AI in 2026: What Happens to Work, SaaS, and Developers Next?

Arham: Welcome back to the AI and Beyond for Business podcast. It’s 2026, we’ve made it and software has changed. It no longer just helps you work, but actually performs a lot of the work for you. I’m once again here with Yurii Yakovlev, the Chief AI Officer at Zazmic.

Yurii is the expert that companies call to automate the heavy lifting from customer service to writing code. Yurii, welcome. You spend your days helping businesses run on autopilot. So I have to ask: is the era of doing things manually slowly coming to an end?

Yurii: Hello, Arham, nice to see you again. To answer your question—no, the era of manual work isn’t over just yet. Human involvement is still essential for most processes. However, we are miles ahead of where we were a year ago in terms of automation, processes, and the use of AI tools.

Just wanted to remind you how a year or two ago we were all amazed that ChatGPT could generate text. Today, that feels basic. We’ve moved past simple prompts; now, we rely on autonomous agents and integrated automations to handle those tasks in the background.

Arham: Technology definitely moves fast. You mentioned that writing an email with AI feels old-fashioned now. What has emerged recently that truly surprised you?

Yurii: The most surprising thing for me is how fast things are moving—the new interaction protocols between AI agents, the whole new terminology we hear and use these days. The shift toward “agentic” workflows that we’re learning to apply daily to our products and services. It’s surprising how far technology has gone in doing most of our routine work. So that’s exciting.

Arham: Going back about our conversation last year, we joked about AI “hallucinating” six-legged horses. Are we still seeing those glitches, or have we finally returned to the standard four-legged horse?

Yurii: We’re still somewhere in between, though things have improved significantly. The issue now is that people are still trying to build “six-legged horses” where they aren’t needed. I see many companies forcing AI into areas where it adds no value.

When we start a project, my first question is: “What real problem are we trying to solve?” We shouldn’t use technology just because it exists. We identify the manual process, determine the goal, and only then choose the right AI stack to solve it.

Arham: That’s a great point. Many business owners risk blowing their budgets on “fancy” tech that doesn’t provide a return on investment. What are the critical building blocks for an organization to figure out which AI they actually need?

Yurii: The biggest immediate benefit is workflow automation. We often perform manual tasks by inertia—searching through mailboxes, writing meeting notes, or copy-pasting data between spreadsheets and CRMs. These are perfect candidates for automation. Otherwise, we are not making the most of technology.

I like to use an analogy: A Large Language Model (LLM) is like the engine of a high-performance sports car. If you use it only through a chat interface, it’s like putting that engine on a bicycle. When you use AI agents to connect systems and automate end-to-end workflows, you’re finally putting that engine into a Mercedes. That is how you become truly effective.

Arham: I love that analogy. We’ve seen this in Gemini Enterprise with “orchestrator” agents that manage sub-agents—for example, a marketing orchestrator we’ve built that creates images, handles SEO, and produces video for a campaign launch. It’s quite fascinating to see how you can almost bundle up a bunch of things within AI agents.

Even so, you mentioned manual work isn’t dead. We still talk about the “human in the loop.” What is the one skill humans bring to the table that AI simply cannot replace?

Yurii: Judgment of the results and final decision-making. I use AI tools, I trust them, but I want to make final calls based on experience and intuition.

For companies afraid to let go of the reins, my advice is: don’t automate everything at once. Take a step-by-step approach. Measure the quality of the AI’s output—whether it’s meeting notes or customer responses—and compare it to your manual standards. Once you see consistent accuracy, you can scale. This incremental approach builds the confidence needed to eventually automate the majority of your processes.

The current AI landscape offers countless tools and possibilities. With so many ways to achieve the same result, the key is to choose the one that fits your company’s needs and personal goals. This tailored approach helps us succeed daily in our own business, and we apply the same philosophy when working with our customers.

Arham: When you consult with business owners about setting goals for AI, how effectively can current applications meet those needs? Specifically, how do we keep customers in the loop and build their confidence, while ensuring the AI acts in their best interests, maintains data security, and hits efficiency targets?

Yurii: Current AI solutions have reached a point where they are incredibly powerful. We’re seeing a tendency where instead of buying generic, expensive third-party platforms, companies are investing in personalized solutions tailored to their specific workflows. This is now much faster and more cost-effective than it was even a year ago.

We’ve also moved beyond thinking of AI as a single system. Today, we talk about multi-agent systems—networks of AI agents making multi-purpose decisions. This involves sophisticated new protocols for communication between these agentic systems, opening up possibilities that simply didn’t exist before.

Arham: That sophisticated “multi-agentic” approach clearly handles the heavy lifting and supports high-level decision-making. But I think about the millions of young graduates entering the workforce. Historically, that “grunt work” was their training ground—the way they learned the ropes. How is the role of a junior employee changing in an AI-native organization?

Yurii: Actually, I’ve noticed that junior developers are growing faster now because they have better tools. A decade ago, to find an answer I had to spend hours searching forums and documentation. Now, the answer is immediate.

The challenge has shifted from finding the solution to analyzing it. When you use an AI assistant for coding, you have to verify: does this actually solve my specific goal? You become an orchestrator. You must assess the AI’s output and learn from it.

I’ve seen junior developers produce incredible results because they have a logical mindset, even if they lack experience with specific libraries. However, for production-ready, enterprise-level systems, experience still matters. AI doesn’t always have the “best practices” context that a senior colleague provides. As a junior, you should use these tools to the max—don’t do the manual stuff, but make sure you truly understand the results the AI gives you. That is how you’ll scale your skills.

Arham: Looking at the market, we’ve seen stocks for major SaaS companies like Salesforce and Adobe fluctuate as investors worry that internal AI agents might replace expensive third-party software. Do you think the era of buying big software packages is ending in favor of “renting” agents from a marketplace?

Yurii: Marketplace agents are certainly gaining ground. While complex legacy systems won’t disappear overnight, companies are realizing they can build personalized, autonomous solutions instead of training staff on bloated software.

The major SaaS platforms are trying to keep up by providing their own “agentic” servers, but smaller “indie” developers are often faster at building personalized solutions that are cheaper and more targeted. There’s no “one size fits all” answer, but the trend is definitely shifting toward bespoke AI workflows.

Arham: What advice would you give to those SaaS companies to survive this shift?

Yurii: Prioritize integration. They need to accept that users are bringing their own AI agents to the platform. The more open and “integratable” a system is, the more valuable it remains. They should focus on building real-world automation scenarios that make the customer’s life easier rather than just adding “AI features” for the sake of marketing.

Arham: I remember something we discussed previously—the idea of “putting the customer hat on” when building these agents. You’ve emphasized that listening to the customer is non-negotiable. Sometimes, that actually means not choosing the AI route for certain use cases if a simpler solution solves their problem more effectively. It’s about more than just padding a SaaS product with “AI features” for the sake of it; it’s about finding a meaningful, value-driven way for the customer to interact with the technology.

I truly appreciate you sharing that perspective. To wrap things up, what are the most ambitious projects you’re currently tackling at Zazmic—either internally or for your customers?

Yurii: We are currently focused on a massive internal project to automate our own departments. It started as a simple Q&A bot, but we’ve evolved it into a multi-agentic system using a knowledge graph. It allows our employees to get immediate answers and automates the background “grunt work” that used to take up so much time. Applying these same advanced architectures to our clients’ projects is what really excites me right now.

Arham: Yurii, thank you for your time and for sharing your insights on how the marketplace is evolving. To our listeners, thank you for joining us. We’ll be back next year to see just how far these AI agents have taken us!