AI in Your Business — How to Start?

Every week brings headlines about companies saving millions with AI. And every week a business owner sits in their office wondering: “This sounds great, but where do I actually start?” The gap between AI coverage in the press and practical implementation in a 30-person company is enormous.

This guide closes that gap. It is not about theory, hype or “digital transformation.” It is about concrete first steps, realistic costs and how to measure whether AI is actually working for your business.

---

Why Most SMEs Still Haven’t Implemented AI

The data is telling: Deloitte’s 2025 AI Study shows that 78% of large companies already use AI in some capacity, while among SMEs the figure is only 23%. The reasons are predictable:

• “We don’t know where to start” — lack of clear methodology
• “We can’t afford it” — perception (often incorrect) that AI requires large budgets
• “We don’t have technical people” — misconception that implementation requires a team of data scientists
• “We’re afraid of failure” — lack of practical examples from companies of a similar scale

The reality is different. In 2026, you can implement the first useful AI tool in your company in 2–4 weeks, for PLN 2,000–10,000, without hiring a single additional IT person. The key is choosing the right first project.

---

AI Implementation in 5 Steps

Step 1: Audit of Pain Points (Week 1)

Before you choose a tool, you need to answer one question: what is consuming the most time in your company?

Create a list of repetitive, time-consuming activities. For each one, note:

• Who does it and how often
• How long it takes
• What could go wrong when done manually
• Whether the task requires human judgement or is mechanical/procedural

Typical candidates for first AI implementation:

• Processing enquiries and emails
• Creating reports and summaries
• Support for offers and proposals
• Answering frequently asked questions (internal or customer-facing)
• Data entry and processing

At this stage you don’t need advanced tools — a shared spreadsheet and 2 hours of honest conversation with your team is enough.

Step 2: Selection of the First Project (Week 1–2)

The ideal first AI project meets four criteria:

1. High time cost — at least 10–20 hours per month in total
2. Low stakes for error — if AI makes a mistake, it’s easy to catch and correct
3. Structured data — the task doesn’t require complex contextual judgment
4. Measurable outcome — you can clearly compare “before” and “after”

Examples of good first projects:

Process	Example AI tool	Estimated time saving
Email processing and triage	Gmail AI / Outlook Copilot	1–2 h/day
Meeting summaries	Otter.ai / Teams Copilot	30–60 min/meeting
FAQ responses to customers	Chatbot + knowledge base	40–60% of customer service volume
Creating proposals and offers	GPT-4 with template	50–70% of writing time
Data entry from documents	IDP (Intelligent Document Processing)	80–95% of manual work

Step 3: Pilot — Quick Win in 4 Weeks

Choose one process. Implement a minimal solution. Measure results.

Pilot structure:

• Week 1: Tool selection and configuration
• Week 2: Internal testing (team uses the tool, gives feedback)
• Week 3: Running in parallel (AI alongside the old method — comparison)
• Week 4: Assessment and decision: scale up or change approach

Budget for a typical pilot: PLN 2,000–8,000 (tool licence + configuration + training)

The goal of the pilot is not perfection — it’s proof of concept. You want to know: does this tool work for our specific business context?

Step 4: Measurement and ROI Assessment

Before you start the pilot, define what you’re measuring. After 4 weeks, compare:

• Time spent on the task (before vs after)
• Error rate (complaints, corrections, rework)
• User satisfaction (the people using the tool)
• Cost (licence + maintenance vs saved working time)

Simple ROI calculation formula:

ROI = (Monthly savings in working time [h] × hourly labour cost [PLN]) – Tool cost [PLN/month]

Example: Meeting summarisation tool

• Before: 8 meetings/week × 30 min notes = 4 h/week = 16 h/month
• Labour cost: PLN 80/h (PLN 150 gross/h with overheads)
• Monthly saving: 16 h × PLN 80 = PLN 1,280
• Tool cost: PLN 200/month
• Monthly ROI: PLN 1,080

This is a small but completely real and measurable effect. After 12 months: PLN 12,960 savings on one process.

