Inside Hostinger’s AI-first approach: What the agentic AI era means for businesses

Instead of waiting for a punctual and returning a consequence arsenic chatbots do, AI agents tin usage tools, make decisions, and return actions to activity toward a goal.

But really good does that really activity today? Where do AI agents still autumn short? And what does an agentic attack look for illustration erstwhile you’re really building and utilizing these systems?

We sat down pinch Tomas Rasymas, Hostinger’s AI Research Lead, to abstracted what’s happening now from what’s still further down the roadworthy – and look astatine what the displacement toward agentic AI already intends for the measurement we work.

The biggest misconception? That agents tin do everything

We asked Tomas what trips companies up first, and he didn’t hesitate: it’s the belief that AI agents tin do everything.

They can’t. At least, we’re not location yet.

For an AI supplier to really beryllium useful, Tomas says it needs 3 things:

• A clear goal. The supplier needs to cognize precisely what it’s expected to achieve.
• Context. It needs the accusation required to run successfully.
• Boundaries. It needs clear limits connected what it should and shouldn’t do.

Without these, moreover a highly tin supplier won’t needfully nutrient useful results.

That’s why Tomas advises against trying to merge AI agents into everything. Instead, prime 1 elemental problem and lick it first.

Once you’ve figured retired really to make an supplier activity for that usage case, you tin iterate, study from the results, and move connected to bigger problems. Starting mini gives companies a measurement to build connected what’s working alternatively than trying to lick everything astatine once.

A amended measurement to adopt an AI supplier is to dainty it arsenic you would a caller employee. If you’re wondering what that looks for illustration successful practice, our guide to maximizing AI agents walks you done really to get the astir retired of them.

Why should companies return AI earnestly now?

AI is already wide utilized by companies to lick existent problems, from proposal systems and fraud discovery to forecasting. But these applications were typically built to lick circumstantial problems wrong defined parameters.

According to Tomas, we’ve now seen a much important shift: AI models tin activity pinch a scope of tools, opening up possibilities beyond predefined applications.

AI isn’t constricted to generating an reply anymore. It tin automate tasks, make decisions, and return action.

That operation of tin models and devices has made AI look little for illustration a passing inclination and much for illustration thing companies request to study really to use.

And successful fact, integrating AI into your stack doesn’t adjacent building a blase system.

Even small businesses tin usage AI successful simple, applicable ways to return activity disconnected people’s plates, from automating repetitive tasks to handling mundane business processes.

The possibilities alteration wide depending connected the business and the problem you’re trying to solve. The important point is to commencement looking astatine where AI tin do useful work.

What does being AI-first really look for illustration astatine Hostinger?

For Tomas, being AI-first intends looking for problems AI tin genuinely lick amended aliases faster, alternatively than adding it to each process conscionable for the liking of it.

At Hostinger, that plays retired simply: if AI tin do it, we look for a measurement to put that capacity to work. If it can’t, there’s nary logic to unit it into the workflow.

That attack shows up successful really we usage AI internally and really we build AI into our products for customers.

How we usage AI internally

One of the clearest examples of really we merge AI into our mundane work is the usage of DEX, our soul AI assistant. Built by our ain developers, DEX is continuously improved pinch caller features to support our activity and lick our ever-changing business operations needs.

Alongside that, we person a supportive situation for experimenting pinch and adopting AI devices to amended our workflows, including for non-technical teams. 

For much precocious usage cases, our dedicated AI and automation squad helps teams crossed the institution automate processes and find amended ways of working.

We’ve besides tally an soul AI Agents hackathon, wherever teams crossed the institution explored really agents could lick existent problems. Some of those ideas person since moved beyond experimentation and into our mundane workflows.

How we bring AI to customers

With the aforesaid attack we return internally, we want to make building and increasing an online business much accessible, including for group without extended method knowledge aliases the clip to grip everything themselves.

That’s wherever Tomas’ squad focuses their work: exploring what’s imaginable pinch AI and uncovering ways to bring it to our customers.

We started by integrating AI to simplify website creation and VPS hosting management, past expanded it to galore different areas, including email marketing, ecommerce, and business tasks management.

As AI capabilities proceed to improve, we’re evolving the exertion down our products and bringing much agentic capabilities crossed Hostinger. Our extremity is to create a much seamless acquisition for customers utilizing Hostinger arsenic a unified platform.

One of the astir caller examples is the improvement of Kodee, our first AI chatbot. Initially designed chiefly arsenic a customer support assistant, Kodee has taken connected a overmuch broader domiciled arsenic its capabilities person grown.

Today, Kodee has evolved into Hostinger Agent, going beyond answering questions to thief users pinch customer support inquiries, business guidance tasks, and different activity they request to get done.

So, arsenic AI gets amended astatine answering questions and taking action, 1 mobility follows naturally: 

What happens to the quality domiciled arsenic AI takes connected much of the work?

As AI agents return connected much repetitive work, Tomas believes the quality domiciled will progressively halfway connected judgment successful ambiguous situations.

AI tin process the accusation it’s given, but it doesn’t ever person the discourse aliases acquisition to cognize what matters erstwhile location isn’t an evident answer. 

There are things we know, understand, aliases prime up connected without ever explicitly putting them into a strategy – and that tin make a quality erstwhile a determination isn’t clear-cut.

At the aforesaid time, the measurement we activity pinch AI is changing. We’re already seeing caller ways of moving look crossed teams, too, arsenic expectations and needs person evolved. 

Our Customer Success team, for example, now includes AI CX Engineers – group who activity down the scenes to amended really Hostinger Agent responds to our clients. They thief fine-tune responses and place ways to make the acquisition amended for customers utilizing Hostinger Agent.

Even arsenic AI becomes much profoundly integrated crossed the company, we’re still hiring crossed departments, pinch group taking connected caller responsibilities and uncovering caller ways to activity alongside progressively tin AI systems.

Asked what would group the champion group isolated if everyone had entree to tin AI agents, Tomas pointed to 2 things: asking the correct questions and owning the outcome.

In different words, the worth of moving pinch AI is knowing what you want to achieve, really to guideline the supplier toward it, and erstwhile to measurement successful and make the call.

The agentic era has conscionable begun

The existent authorities of AI agents is little futuristic than the hype sometimes suggests.

Today’s agents still request clear goals, the correct context, and boundaries. They tin return connected meaningful tasks, but they aren’t cosmopolitan autonomous assistants that tin reliably grip thing you propulsion astatine them.

Still, we’re already seeing what happens erstwhile AI tin return action. 

At Hostinger, that displacement is evident successful really we build products, troubleshoot issues, and thief customers move ideas into websites.

If you want to commencement bringing AI into your business operations, enactment funny and support asking wherever it tin lick problems amended aliases faster.

Watch the afloat question and reply pinch Tomas connected Hostinger Academy: