Low-code changed the rules by letting people assemble software instead of writing it

Aayush Kumar, CEO and Founder, Zyven Technologies.

How are low-code platforms changing the way enterprises approach Industry 4.0?

For decades, automation worked like a members-only club. If you wanted a workflow built, you filed a ticket, waited, followed up, waited some more, and eventually got something that solved last year’s problem.

Low-code changed the rules by letting people assemble software instead of writing it. The next step is even simpler: describe what you want in plain language and let an AI agent do the assembly through connected tools. The Model Context Protocol (MCP) is what makes this practical, because it gives agents a standard way to plug into enterprise systems instead of a custom adapter for every one. Automation moves from “IT’s problem” to “everyone’s tool”.

What impact can low-code have on speed from idea to scale?

Most automation projects don’t die in development. They die in the waiting room. Between “this could be better” and “this is live” sits a queue of approvals, meetings and “let’s circle back” emails long enough to make a government office blush.

I’ve seen this first-hand. In 2023, for iCreate’s SparkUp Fund, I built a concept for a smart vehicle monitoring system. A device plugs into a car’s OBD2 port, pulls live data from the ECU, adds GPS, and sends it all to AWS, where machine-learning models flag anomalies, accidents and rash driving. A companion app adds geofencing, maintenance prediction based on usage patterns and, my personal favourite, a chat interface so you can simply ask your car how it’s feeling. The judges liked it enough to name me a winner, alongside two other participants. It’s a concept, not a production system, but the lesson holds. The data already existed, and the speed came from connecting it to models and a plain-language interface instead of building everything from scratch.

Swap the car for a CNC machine, a packaging line or a plant’s fleet of forklifts and the pattern is identical: read the data you already have, spot the patterns, and let people ask questions in plain language. Then the same idea gets copied to the next plant, and the next. The unit of delivery shrinks from a project to a workflow, and time-to-value stops being a number you apologise for in the quarterly review.

How can low-code help bridge the skills gap?

The industry has spent years complaining about a shortage of automation engineers, as if they were hiding somewhere and could be lured out with a better coffee machine. They aren’t. You can’t hire your way out of this, so you have to upskill your way out.

The good news is that your workforce already has the hard part: they know how the plant runs. What they’ve lacked is a way to say it to a computer without learning a programming language. Plain-language tools let them read what an automation does, tweak it and test it. Your specialists, meanwhile, stop spending their days as a human help desk for routine requests and get back to architecture and the genuinely tough problems. Nobody has to learn to code. They just have to describe their process precisely, a skill anyone who has written a work instruction already has.

How can enterprises avoid creating yet another isolated layer?

The cure is a common integration layer instead of a spaghetti bowl of point-to-point links. For IT systems, MCP does that job. ERP, databases, ticketing, email, document stores and any other platform with an API can be exposed once as tools, then used by many agents and workflows. That’s the layer we focus on at MCP Admin: one place to connect, manage and monitor those tools, including multiple accounts of the same system, which matters when different plants or business units each run their own instance.

A common layer also settles an argument that never ends: which AI is best. One engineer swears by Claude, another by ChatGPT, a third lives in Codex or inside their IDE, and the debate is about as productive as a cricket-team argument at tea time. But everyone needs the same tools: the same ERP, the same databases, the same tickets. MCP Admin connects to all of those assistants, so people can use the one they like while the tools, permissions and rules stay in one place.

Let’s also be honest about scope. Real-time control logic in PLCs and SCADA should stay exactly where it is, doing its deterministic, safety-certified thing. Agent-based automation belongs above that layer, working with data that MES, historians and ERP already publish. The golden rule: the new layer reads from your systems of record. It never replaces them or goes around them.

What governance, cybersecurity and quality controls are essential?

Giving everyone the power to build automation without guardrails is like handing out forklift keys at a company picnic. Enthusiasm is high, and so are the odds of somebody driving through a wall. An AI agent with broad access to enterprise tools is powerful, and powerful things need adult supervision.

So here’s what a grown-up setup looks like:

• Least privilege. Give each agent only the tools its job needs. An agent that summarises reports has no business deleting records, any more than the office intern needs the keys to the server room.

• Action-level approval rules. Not all actions are equal. Reading a report can run on autopilot. Creating a purchase order, publishing content or deleting data should wait for a human to say yes. Rules should work per tool and per payload, so you can say: auto-approve below a threshold, require approval above it, and block this one entirely.

• Approval where the approver is. Managers live on the plant floor, not glued to a desk. Requests should reach them on a messaging app, with the option to wait for an answer or be notified later.

• A full audit trail. Every tool call should record who or what triggered it, what was asked and what was decided. When something goes sideways, “we’re not sure what happened” is not an answer anyone wants to give a regulator.

• Central standards, distributed building. IT and security own the connections, permissions and approval policies. Business teams own the workflows on top.

That’s the principle behind MCP Admin: freedom to build, plus control over what agents are allowed to do, whichever assistant is asking. Keep sensitive data in its system of record, show agents only what a task needs, and test before production. Hope is not a security strategy.

How should enterprises measure success, and what does a roadmap look like?

If your success metric is “we launched a platform”, congratulations, you’ve measured the wrong thing. Measure outcomes instead, in four groups:

• Speed. Time from idea to working automation, and from pilot to scaled rollout.

• Productivity. Hours of manual work removed per workflow, and the share of routine requests handled without raising an IT ticket.

• Quality and risk. Error rates, approval-rule hits, blocked actions and audit findings. A pile of blocked actions in the early days isn’t failure. It means the guardrails are doing their job.

• Adoption. The number of active builders outside IT and how often connected tools get reused. If only three people use it, you’ve bought an expensive hobby.

As for the roadmap, don’t boil the ocean:

1. Foundation (first quarter). Connect a few low-risk IT systems, define the permission model and approval policies, and pick two or three pilot workflows with named owners.

2. Expansion (next two quarters). Open the doors to more plant and business teams, add systems gradually, and turn winning pilots into shared templates.

3. Scale (after that). Standardise across plants and units, review the numbers regularly, and treat governance as a living process that tightens or loosens as agents earn trust.

The enterprises that win with democratised automation won’t be the ones that hand everyone the biggest toolbox. They’ll be the ones that made it safe to hand out toolboxes in the first place, because the fastest way to democratise something is to make sure it doesn’t blow up.

(The views expressed in interviews are personal, not necessarily of the organisations represented.)

Aayush Kumar is the CEO and Founder of Zyven Technologies, the company behind MCP Admin (mcpadmin.cloud), a platform for connecting, managing and governing the tools that AI agents use across enterprise systems. He works on making AI-driven automation accessible to non-developers while keeping it secure, auditable and under human control. He is also a 2023 winner of iCreate’s SparkUp Fund for a connected-vehicle monitoring concept that combines OBD2 data, GPS and machine learning on AWS.