HCLTech's software division has completed its acquisition of Robotiq.ai for a reported €9 million, according to Croatian business outlet Konsulteer, roughly a month ahead of the November 2026 close originally expected. The deal adds enterprise robotic process automation (RPA) to HCL UnO Agentic, HCLSoftware's platform for orchestrating AI agents, bots, people and deterministic workflows. For industrial teams, the significance is not that HCLTech is buying into PLCs. It is that agentic AI is acquiring the execution layer it needs to reach the systems that surround the plant.

HCLSoftware announced its intent to acquire Zagreb-based Robotiq.ai on 28 September 2026 and said the transaction would close in November. Konsulteer reported on 27 October that the deal had already completed. HCLSoftware did not disclose the price in its announcement; reports put the transaction at €9 million. PetroPartsOnline could not independently verify the closing, and HCLSoftware had not published a separate completion statement at the time of writing.

Key facts at a glance

What HCLSoftware actually bought

Robotiq.ai is an enterprise RPA platform vendor headquartered in Zagreb. Its legal entity is Robotic Process Automation d.o.o., and its chief executive and co-founder is Darko Jovišić. The platform is used by large banks, insurance groups and telecom providers, and HCLSoftware highlights its ISO-certified security controls, audit logs and flexible deployment options.

It is not the same company as Robotiq Inc., the Canadian manufacturer of robot grippers and cobot accessories. The shared name is a coincidence that buyers occasionally confuse; the two businesses serve different markets and have no corporate relationship.

RPA is often described as "software robots" because it automates repetitive work by interacting with applications through their user interfaces or APIs. In banking and insurance, that means processing forms, reconciling records and moving data between systems that were never designed to talk to each other. In manufacturing, the same technique applies to legacy CMMS, MES and supplier portals.

HCLSoftware is the software arm of HCLTech, an IT services group with more than 223,000 employees and $14.8 billion in trailing-12-month revenue as of June 2026. HCLSoftware says it serves more than 20,000 organisations, including most of the Fortune 100 and almost half of the Fortune 500.

The destination for Robotiq.ai's technology is HCL UnO Agentic, which HCLSoftware launched in May 2025 as an extension of HCL Universal Orchestrator. The platform coordinates AI agents, bots, people and systems, and combines deterministic workflows with probabilistic agent decisions. Its tooling includes visual builders for agents and workflows, hybrid workflow management, and connectors to enterprise systems, ERP, mainframes and cloud applications.

Why an agentic platform needs RPA

Agentic AI is good at interpreting a goal, breaking it into steps and choosing a tool. It is less good at the last mile: logging into a 15-year-old supplier portal, reading a screen that has no API, or moving a record through an application that only accepts keyboard input. That is the execution gap HCLSoftware is buying.

In its announcement, HCLSoftware said Robotiq.ai's RPA capabilities allow AI-driven workflows to automate tasks in applications where APIs are unavailable or insufficient, extending HCL UnO Agentic's orchestration "from decision-making to execution." The company's president, Kalyan Kumar, framed the deal around production-scale automation that is "secure, governed and reliable" rather than experimentation.

Buying an installed RPA engine is faster than building one. It brings a mature runtime, UI connectors, audit trails and a customer base that already runs automation in regulated environments. It also brings the weaknesses of RPA: bots are brittle when screens change, they accumulate technical debt, and they need monitoring, version control and clear ownership.

The combination is more useful than either technology alone. An agent can decide that a purchase requisition is missing a supplier certificate; RPA can open the supplier portal, download the document and attach it to the record. The agent handles variation, while the bot handles the repeatable interface steps. Governance determines what the system is allowed to do without a human.

Where agentic AI fits in a plant: an ISA-95 view

Industrial automation is organised in layers. ISA-95, also published as IEC 62264, separates enterprise systems at Level 4 from manufacturing operations at Level 3, supervisory control at Level 2, control at Level 1 and the physical process at Level 0. Agentic AI and RPA fit comfortably in the upper layers. They do not belong in the control loop.

