Behind on AI? Here's What You Can Already Put to Work in Your Company

As many as 39% of Polish small and medium-sized companies that use AI use it mainly to write content: e-mails, newsletters, social media posts. Creating graphics and video came second (33%) and designing websites third (28%), according to home.pl's report "Cyfrowy puls polskich MŚP 2026" (The Digital Pulse of Polish SMEs 2026). In other words: most companies that have reached for AI at all have stopped at the "better ChatGPT for content" stage.

That's not a criticism, it's a starting point. Because if AI implementation in your company ends with writing copy, you're probably not using even a fraction of what it is genuinely good for today: handling enquiries, analysing data, preparing quotes, working with product data or automating entire processes. In this article we show concrete, practical examples of such implementations, with no futurism and no promises that "AI will change everything", to make it easier to judge what could be carried over to your own business.

Before you implement AI: start with the problem, not the tool

The most common mistake when implementing AI in a company is doing it backwards: first choosing a fashionable tool, then looking for something to use it on. It works far better the other way round: start with a specific, recurring problem and only then pick a solution for it.

Good candidates for AI implementation are usually situations where:

• an employee spends an hour every day on the same task,
• a salesperson manually analyses and qualifies incoming enquiries,
• someone regularly copies the same data between systems,
• preparing a single quote takes several hours,
• working with hundreds or thousands of products means repeating the same manual steps,
• the business owner is the "bottleneck" of a process, because only they know how to do something correctly.

Not every one of these problems calls for building a custom solution. Sometimes a well-configured, off-the-shelf tool is enough. The key question isn't "which AI should we implement" but "which process in my company will actually benefit from it".

Case #1: AI in B2B, from enquiry to a ready quote

This scenario shows well how AI can support an entire process rather than a single step. The classic path in a B2B company looks like this: customer enquiry → analysing the message → extracting parameters → qualification → preparing a reply → gathering data → quote → entry in the CRM.

In practice it might work like this: an enquiry arrives through the contact form on the website, the data lands in a CRM (for example HubSpot), information about the prospect is pulled from there, and Claude prepares a preliminary quote based on it. Thanks to an integration with the mailbox, the same reply can be prepared for sending straight away, with a human who verifies the content before it reaches the customer.

That last sentence matters a great deal. The question "should AI send the quote on its own" has no single right answer. It depends on how sensitive the data AI works with is and where in the process a human has to step in (the so-called human in the loop). In most B2B implementations the more sensible model is one where AI prepares and a human verifies and approves the proposal.

The value of such an implementation is easier to measure than it seems: how much time a salesperson spent preparing one quote before the implementation, and how long it takes them now. No complicated metrics are needed here, just a "before" and "after" comparison.

Case #2: AI on marketplaces and in e-commerce

This is one of the areas where scale makes an enormous difference. If you sell a dozen or so products, writing descriptions by hand is manageable. If you sell several thousand, it's a completely different story.

A typical scenario looks like this: product data is pulled from the manufacturer's site (for example a B2B portal) and then "enriched": AI improves and expands descriptions, adapts titles and parameters to the requirements of a specific sales channel, and the finished data flows through an integration (for example with Shopify) straight into the store.

In this area AI can support:

• creating and optimising product descriptions for a specific sales channel,
• classifying products and matching them to marketplace categories,
• preparing titles that meet a given platform's requirements,
• analysing customer reviews and spotting recurring problems in listings,
• analysing sales and flagging products that need attention.

More important than "will AI replace people here" is the question of whether it eliminates tedious manual work across a large number of products, which is not the same thing. With 100 products the difference is often small. With 10,000 products AI stops being a curiosity and becomes a genuine precondition for keeping up with the size of the catalogue at all.

Case #3: AI as the company's data analyst

The data source can be practically anything the company already collects: sales, CRM, Google Ads, Google Analytics 4, e-commerce data, quote requests, complaints or customer reviews. Instead of waiting for a weekly report prepared by hand, an owner or manager can ask AI a direct question, for example which products lost sales in the last quarter, or whether there's a common pattern in incoming complaints.

One important caveat is needed here, though: AI can help find a pattern in the data, but that doesn't automatically mean it understands the business context behind that pattern. A drop in sales of a given product may stem from seasonality, a mistake in an ad campaign or a competitor's price change. AI will point to the correlation, but interpretation and the decision still belong to a human.

Case #4: AI connected to the company's knowledge

This is the moment when AI stops being just "ChatGPT used at work" and starts being genuinely embedded in the organisation. Companies usually have scattered, hard-to-search knowledge: instructions, documentation, procedures, quotes, product catalogues, FAQs, technical materials, or knowledge that individual salespeople keep only "in their heads".

Instead of searching through all of this by hand across different files and systems, an employee asks AI a question, AI searches the company's connected knowledge base and returns an answer, and a human verifies it before using it. This solves a specific, recurring problem: knowledge locked in the head of one person who happens to be on holiday or no longer works at the company.

This is also a good moment to distinguish two completely different situations: "we use ChatGPT" and "we have AI embedded in a company process, connected to our data and documents". The first is a tool on one person's desk. The second is part of the company's infrastructure, and it's that second level that makes a real difference at the scale of the whole organisation.

What NOT to implement just because it's AI

Not everything that can be automated with AI is worth automating. A few typical mistakes worth sparing yourself:

• an AI chatbot on the website that nobody uses, because nobody needed it,
• automating a process that takes five minutes a month anyway,
• generating enormous amounts of content just because it's technically possible,
• deploying an AI agent where a simple form would do,
• an implementation with no human control over the output,
• buying yet another AI tool without integrating it with the systems the company already has.

The opposite is also worth remembering: if you as the owner don't see an obvious use for AI in your company, that doesn't necessarily mean there isn't one. Sometimes your team, working closer to a specific, recurring problem, knows better where AI would actually help. It's worth giving them room to say so.

Where are you with AI today?

It's worth honestly assessing which of these three levels your company is actually at:

1. "AI? We have ChatGPT." Individual people use individual tools, mainly to write and polish text.
2. "AI helps us do the work." AI is part of specific, recurring processes: it supports handling enquiries, working with product data or analysing data.
3. "AI does part of the processes for us." AI is connected to the company's systems, data and automations, not just to a single chat window.

If, after reading this, you recognise your company at level one, it doesn't necessarily mean you're "behind". It rather shows where the biggest, still untapped reserve lies. It's worth starting with one specific implementation you can evaluate as early as next month, rather than trying to automate the whole company at once.

Summary

Real AI implementations in companies rarely look like "digital transformation" presentations. Usually it's a specific, recurring problem that, thanks to AI, takes less time and less manual work: a faster response to a quote request, less manual work across thousands of products, quicker access to sales data or easier access to knowledge scattered across the company.

If you want to find out which process in your company is genuinely suited to such an implementation, before you invest time and budget in the wrong place, we'll help you assess it and design a specific automation.