“So what are you doing about AI?”
You’ve probably heard it once or maybe a hundred times already.
AI is no longer just a nice-to-have. It’s an expectation. And for many enterprise leaders, that brings an insane amount of pressure: 74% of CEOs say they’re worried about losing their jobs if they don’t show clear business results from AI.
That pressure is pushing companies to ramp up their investments. As a result, enterprise spending on AI is expected to grow 10x between 2023 and 2028.
From what we see in the market, much of that money is being poured into isolated tools, like copilots to help draft emails or LLM chatbots to handle support tickets. Are they useful? Sure. But these are surface-level use cases for GenAI that don’t reflect the bigger shift happening to enterprise software.
At Modeso, we believe the real impact of AI won’t come from how many “AI-powered” tools you plug into your stack. It’ll come from how AI changes enterprise software itself — how it’s built and how it’s used.
We think agentic AI will be a key driver of that change.
Since ChatGPT exploded in 2021, we’ve been riding wave after wave of AI hype: first LLMs, then retrieval-augmented generation (RAG), then multi-modal models.
In 2025, AI agents got the spotlight.
There’s a clear business incentive behind this. Leaders like OpenAI’s Sam Altman are calling AI agents the next step, saying they’ll “join the workforce.” Microsoft’s Satya Nadella says agents will disrupt standard software entirely by automating business logic. Tech giants introduce new models for the “agentic era.”
Whether this buzz is to justify their investments in agentic models — or winning the post-ChatGPT arms race among tech giants — we don’t know for sure.
But what we do know is that in 2024 alone, startups building AI agents raised over $8.2 billion. So yes, the hype is working.
Despite all the hype, no one seems to agree on what AI agents actually are. Right now, the term refers to everything, from chatbots to automation flows. In many cases, it means “AI assistant.”
But when we talk about AI agents, we’re more interested in where this technology is heading.
Modeso’s CTO, Ahmed Mohsen Gharib Farag, sees agents not just as tools that automate tasks but as a new layer of software. One that understands business logic, reasons through steps, and takes action accordingly. As Ahmed puts it, “the LLM is like the brain, but the AI agent is the full body.”
The image above explains the core functionality of AI agents: not just tools that respond but systems that think through problems, take action, and coordinate across tools, tasks, and even departments.
Yes, our team sees it as a shift from AI being reactive (just answering) to being proactive (taking action).
Each stage in the GenAI journey for consumers — LLMs, RAG, and now agents — has brought AI closer to solving real, specific problems for real users in enterprise settings.
AI agents bring together different types of AI — language processing, perception, and decision optimization — into systems that can act autonomously. This grouping is based on a framework by data expert Dylan Anderson, who mapped out four key categories of AI solutions: language-based, perception-based, decision-optimization, and autonomous systems (agents) that combine all three.
Okay, but what does that mean for how enterprise software is built?
The next wave of enterprise software won’t just use AI. It will be built around it. That shift will reduce the need for heavy interfaces, move logic to the AI layer, and fundamentally change how software works.
Still, no one can predict exactly how this will unfold. But the overall direction is clear: AI is impacting how software is made. That’s the shift Ahmed believes will define the next era of enterprise tools, and it’s something we’ve been thinking about at Modeso for a while.
Today, there are already plenty of tools that let you “build your own AI agent.” Most focus on things like generating no-code websites, automating small tasks, or chaining prompts together to solve narrow, well-defined problems like writing emails.
AI-powered solutions may look futuristic, but under the hood, they still work like traditional software: many dashboards, buttons, and data fields. The AI might help here and there, but the structure is the same. The user still drives the process.
Here are the two things we think will change.
Enterprise apps today rely on complex user interfaces (dashboards, forms, filters, and charts) because users need to structure data, apply logic, and make decisions manually.
In an AI agent-driven world, that structure starts to disappear. Instead of navigating through tabs and fields, you’ll just say or type what you need, like ‘show me the top 3 best-selling products from last year, with a short analysis.’
That’s it. One simple input, like a ChatGPT-style prompt box, is all the AI needs.
The business logic — a set of rules according to which software works — will become something that lives in the brain of the AI agent, not in the software’s backend.
We’ll be able to describe what the software should do in plain language or structured templates, and the AI agent will read, interpret, and act on them.
For example, today, a developer writes code to define what the system should do in a specific situation – like: “if the invoice is overdue more than 30 days & Value < 20.000 → send a reminder email. Else → Notify Management.”
This kind of logic is built into the backend and executed step by step: the system gathers data through forms or dashboards, applies the business rules, and returns the result.
If the rule itself needs to change (say, to include a new condition like invoice amount or customer type) that usually means updating the codebase or using a complex rule engine. It’s doable, but it takes effort and developer time.