The timing difference matters. A PLC executes logic in milliseconds and must behave deterministically every cycle. An agent may take seconds or minutes to reason, call a tool and write a result. That makes agents suitable for supervisory and business workflows, not for closed-loop control, motion control or safety instrumented functions.

Most plants already have the data path: OPC UA, historians and asset databases expose condition and production data to Level 3 systems. The missing piece is usually the action layer — the ability to open a work order, check stock, request a quote or update a supplier record. That is where RPA and agentic orchestration can add value without touching the PLC program.

A realistic maintenance workflow

Consider a compressor train monitored by Bently Nevada condition-monitoring systems, including 3500 series machinery protection systems. A vibration channel crosses an alert threshold. In a conventional workflow, an engineer reviews the trend, checks the asset history, confirms whether the signal is real, creates a work order, checks spare availability and contacts procurement.

An agentic workflow can prepare most of that work:

1. The historian raises an event and the agent retrieves the asset tag, maintenance history and recent work orders.
2. The agent checks whether the excursion matches a known pattern and pulls the relevant procedure for human review.
3. It queries the CMMS for open work orders and the ERP for spare-part stock, including analog input modules such as the Allen-Bradley 1756-IF8 or other I/O and HMI and operator interface hardware.
4. If a part is missing, RPA opens the supplier portal, checks lead time and drafts an RFQ. If the CMMS has no API, the bot creates the work order through the screens a planner would use.
5. The planner receives a package — evidence, asset context, proposed work order and quote request — and approves or rejects it.

The human still decides. The agent removes the administrative loop around the decision, which is often where maintenance teams lose hours. The same pattern applies to industrial power supplies, drive replacements and other routine MRO purchases.

What it means for MRO and spare-parts teams

The near-term value is not autonomous purchasing. It is faster, cleaner preparation. Agents can cross-reference legacy part numbers, flag discontinued revisions, compare lead times and assemble a quote request before a buyer opens the ERP. RPA can push that request into portals that still require manual entry.

For plants running mixed estates, the useful targets are the brands with long lifecycles and deep installed bases: Siemens S7 controllers, Schneider Electric Modicon and Altivar ranges, ABB AC 800M and drive systems, Honeywell ControlEdge and safety I/O, and Allen-Bradley ControlLogix and CompactLogix platforms. A bot that can read a nameplate, search a catalogue and populate a requisition is more valuable when the catalogue spans dozens of series and revisions.

The constraint is master data. If part numbers, descriptions and asset tags are inconsistent, the agent will automate the inconsistency. Plants that clean their asset register and standardise spare-part descriptions before deploying automation get better results than plants that start with the bot.

Engineering approval also remains necessary. Form, fit and function are not interchangeable across revisions, and a cheaper equivalent can be wrong for a safety function. An agent can propose; an engineer should approve. For examples of the hardware involved, see variable-speed drives such as the Schneider ATV71, CompactLogix controllers such as the 1769-L30ERM and 32-channel DC input modules such as the 1756-IB32/B.

The limits: safety, determinism and cybersecurity

Determinism

Probabilistic systems are unsuitable for functions that must produce the same output every cycle. Control logic belongs in a PLC or DCS programmed to IEC 61131-3, with validated changes and a known scan time. Agents can recommend a setpoint, but they should not write one directly.

Functional safety

Safety instrumented systems are designed, verified and maintained under standards such as IEC 61508 and IEC 61511. Their independence from the basic process control system is a design requirement, not an implementation detail. An agent must never sit inside a safety loop, and RPA must never be used to bypass an interlock or simulate a safety signal.

OT cybersecurity

RPA bots are privileged identities. They log into applications, hold credentials and perform actions at machine speed. That makes them an attractive target and a potential insider risk. IEC 62443 provides a useful framework: segment IT and OT zones, define conduits, apply least privilege, vault credentials, enforce multi-factor authentication where possible and log every action with a user, timestamp and input.

Data quality and accountability

Agents inherit the quality of the data they retrieve. A wrong asset tag or a stale spare-part record can produce a confident but incorrect recommendation. Audit logs, human approval gates and a clear escalation path are not optional features; they are the controls that make the automation usable in an industrial environment.

