Yesterday, we woke up to yet another proclamation of a technological revolution. This time, Forbes was heralding the arrival of Manus AI, an autonomous agent developed by a Chinese company that supposedly “changes everything” for workers worldwide. It sounded ominous. And while the article highlights some notable advancements, it also overlooks several critical points. So, let’s dial back the hype: human jobs are not at risk—at least not from Manus.
The Promise of Manus AI
According to Forbes, Manus AI can:
-
Decide which workflows to work on (“task initiation”)
-
Apply established parameters to make decisions
-
Move through each step of a workflow until completion
At first glance, this sounds impressive. An AI that initiates tasks and makes decisions? No wonder the article claims everything has changed.
The Reality Check
But as someone who’s spent nearly 20 years in enterprise software—both as a vendor and a user—I’ve seen these grandiose claims before. And they often ignore a crucial distinction: there’s a world of difference between an AI that operates in a silo (e.g., schedules vacations, screens resumes) and one that navigates workflows requiring multiple decision-makers. Which, by the way, is almost all business workflows.
The Difference Between Doing Work and Making Decisions
Here’s where the Forbes article, like many others, gets it wrong: it conflates doing work with making decisions. Those aren’t the same thing, and AI is still nowhere near replicating human decision-making in enterprise settings.
Why? Because companies don’t run on generic processes. Every business has unique, interconnected workflows with variables that shift based on real-world inputs. Even if you trained a dozen neural network models, they wouldn’t fully capture the nuances of how these processes evolve in real time.
Unique Workflows Pose Challenges for Generic AI Agents
Take a seemingly simple task: sending a quote to a potential customer. In theory, an AI agent could:
-
Add the prospect to a CRM
-
Ask a few questions about buying volume
-
Run a D&B search to qualify them
-
Send a quote based on a pricing tier
Sounds good in principle. But in reality? That’s not how most businesses operate.
Most new accounts are vetted by multiple people—sales reps, sales managers, finance teams, and sometimes even legal. They’re making judgment calls:
-
Is this prospect a good fit?
-
Could working with them jeopardize existing customer relationships?
-
Are they just fishing for a quote to benchmark against competitors?
These aren’t yes/no questions. They require context, experience, and intuition—things AI simply doesn’t have.
The Human Element in AI Decision-Making
What many fail to realize is that AI doesn’t make decisions. It follows rules set by humans. Without human-defined standards, Manus wouldn’t know what to do next.
Consider a hiring scenario: How many years of Java experience should a VP of Engineering require for a role? The answer isn’t fixed:
-
In a tight labor market, the requirement might loosen.
-
If a project is behind schedule, they might prioritize availability over experience.
Humans adapt to changing circumstances. AI doesn’t—at least not yet.
One Business Process Might Require Multiple Agents
Even a basic workflow, like discounting slow-moving inventory, introduces complexity:
-
Option A: Discount item A, but keep margins since 25% is returnable for credit.
-
Option B: Discount item A, then adjust pricing after X days based on sales velocity.
-
Option C: Discount item A, but only if item B (a complementary product) isn’t also slowing down—otherwise, bundle them.
Now layer in operational constraints:
-
Do we print new price tags?
-
Can the POS system be updated in time?
-
Are there contractual pricing restrictions?
Even for a “simple” process, decision variables multiply fast. And AI struggles with that level of nuance.
What AI Agents Can’t Do (Yet)
Despite the hype, AI agents like Manus aren’t capable of:
-
Creating workflows from scratch
-
Evaluating qualitative aspects of decisions
-
Navigating office politics
-
Adapting decision-making to shifting business conditions
This is particularly challenging for companies with thousands of workflows. Teaching an AI agent to handle these at scale—while optimizing for best practices—remains a massive undertaking, even for the biggest software vendors.
The Long Road Ahead
Building AI to handle complex business processes isn’t a quick fix—it’s a long-term project, similar to ERP or CRM implementation. And after 30 years, ERP rollouts are still complicated and time-consuming.
Why? Because:
-
Every industry is unique.
-
Every business is unique.
-
Every process within those businesses is unique.
-
And the decision-makers behind them are constantly changing.
We don’t expect one AI to run an entire department anytime soon—let alone manage thousands of workflows across an organization.
What to Expect in the Near Future
Over the next few years, we’ll likely see:
-
Incremental AI enhancements in ERP, CRM, and accounting software
-
Collaboration between AI vendors and industry experts to refine ML models
-
A slow but steady process of capturing “institutional knowledge” that’s often undocumented (and arguably impossible to fully document)
The Bottom Line
AI agents like Manus represent real progress in automation, but they’re not replacing human workers in enterprise decision-making anytime soon. The complexity of business processes, the need for adaptive thinking, and the challenge of codifying institutional knowledge mean that human oversight will remain essential for the foreseeable future.
So, workers of the world, you can breathe easy—at least until next Sunday, when another article will undoubtedly proclaim that yet another AI agent has “changed everything” again.