In the future, that logic might be written in plain text or structured templates by a business user, not a developer. The AI agent would interpret that rule, apply it in real time, and take the right action without requiring the software to be rebuilt.
With this shift, business logic becomes dynamic, editable, and human-readable. It’s less about maintaining backend code and more about clearly describing what needs to happen.
Messy enterprise data isn’t ready for the agentic AI shift. Integration is tricky. Security is an open question. And data quality? Still, a long way to go.
The upside is that AI hype pushes more companies to finally get serious about data cleanup, governance, and modeling. That’s a step in the right direction. But it doesn’t happen overnight.
Even the companies building this tech can’t be 100% sure. No one can.
But the market is clearly moving in this direction. Big players are investing. Startups are experimenting. Enterprises want to make AI actually work, not just as a feature.
So, if you’re thinking long-term, this is the shift to watch.
Based on how the tech is evolving and how enterprises respond, here are four things we think could shape the future.
Enterprise standard software, or SaaS, won’t look like one big suite. Instead, we’ll see smaller, more specialized tools — an inventory micro-app here, an HR action tool there — while the intelligence that connects and drives them lives in the AI agent layer.
If business logic moves to the AI agent layer, SaaS tools become interchangeable components. They’ll still be needed, but their strategic value drops. They’re no longer where decisions are made or workflows live.
In short, SaaS becomes the pipes and systems doing the actual work. An AI agent is the intelligence deciding what work needs to happen and why.
The more companies rely on the same AI tools, the harder it becomes to stand out.
If everyone uses the same off-the-shelf AI agents, everyone starts solving problems the same way. When your AI agent makes the same decisions as your competitor’s, you lose what made you different.
The risk grows if you don’t control your data. When companies let external tools learn from their internal processes, their unique know-how can leak into the models — models others can use. “Your expertise feeds AI models,” Ahmed warns. “And if others use that same model, they gain what made you unique.”
Over time, businesses may lose internal skills, too, just like people forgot how to do long division when calculators became standard. That might be fine for routine work. But for fields like logistics, finance, or compliance, your expertise is your product.
So that’s what’s risky: AI doesn’t just automate your work. It might also automate away your edge.
But that leads us to the next prediction.
Just because an AI agent can decide doesn’t mean it always should.
We believe the biggest difference between standard SaaS tools and custom-built AI agents will be this: who’s in control and when. Most out-of-the-box tools don’t let you design for human-in-the-loop moments. As Ahmed points out: “With standard tools, you have no real control over the workflow. It’s not easy to decide which steps need human confirmation — and which don’t.”
But with custom AI agents, you can balance when the AI needs to check in with a human and when it can act on its own.
Does it mean we think that custom solutions will overtake standard software? Not quite.
AI agents won’t eliminate the need for either standard or custom software. They’ll just make the tradeoffs more visible.
Our survey shows most enterprises use a mix of custom and standard software. That won’t change that overnight, if ever.
The “buy vs build” debate will only get more nuanced in the AI era. The key question will shift from “Should we build or buy?” to:
AI’s real impact isn’t just in single-use tools like LLM chatbots or smart dashboards. It’s much bigger than that. It’s changing what enterprise software is: how it’s built, used, and valued.
Too many companies are stuck in “we need to do something with AI” mode without a clear plan.
But things are getting better. If we are in prediction mode, here’s another one from a data expert Dylan Anderson: “The primary role AI will take in organizations [in 2025] is small-team experimentation and delivery.”
Instead of big, risky moonshots that add to the AI project failure rate, companies may start focusing on smaller, proven AI use cases, often built on top of their existing data and analytics work. They start taking a smarter approach, moving slowly, and testing ideas.
The most important thing here? Making sure their data and workflows are ready for scaling with AI.
Take a step back and ask: Are we just adding tools — or are we actually making AI work for our business?
Assess your data foundation. AI is only as smart as the data you feed it. If your data is messy, fragmented, or poorly governed, your AI efforts won’t scale. So start from here.
Yes and no. Companies like ours will keep building custom software, but what “custom” means is shifting. It’s not just about building software that works. It’s about helping companies figure out what kind of software they need in an AI-driven world and how to make long-term decisions around that.
That means helping our clients figure out:
The agentic AI shift is bigger than any one tool or feature. It’s a rethink of how software is built and what it’s for. And we’re already helping businesses lead that shift, not just by building smarter systems but by taking real ownership of the product thinking behind them. If that’s something you're exploring, we’d love to hear from you.
Privacy is important to us, so you have the option of disabling certain types of storage that may not be necessary for the basic functioning of the website. Blocking categories may impact your experience on the website. More information