How to evaluate an agentic automation pilot

Most failed pilots start too broad. A better approach is one workflow, one site and one measurable outcome, with a manual fallback that can be restored immediately.

A practical 90-day sequence is: two weeks to map the current process and data sources, four weeks to build and test with historical cases, four weeks to run in parallel with human approval, and two weeks to review exceptions and decide whether to scale. The review should include maintenance, IT and OT security. If any group cannot explain what the agent did and why, the pilot is not ready to expand.

What to watch next

• Integration depth. The interesting question is not whether Robotiq.ai remains a standalone product, but how quickly its runtime and connectors appear inside HCL UnO Agentic workflows.
• Pricing and packaging. Enterprise RPA is usually licensed by bot, process or consumption. HCLSoftware has not published post-acquisition packaging.
• Industrial references. Banking and telecom references prove the technology works at scale. Manufacturing references will show whether the same governance model survives OT constraints.
• Partner ecosystem. HCLTech's engineering services arm can implement the platform, but plants often buy through system integrators and automation distributors. Channel coverage will determine reach.

The €9 million price is small for a company of HCLTech's size. The strategic signal is larger: IT services and enterprise software vendors are moving toward the execution layer that connects AI decisions to real systems. For plant teams, the practical response is to prepare the data, the approval paths and the audit trail now, so that agentic automation can be adopted on their terms rather than imposed by a vendor roadmap.

For more coverage of automation, maintenance and supply-chain technology, see our industrial automation news and analysis. To discuss spare-part availability or a replacement project, contact the PetroPartsOnline team.

Frequently asked questions

Who acquired Robotiq.ai?

HCLSoftware, the software division of HCLTech, agreed to acquire Robotiq.ai on 28 September 2026. Konsulteer reported on 27 October 2026 that the transaction had closed, ahead of the November timetable given at announcement.

How much did HCLTech pay for Robotiq.ai?

Reports put the price at €9 million. HCLSoftware did not disclose financial terms in its announcement, so the figure should be treated as reported rather than confirmed by the buyer.

Is Robotiq.ai the same company as Robotiq Inc.?

No. Robotiq.ai is a Croatian enterprise RPA software vendor based in Zagreb. Robotiq Inc. is a Canadian manufacturer of robot grippers and cobot accessories. They are separate companies with no corporate relationship.

What is HCL UnO Agentic?

HCL UnO Agentic is HCLSoftware's agentic orchestration platform, launched in May 2025 as an extension of HCL Universal Orchestrator. It coordinates AI agents, bots, people and systems, and combines deterministic workflows with probabilistic agent decisions.

Will agentic AI replace PLCs or DCS?

No. PLCs, DCS and safety instrumented systems execute deterministic control in milliseconds and remain governed by IEC 61131-3 and functional-safety standards. Agentic AI is better suited to Level 3 and Level 4 workflows such as maintenance triage, procurement and reporting.

Can RPA work with industrial software that has no API?

Yes. RPA interacts with applications through their user interface, which is why it remains useful for legacy CMMS, MES and supplier portals. The trade-off is brittleness: screen changes can break bots, so version control, monitoring and ownership are essential.

What should plant teams do now?

Start with a read-only or human-approved workflow around maintenance or procurement, clean up master data, define audit and rollback requirements, and keep agents outside the control loop. A narrow pilot with a measurable outcome is more useful than a broad platform evaluation.

Sources

• HCLSoftware to Acquire Robotiq.ai, Strengthening Enterprise Agentic Automation, PR Newswire, 28 September 2026.
• Robotiq.ai, official company site and platform documentation.
• HCL Universal Orchestrator / HCL UnO Agentic, HCLSoftware product page.
• TMTPost analysis of HCLTech's agentic AI strategy, which reports the €9 million price.
• ISA-95 enterprise-control system integration standard, International Society of Automation.
• IEC standards for industrial automation and OT security, including IEC 61131-3, IEC 62443 and IEC 61511.

Editorial note: this article was prepared by the PetroPartsOnline editorial team. Product links are provided for context; the article is not a recommendation to deploy any specific automation platform without an engineering and security review.