Results of mid-2025: What's new in AI and business automation

Artificial intelligence is no longer a toy: Realities of 2025

The second half of 2025 approved a new standard: AI is no longer just a generator of beautiful pictures or texts. Artificial intelligence has deeply penetrated the architecture of corporate ERP, CRM systems and automation platforms.

According to McKinsey report for 2024-2025, more than 65% of mid-market companies have already implemented generative AI in at least one business process. Having developed CRM solutions with the integration of neural networks, we at NBM-IT see the same statistics among our clients.

We analyze the main technological shifts in recent months, which will determine the competitiveness of business in 2026.

---

1. Open and corporate LLM solutions

The automation market has shifted from public ChatGPT to local (On-Premise) language models.

• Corporate LLM assistants: Large integrators (and we are among them) began to massively deploy open-source models (based on LLaMA 4 or Qwen level architectures) directly on customer servers.
• Main plus: Absolute data security (Data Privacy). The model can be “fine-tuned” using the company’s internal regulations, history of correspondence with clients and financial documents without the risk of leaking to the public Internet.
• Integration with 1C and Bitrix24: Boxed modules have appeared that independently analyze incoming emails, distribute leads into funnels and generate ready-made drafts of commercial proposals.

---

2. Low-code automation with AI agents

If earlier a developer was required to set up a complex script in n8n or Make, now they have entered the arena AI Agents (Autonomous agents).

You simply write a prompt: “Every morning, check your mail for reconciliation reports, extract the TIN and amount from them, check them with the table in Google Sheets and send a notification to the accountant in Telegram.”. The AI ​​itself selects the necessary modules (API), builds a chain and generates JavaScript/Python code for non-standard nodes.

What does this mean for business?

• Reduce time to develop internal integrations by 40–60%.
• Ability to quickly test business hypotheses.

---

3. Generative AI in Marketing and Sales (AEO/GEO)

The effectiveness of sales and SEO teams in 2025 directly depends on AI pipelines.

• Mass personalization: AI generators have been implemented to create unique product cards for online stores. Not only SEO is taken into account (including standards AEO and GEO), but also the tone of the brand.
• Call center automation: Autoresponders have transformed into full-fledged AI operators who are able to accompany a transaction from the client’s first question to sending a payment link. They understand interruptions, sarcasm, and complex technical issues.

---

4. Risks 2025: Data Quality and Regulation

Implementing AI has become easier, but managing it has become more difficult. New threats and demands have come to the fore:

• Data Quality: A neural network is useless (or even dangerous) in a company where data is scattered and chaos reigns in the databases. If the AI ​​is fed outdated price lists, it will give customers incorrect prices. Data preparation now takes up 70% of an AI project's time.
• Explainable AI: Legislation (including European regulations) obliges companies to have an audit tool: why the AI denied a client a loan or recommended dismissing an employee.
• Cybersecurity (Prompt Injection): Attackers have learned to “hack” public corporate bots, forcing them to sell goods for 1 ruble or reveal internal instructions.

NBM-IT expert advice: Always isolate external chatbots from your main ERP system through hardcoded API scripts with Read-Only permissions.

---

Practice: What should business do right now?

In order to keep up with the market, we recommend launching the AI transformation process in 4 stages:

Stage 1: Bottleneck Audit

Make a list of tasks on which your employees spend 2 to 5 hours a day of monotonous work. This could be reconciling documents, answering similar questions in chat, or collecting reports. These are prime candidates for AI integration.

Stage 2: Launching the Pilot Project

Don't try to implement AI across your entire corporation at once. Take one department (for example, help desk) and launch a simple bot trained on your knowledge base.

Stage 3: Digital hygiene

Before large-scale implementation of ML algorithms, get your IT infrastructure in order. Clear CRM databases of duplicates, structure regulations. AI only works well with clean, structured data.

Stage 4: Development of in-house expertise

Train the team in basic prompt engineering. An employee who knows how to correctly formulate tasks for a neural network works 2 times more efficiently than a colleague who does not do this.

Summary

In mid-2025, artificial intelligence is no longer a trend of the future, but a basic tool for maintaining profitability. Companies that delay implementing automation risk facing incredible labor costs compared to their more digitally oriented competitors.

Are you ready to discuss routine automation and the implementation of corporate LLM in your business? Leave a request to NBM-IT technical specialists — we will conduct a free express audit of your business processes.