OAuth in Google Cloud Console). 2. In n8n, add a Gmail trigger that listens for new messages. 3. A Claude node classifies the email into a category: `urgent_client`, `new_lead`, `routine_inquiry`, `vendor_bill`, `newsletter`, `spam`. 4. A switch node routes the action: auto-reply (for routine inquiries), label + flag (for urgent), forward to accounting (for bills), archive (for newsletters). 5. For `urgent_client` and `new_lead` — notification to Slack or SMS.

Expected ROI: 45-60 hours saved per month × €20/h (real cost of your time) = €900-1,200/mo. Setup cost: 2-4h of work + €20/mo n8n Pro + ~€5/mo Claude API. Payback: first week.

The trap: don't automate replies for 100% of email in the first 30 days. Start with classification + labeling only. After 2 weeks, check whether the categories are accurate, then layer on auto-reply rules. A deeper walkthrough of small-business automation principles is in the [pillar post on AI automation](/en/blog/ai-automation-for-small-businesses).

Process 2 — Data entry from PDF documents and invoices

The problem. 30-100 vendor invoices per month. Each one needs to be opened, read, and its IBAN, amount, VAT ID, and due date retyped into your accounting system. 2-5 minutes per invoice × 60 = 2-5 hours per month of pure data entry. On top of that: transcription errors lead to wrong payments and reconciliation pain.

Tooling. Claude API (for OCR + structured extraction) + Supabase or Google Sheets as output. Alternative: [Make.com](https://www.make.com) with an OCR module if you don't want to write any code.

Setup:

1. Watcher folder (Dropbox/OneDrive) where new PDFs land. 2. A trigger runs a Claude call with the prompt: "From this invoice, extract: issuer (name + VAT ID), amount, date, IBAN, reference number. Return as JSON." 3. A validator checks that the VAT ID has the right format and the amount is numeric (guard against hallucinations). 4. Insert into the database + notify accounting.

Compliance note (EU / Croatia): from 2026 onward, e-invoice (e-Račun) is mandatory for B2B transactions. There are turnkey services (Moj eRačun, Digitalni Servis) that emit structured XML directly — no OCR required. But for invoices from foreign suppliers and legacy PDF archives, this flow is unavoidable.

Expected ROI: 2-5h/mo × €20/h = €40-100/mo + materially fewer payment errors. Setup: 4-6h.

Process 3 — Social media monitoring and auto-replies

The problem. DMs hit LinkedIn, Facebook, Instagram, and WhatsApp Business with routine questions: "Are you open tomorrow?", "What's the price?", "Do you have it in stock?". The answers are almost always the same, but the customer waits 2-8 hours before someone replies — and in the meantime, they go to a competitor.

Tooling. [WhatsApp Business API](https://business.whatsapp.com/products/business-platform) + n8n + Claude. For Instagram DMs — Meta Graph API.

Setup:

1. Register a WhatsApp Business number through Meta Business Suite. 2. An n8n webhook receives every incoming message. 3. Claude classifies: `price_inquiry`, `availability_check`, `opening_hours`, `complex_custom`. 4. For the first three — auto-reply with a predefined template (filled with dynamic data from a Google Sheet). 5. For `complex_custom` — forward to the owner + "We'll get back to you within 2 hours."

GDPR note: WhatsApp requires explicit opt-in for business messages. You cannot send proactive outbound messages without consent.

Expected ROI: response time drops from 4h to under 60 seconds = 2-3× better inquiry-to-booking conversion. For a restaurant, salon, or trade business, that is a direct +20-40% bookings.

Process 4 — Generating monthly reports from multiple sources

The problem. End of month — pull a sales report from Shopify (or your webshop), Google Analytics, CRM, and accounting. Copy-paste across 4 sources, format in Excel, send to management. 4-8 hours every single month.

Tooling. n8n (orchestration) + Google Sheets (staging) + Claude (narrative summary) + Gmail (delivery).

Setup:

1. A scheduler trigger on the 1st of each month at 08:00. 2. Parallel API calls: Shopify Orders, [GA Data API](https://developers.google.com/analytics/devguides/reporting/data/v1), HubSpot Deals, ERP export. 3. Normalize data into a common format → Google Sheets tab. 4. Claude writes a narrative summary on top of the structured data ("Top 3 products, biggest YoY swing, why..."). 5. An email template with a PDF attachment + a link to the Sheet for drill-down.

Expected ROI: 4-8h/mo reclaimed + a consistent format. Setup: 6-10h once; marginal cost per month: zero.

Process 5 — Lead enrichment and outreach

The problem. You have a list of 500 companies that could become clients. For each one you need the actual decision-maker's name, email address, and recent business events (new hires, funding, announcements) — so outreach is personalized and not generic spam. Manual: 10-15 minutes per company × 500 = 100+ hours.

Tooling. [Apollo.io](https://www.apollo.io) (275M+ B2B profiles, API available) + n8n + Supabase for deduplication + [Instantly.ai](https://instantly.ai) or Lemlist for delivery.

Setup:

1. Apollo Search API with a filter (industry, size, geography). 2. Batch enrichment in pages of 100. 3. Deduplication against previously contacted records (Supabase query). 4. Claude generates a personalized opening line for each prospect based on the company's LinkedIn summary. 5. Instantly.ai sends emails (30-50 per day per mailbox, with SPF + DKIM + DMARC configured).

Result from a real project: inside a Fraviz engagement we built a list of [1,000+ contacts in 90 minutes](/en/blog/how-i-scraped-1000-contacts-in-an-hour) using a variant of this pipeline. That campaign generated €10,000+ in revenue in the first 6 months.

Legal note: Apollo's data is legitimate (legal basis: opt-in database + GDPR compliance). LinkedIn scraping violates Terms of Service — not recommended.

How to decide WHICH process to automate first

Don't try all five at once. The framework:

3 questions:

1. Which process eats the most hours per week? (time × frequency) 2. Which process annoys you (or the team) most? (motivation factor) 3. Which process has the clearest rules? (easiest to automate)

Ideal first candidate: the process that shows up on all three lists.

ROI calculator:

``` monthly savings (€) = hours_per_week × 4.3 × hourly_cost × automation_coverage ```

Where: - `hourly_cost` = the real cost of your time (€15-50 for most SMB owners) - `automation_coverage` = the fraction of the process AI can actually cover (60-90% is realistic)

Example: email triage, 8 hours/week, €25/h, 70% coverage:

`8 × 4.3 × 25 × 0.7 = €602/month = €7,224/year`.

One-time setup cost: €300-600. Payback: one month. After that, pure ROI.

Mistakes beginners make

After three years of implementing workflows like these, three mistakes repeat:

1. Automating without first documenting the existing process. If the current process isn't clear to the humans doing it, automation will simply enforce the existing chaos at 10× the speed. Map first, automate second. 2. A 100%-automation-first mindset. AI doesn't need to replace every step. The best workflow is often 70% automated + 30% human-in-the-loop for edge cases. Design with failure states in mind. 3. No monitoring. A workflow that runs today won't necessarily run tomorrow — vendors change APIs, Claude changes model defaults, an n8n update can break a trigger. Every production workflow needs a health check that pages you when something stops working.

Next steps

If you're ready to move:

• Pick one process from the five above.
• Calculate the ROI using the formula.
• Book a [free 15-minute call](https://cal.com/fraviz/15min) so we can confirm it's a good automation candidate.
• Or send a direct [message through the contact form](/en#contact) with the specific challenge.

You don't need to know the tech in advance — we identify the right tool and ship a working workflow. The first implementation typically takes 1-3 weeks and pays for itself in the first month.