Manta Industrial: AI-Native Systems Engineering in the Field

There’s a category of engineering company that barely appears in case studies but is responsible for a significant share of the world’s deployed automation: the systems integrator. Not the OEM that builds the robot. Not the software vendor that sells the SCADA platform. The integrator who takes all of it—conveyors, vision systems, PLCs, safety controllers, custom end-of-arm tooling, customer-specific process requirements—and makes it work together in a real factory, on a real timeline, for a customer who will not accept a delayed production start.

Manta Industrial is that kind of company. Based in the Midwest, Manta designs and deploys custom automated systems for manufacturing customers across automotive supply, consumer goods, and food processing. Their programs range from single-station robotic weld cells to multi-cell flexible assembly lines integrating dozens of subsystems from different vendors. Each program is, in engineering terms, a unique system. And each customer has their own requirements vocabulary, their own safety expectations, and their own definition of what “done” looks like.

That combination—high variability, high complexity, and real regulatory stakes—makes requirements management not a paperwork exercise but a core engineering discipline. Manta started using Flow Engineering roughly eighteen months ago. What follows is an account of why they made the change, what they’ve found works, and where the tool still has room to grow.

The Problem with Managing Requirements at Scale Across Diverse Programs

Before Manta adopted Flow Engineering, their requirements process looked like what most mid-sized integrators use: a combination of Word documents, Excel-based requirement lists, and PDF deliverables exchanged with customers. For a single-cell program with a short schedule, that works. The team is small enough that the lead engineer knows where everything is. When someone asks whether the requirement for guarding access is covered, the answer lives in someone’s head.

The problem compounds fast when you’re running six programs simultaneously, each for a different customer, each with a different requirements baseline, and each progressing through a different phase of the engineering lifecycle. At that scale, the document-based approach produces three failure modes that Manta’s engineering leadership had seen repeat across programs.

The copy-paste drift problem. Manta builds similar systems for different customers. A palletizing cell for a consumer goods manufacturer and a palletizing cell for a food processor share 70% of their engineering content. But if that shared content lives in separate Word documents, changes propagate by finding all the copies and updating them manually. The copies drift. What was a common requirement becomes customer-specific variations that nobody tracked intentionally.

The subsystem boundary problem. An industrial automation system is not a software system with a clean API boundary. When a mechanical design change affects the clearance envelope of a robot, it may invalidate a previously accepted guarding design, which may have been the basis for a safety function claim under ISO 13849. That chain of implications crosses the mechanical, electrical, and software domains. In a document-based world, those documents are maintained by different engineers and rarely linked explicitly. The connection exists in someone’s understanding, not in the artifact.

The compliance demonstration problem. IEC 62061 and ISO 13849 are not checkbox standards. They require integrators to demonstrate that safety functions have been identified, that the architecture of each safety function has been designed to meet a claimed performance level or safety integrity level, and that verification activities are traceable back to those claims. Assembling that demonstration from fragmented documents at the end of a program is expensive and error-prone. It also tends to reveal gaps—requirements that were implied but never formally captured, verification activities that were performed but not linked to any claim.

None of these problems are new. What’s changed is that the programs are getting more complex, the schedules aren’t getting longer, and customers in safety-regulated sectors are asking harder questions about compliance evidence earlier in the program.

What Manta Was Looking For—and What They Ruled Out

Manta’s engineering leadership evaluated the standard options. IBM DOORS and DOORS Next are the dominant enterprise tools in safety-regulated industries, and for good reason—they have deep traceability capabilities, broad adoption in automotive and aerospace, and well-established compliance workflows. But for a company running programs of Manta’s scale, the implementation and administration overhead of a DOORS deployment is substantial. DOORS Next, in particular, requires configuration that assumes a dedicated tool administrator and a relatively stable process model. Manta has neither. Their process has to flex to each customer’s requirements format and phase gate structure.

Jama Connect was a closer fit in terms of deployment simplicity, and Manta ran a pilot. Jama’s review and approval workflow is well-designed, and its reuse features reduce some of the copy-paste drift problem. Where it fell short for Manta was in the multi-domain traceability model. Jama’s data model is item-based and relationship-typed, which is adequate for software requirements management but becomes awkward when you’re trying to model the relationship between a mechanical interface specification, an electrical load requirement, and a software safety function—across a customer program with its own requirement hierarchy on top.

Codebeamer is worth naming here because it’s increasingly common in functional safety contexts, and it handles the combination of requirements, risk assessment, and V-model verification workflows with more integration than most tools. For teams already in an automotive Tier 1 environment with established Codebeamer processes, it’s a serious option. Manta’s concern was the same as with DOORS: configuration overhead relative to program volume and team size.

How Flow Engineering Fits Manta’s Model

Flow Engineering’s core architectural decision—treating requirements and their relationships as a graph rather than a document hierarchy—is the property that made it relevant to Manta’s situation.