Step 5: Scaling

After a successful pilot — extend to other processes. Build on what works.

Scale-up strategy:

• Month 1–2: pilot (1 process)
• Month 3–4: extension (2–3 processes)
• Month 5–6: integration (connect AI tools between themselves and with existing systems)
• Month 7+: optimisation and advanced use cases

The key to successful scaling is not adopting the most advanced tools immediately. It is systematic expansion of what already works.

---

Practical AI Tools for SMEs in 2026

Communication and Email

Microsoft Copilot for Microsoft 365 (PLN 120/user/month) Best choice for companies already using Office 365. Summarises meetings, drafts emails, creates presentations, answers questions based on company documents.

Google Workspace Gemini (from PLN 80/user/month) Equivalent for Google ecosystem users. Gemini in Gmail, Docs, Drive, Meet.

Superhuman / SaneBox (from PLN 50/user/month) AI triage and email prioritisation for heavy email users.

Customer Service

Tidio (from PLN 100/month) AI chatbot for the website — handles FAQs, collects leads, escalates to a consultant. Minimum configuration (without custom knowledge base): 2–4 hours.

Intercom Fin (from PLN 400/month) More advanced chatbot with AI answering on the basis of documentation. Better for companies with developed knowledge bases.

Freshdesk Freddy AI (from PLN 150/user/month) AI support for the service team — answers suggestions, ticket classification, analysis of response patterns.

Content and Documents

ChatGPT Plus / Team (PLN 80–400/month) The most versatile generalist tool. Good for: proposal drafting, content, summaries, analysis, brainstorming.

Notion AI (PLN 30/user/month in Notion) AI integrated with team documentation. Good for: meeting summaries, knowledge base management, task descriptions.

Gamma.app (from PLN 50/month) AI presentation creation. Input: text description → output: ready-made presentation with design.

Sales and CRM

HubSpot AI (in the CRM subscription) AI assistant for salespeople — deal summaries, email drafting suggestions, sequence optimisation.

Salesforce Einstein (from PLN 800/user/month) Enterprise-grade AI, but available in lower tiers for SMEs. Best for companies already on Salesforce.

Apollo.io AI (from PLN 200/month) AI support for prospecting and outreach — personalisation of messages, lead analysis.

Automation and Processes

Make + AI modules (from PLN 500/month) Automation platform with built-in AI models (OpenAI, Claude, Gemini). Good for: connecting existing tools, process automation with AI decisions.

n8n (free self-hosted / from PLN 200/month cloud) More technical alternative to Make — greater flexibility, lower operating costs at scale.

---

Most Common Mistakes When Implementing AI in an SME

Mistake 1: Starting with the Most Complex Solution

Companies often want to implement the “best and most advanced” tool immediately — instead of the simplest one that solves the real problem. Result: high costs, long implementation time, and ultimately abandonment.

Solution: Start with the simplest tool that solves 80% of the problem. You can always upgrade later.

Mistake 2: No Measurement Baseline

Many companies implement AI and then can’t say whether it worked because they didn’t measure what “before” looked like. “We think it’s faster” is not a result.

Solution: Before starting the pilot, measure the baseline — time, frequency, cost, error rate.

Mistake 3: Skipping the Pilot Stage

Companies go straight from “great idea” to “company-wide roll-out” without a controlled pilot. When something goes wrong (and it always goes slightly differently than expected), the scale of the problem is much larger.

Solution: Always pilot first — in one team, on one process, for 4 weeks.

Mistake 4: Ignoring Team Resistance

AI implementation without communicating the purpose to the team generates resistance. People fear for their jobs, see AI as a criticism of their work, or simply don’t trust new tools.

Solution: Involve the team from the start. Explain what AI is supposed to do (take over mechanical tasks) and what it won’t do (replace people in high-value tasks). Make team members partners in the implementation, not its subjects.

