Nexly AgentsBlog

Nexly Agents › Blog › Position

Position

Don't start AI with a revolution

Published 2026-09-15·5 min read·Author: Jakub Milewski, Nexly Lab

Photo: Unsplash

Want your company to use AI but don't know where to start? Definitely not with a revolution.

At Nexly Lab we walked this road and deployed teams of AI agents in the hardest areas we have. One of them is the full automation of software development. Despite that, our team didn't shrink and nobody was replaced by AI. We are not an exception: a year ago McKinsey asked companies about expected AI-related headcount reductions, 32 in 100 expected them, and 14 in 100 actually saw them. The same team runs more projects at once today, because agents do what is repetitive and people do what requires a decision. Now we help other companies start this process.

Don't start with the foundation. Don't start with what brings stability: production, sales, serving your biggest customers. It is tempting to think that this is where most happens, where the biggest costs are and where AI will surely give the most. Except that this is also where you can lose the most. The foundation of the business is your source of financial stability. You will transform that area at the very end of the road, when the company already understands what AI can and cannot do.

Start with small, daily, repetitive things. The ones that are tedious, monotonous but important, and that somebody in the company does the same way every day. Give yourself and the company time to get used to handing part of the work to an agent that is meant to support the team, not replace it. The technology is ready on day one. Getting people used to it takes weeks.

McKinsey's August 2026 report surveyed 1,719 companies in 97 countries: almost 9 in 10 use AI regularly in at least one function, 44 in 100 are scaling it across the organization, but only 37 in 100 see a positive impact on their financial result. Companies where AI delivers at least 5 percent of profit: 6 in 100. BCG reached the same number a week later: 82 percent of CEOs are more optimistic about AI returns than a year ago, and 6 percent of companies see real value in revenue or costs. These 6 percent don't run dozens of pilots. They pick a few processes, prove the result, and only then rebuild the whole process around AI. That is exactly what happened at our company: we rebuilt the foundation last. MIT adds that a tool bought from a specialized vendor succeeds twice as often as one built in-house, though this comes from a small sample. In Poland 8 in 100 companies use AI, against 20 in the EU. Only Romania is lower. The most common reason for giving up: not knowing how to go about it (71 in 100), with costs at 33 in 100.

Three places where companies start most often and with the best results:

  1. Meetings. A transcript of a customer call that turns itself into notes and a task list, with the finished report landing in the tool you work in (Jira, Trello, e-mail). Nobody wastes time writing up agreements, and the specialist gets to work right away.
  1. Repetitive documents. Data from invoices and PDF orders entered into the system instead of by hand. Product descriptions and their translations for new markets. A first draft of a quote from a template, based on a customer's e-mail. A human checks the result before it goes further.
  1. The first line of customer contact: a chat on your website with an AI agent that knows your offer and your industry as well as your best employee. It answers customer questions without human involvement.

For this last step we built a ready-made solution, because it is the least invasive. Nobody on the team has to change how they work. The agent doesn't touch your systems and doesn't go directly into production or sales. When it detects sales potential, it hands the customer over to a human. It talks to customers on the website while your team watches and learns what comes out of it.

Deployment is one line of code on your website. Before the agent answers its first customer, it goes through an evaluation on your documentation until it passes an exam of 100 control questions. Then every morning the team gets a report: what customers asked about, where answers were missing, what is missing in the documentation. And when a customer wants to buy or asks for a human, the agent hands the conversation to a consultant or a salesperson.

Your company gets two things at once. A new customer contact channel, open 24/7 in more than 200 languages, and proof that AI takes repetitive questions off people's plates and lets them spend time on more important matters instead of taking their jobs. After a month, the conversation about the next areas looks different, because the team has seen for itself where the agent helped and what is worth automating next.

There is a third thing that hardly anyone thinks about when deploying a chat. Every conversation is a record of how the customer names your product, what they are looking for and what stops them from buying. After a month you have hundreds of such records, after six months thousands, and they grow every day, without surveys and without a research agency. For marketing this is ready material: the customer's words for ads and website headlines, unanswered questions as article topics, pre-purchase objections as arguments for salespeople. Marketing stops guessing what customers think, because customers write it themselves, every day.

Want to see how the agent handles your customers' questions? Enter your website address and e-mail in the demo form. In 20 minutes you get an agent trained on your website's offer, ready to be quizzed.

Sources

See the AI agent on your own website

Give us your website address and e-mail, and in 20 minutes you get an agent trained on your offer, ready to be quizzed.

Get a free demo

Read also

The most important value of a chatbot is not answering questions How do you know AI is telling your customers the truth? A chatbot should not close the sale