In a graph-based model, a requirement doesn’t belong to a single document. It can participate in multiple traces simultaneously. A guarding interface requirement can be linked to the customer’s top-level safety specification, to the mechanical envelope requirement that governs it, to the IEC 62061 safety function it contributes to, and to the test case that verifies it—all without copying the requirement or choosing which document it “really” lives in. When the mechanical envelope changes, the engineering team can traverse the graph and identify every downstream artifact that needs review. The system makes the dependency visible; it doesn’t rely on someone’s memory.

For Manta, that property directly addresses the subsystem boundary problem. Their programs involve requirements contributed by mechanical engineers, electrical engineers, PLC programmers, and safety engineers—often in parallel, often across different tools and formats. Flow Engineering’s import and linking capabilities let Manta’s systems engineers connect those artifacts into a unified program graph without forcing every discipline onto the same authoring tool.

The multi-program management capability matters too. Manta maintains what they call a reference architecture library—a set of validated requirement blocks and safety function templates that recur across programs. In Flow Engineering, those blocks exist as reusable nodes in the graph. When Manta starts a new palletizing program, they instantiate from the reference architecture, then extend and specialize for the customer’s specific requirements. Changes to the reference architecture can be propagated explicitly, with review, rather than through manual find-and-replace across documents.

AI-Assisted Gap Detection: Where the Practical Value Shows Up

Flow Engineering’s AI capabilities are the feature Manta’s team reports as the most operationally surprising—not because AI assistance in requirements is a new concept, but because the specific application to gap and inconsistency detection has direct value in their workflow.

The use case is straightforward to describe and harder to replicate manually. When a new customer requirement arrives—often stated in the customer’s language, at whatever level of abstraction the customer’s requirements team used—Flow’s AI analysis can identify whether the requirement is covered by existing architecture, whether it conflicts with a previously captured requirement, and whether it implies derivative requirements that haven’t been captured yet. That analysis happens at intake, not at integration test.

Manta’s senior systems engineer described a specific example from a recent food processing program. The customer’s requirements included a statement about wash-down compatibility for the end-of-arm tooling that was phrased in operational terms, not IP rating terms. Flow’s AI flagged that the requirement implied a specific IP69K compliance need that wasn’t present in the electrical enclosure requirement baseline. That gap, caught at requirements review, took an afternoon to resolve. Caught at factory acceptance test, it would have meant tooling rework and a delayed customer delivery.

That kind of catch isn’t guaranteed, and Manta’s team is clear-eyed about it: the AI analysis is a filter that improves the odds of catching gaps early, not a guarantee that no gaps exist. The value is probabilistic. But in a program model where integration test surprises are the most expensive failure mode, shifting even a fraction of gap detection earlier in the cycle has measurable impact.

For IEC 62061 and ISO 13849 compliance, Flow Engineering’s traceability graph provides the evidence structure that safety auditors expect. Manta can demonstrate, from the graph, the chain from hazard identification through safety function specification through architecture justification through verification. That demonstration used to be assembled manually from document artifacts in the weeks before a program closeout. Now it’s a query on the live graph, updated as the program progresses.

Where Flow Engineering’s Focus Shows

Flow Engineering is not a full-lifecycle ALM platform. It doesn’t include integrated change control workflows with the configuration management depth of a tool like Polarion. It doesn’t have the breadth of out-of-the-box compliance templates that some teams expect from a tool positioning in functional safety. For Manta, neither of those gaps has been program-blocking—their change control process uses separate tooling, and their safety templates live in the reference architecture library they’ve built in Flow. But integrators evaluating Flow Engineering for larger programs, or programs with more prescriptive customer-mandated tool chains, should assess those dimensions explicitly.

What Flow Engineering has optimized for—an AI-native requirements graph that scales across programs without administrative overhead—is exactly the value proposition that fits Manta’s operating model. That’s not a coincidence; it’s a deliberate scope decision that makes the tool genuinely useful for a specific kind of engineering team.

What Other Integrators Should Take From This

Manta Industrial’s experience points to something broader than a tool selection. The systems integrator sector has historically lagged larger OEMs and Tier 1 suppliers in requirements engineering maturity, not because the engineers are less capable, but because the tools available were sized for organizations with dedicated tool administration teams and multi-year implementation budgets.

AI-native tools designed for practicing engineers—not for tool administrators—are changing that calculus. The barrier to rigorous requirements traceability, including the traceability that safety standards require, is no longer primarily a tooling cost problem. It’s a process adoption problem, and that’s a more tractable challenge for a company like Manta.

The integration complexity of industrial automation systems isn’t going down. Customer expectations for compliance evidence aren’t going down. The engineering teams responsible for delivering those systems need tooling that scales with program complexity without adding process overhead that consumes the margin the program doesn’t have. Manta’s adoption of Flow Engineering is one data point in what looks like a broader shift in how mid-market systems integrators approach requirements engineering—and it’s a shift worth watching.