Mistake 5: Underestimating Data Quality

AI works as well as the data it is fed. If your documents are disorganised, data is scattered across five different systems and naming conventions are inconsistent — AI will not magically sort this out.

Solution: Before implementing AI, do a mini audit of data quality in the area you want to automate. Often, 2–3 days of cleaning and organising existing data significantly improves AI performance.

---

How Much Does AI Implementation Cost in an SME?

Scenario 1: Self-Implementation (Budget: PLN 500–2,000/month)

Suitable for: companies with a technically competent person on the team (doesn’t have to be a programmer — a curious “power user” is enough).

Tool	Cost
ChatGPT Team (5 users)	PLN 400/month
Make Pro (automation)	PLN 500/month
Tidio (chatbot)	PLN 100/month
Training and configuration (one-off)	PLN 2,000–5,000

First-year cost: PLN 12,000–17,000 Typical saving: PLN 30,000–80,000/year (20–50 hours/month × PLN 80/h × 12) ROI: 150–400% in the first year

Scenario 2: Consulting-Supported Implementation (Budget: PLN 3,000–8,000/month + consulting)

Suitable for: companies without an internal technical resource, or those wanting faster and more reliable implementation.

Item	Cost
Tools (suite of several applications)	PLN 2,000–5,000/month
Consulting — audit and strategy	PLN 10,000–25,000 (once)
Implementation and configuration	PLN 15,000–40,000
Training and documentation	PLN 5,000–10,000
Monthly support	PLN 2,000–5,000/month

First-year total cost: PLN 60,000–130,000 Typical saving: PLN 120,000–350,000/year (companies from 20 employees) ROI: 100–300% in the first year

Scenario 3: Comprehensive AI Strategy (Budget: PLN 15,000–50,000/month)

Suitable for: companies at 50+ employees, wanting AI integration across multiple business areas.

This is a multi-project programme — typically 3–5 simultaneous implementations in different departments, with a dedicated internal AI coordinator.

---

FAQ — AI Implementation in an SME

Does my company need a data scientist or AI engineer?

Not for first AI implementations. Tools like ChatGPT, Make, Tidio or Copilot require no programming knowledge. They need a “power user” — someone who is curious about technology and willing to spend 2–4 hours learning a new tool. An internal expert is worth bringing in when you’re building custom AI solutions or integrating multiple systems.

Is AI data safe? What about GDPR?

It depends on the tool and how you configure it. Business versions of most tools (ChatGPT Team, Microsoft Copilot, Google Workspace) have GDPR-compliant contractual agreements and data processed in the EU or with appropriate transfers. Key rules: don’t enter personal customer data into generalist tools without a data processing agreement; use Business/Enterprise tiers, not Free. We discuss details in our GDPR and AI article.

How long does it take to see first results from AI?

For a well-chosen first project: 2–4 weeks. The pilot clearly shows whether the tool works for your specific case. Measurable effects (time savings, error reduction) appear immediately after implementation. More complex effects (customer satisfaction improvement, better decision quality) are visible after 3–6 months.

Can AI completely replace some of my staff?

In some cases — yes, for highly repetitive tasks. But at the SME level this is rarely the primary goal. The typical result of implementing AI in an SME is not a headcount reduction, but a growth in productivity without a growth in costs. Employees released from mechanical tasks move to higher-value work — sales, customer relationships, complex problem-solving.

---

What’s the Right First Step?

You now have a framework — but every company has a different starting point. What works for a marketing agency may not work for a manufacturing company. What is a “quick win” for a 100-person firm may be overkill for a 10-person team.

The most common mistake is choosing the wrong first project — either one that’s too complex, or one where the ROI is too low to be convincing.

QA10 Process Intelligence Audit is a structured analysis that, in 2–3 weeks:

• Maps your processes and identifies those with the highest AI automation potential
• Estimates ROI for each candidate
• Recommends concrete tools matched to your specific situation
• Creates an implementation roadmap with timelines and budget

Without guessing — based on data from your company.

➡ Schedule a free initial consultation