Long seen as the "poor relation" of digital transformation compared with the ERP, PLM (Product Lifecycle Management) is establishing itself as the foundation of product data — and the prerequisite for any reliable R&D AI. At the Tech For Industry Show, five experts compared notes: Emily Lara Rosales, Global R&D PMO & PLM Director (Pernod Ricard), Laure Burtin, DDAI Strategic Office Director Data, Digital & AI (Danone), Sarah Stocco, Principal PLM Solution Consultant (Infor), Hélène Boileau, Senior CPG PLM Consultant (Accenture) and Frédéric Russo, SVP, General Manager Central & Southern Europe (Infor).

Every digital project promises a better time to market — so that's no longer a differentiating criterion for launching a PLM program. The compelling criteria: getting it right the first time and securing license to operate, with significant cost avoidance and cross-team efficiency as a result. A second issue: data, a potential AI accelerator — that's where real acceleration happens.

The world's second-largest wine and spirits company, Pernod Ricard constantly manages a wide split: aged spirits whose development can take decades, and unaged formulated products with a time to market of just a few months. With a market being reshaped — lower barriers to entry, blurred lines between spirits, soft drinks and food — those who succeed will be the ones who start from the consumer, their needs and occasions, beyond the traditional industrial view.

PLM centralizes all information and processes tied to a product, from design to delisting: recipes, formulations, packaging, bills of materials, specifications. It doesn't replace the ERP — it makes ERP execution more reliable. Sarah Stocco (Infor) takes the image further: PLM isn't the ERP's little brother but "the parent of product data." If upstream data isn't reliable, downstream only inherits that unreliability — and building in design to cost and design to sustainability early turns an imposed constraint into a design lever.

Beyond a single, reliable, centralized data source, one often-overlooked lever: the HR lever. When data is scattered across R&D, quality, SharePoint, Excel and ERP systems, teams exhaust themselves checking, searching, re-entering data — leading to frustration, even attrition. PLM frees up time for the core job: developing products. Add to that avoiding costly reworks and errors (incorrect nutritional labeling) for smooth, on-time, compliant launches — without overlooking the unquantified value in the business case.

Identified 4-5 years ago and raised by the field, the need found its triggers: globalizing the R&D function — moving from around twenty variably structured teams toward a global function sharing knowledge, projects and validated data, opening the door to reuse and platforming; the agility challenge posed by raw material and energy costs, US tariffs, and regulation that has become a lever of economic warfare — with structured specifications allowing a move from one site to another; and the ongoing ERP/master data transformation, which requires structured input data upstream.

Three years after launch, three key factors: top-level sponsorship ("otherwise we don't get started"); engagement across functions — this isn't an R&D-only project, and nothing builds buy-in better than quickly demonstrating value on two integration points; and positioning PLM as a cornerstone of design to delivery, down to the storytelling. The anecdote: at Danone, the project is called Cybèle — the equivalent of Rhea, goddess of creation, one of the twelve Titans — seated at the table of Themis… the name of the in-house ERP. "We won't go any lower in the hierarchy." Final advice: cut the elephant down to size — bill-of-materials integration started at one plant, then ten, before considering all 200.

At Pernod Ricard — "creators of conviviality," a group built through successive acquisitions where each entity developed its own culture — PLM transformation is first and foremost a process transformation, supported by a digital solution, not the other way around. Faced with low-turnover R&D populations, sometimes working in their native language and on Excel, announcing a global AI-powered tool can create anxiety: the benefit needs to be co-built with them, not imposed top-down. And the program leader must be patient and resilient — years can pass before the trigger point, and that's only the beginning.

Key success factors on the change management side: mapping stakeholders as early as possible (resistance isn't always where you expect it, and the best sponsor isn't necessarily the most senior — but the one who genuinely believes in the project); building a network of ambassadors — trusted relays in the field, not just trainers; and building a collective: a PLM project is cross-functional (R&D, marketing, quality, supply chain, regulatory), ideally led by a physical or virtual project team where business and IT speak the same language. Down to the collective choice of the project's name — the first step toward ownership.

Implementing a PLM is like "putting the church back in the middle of the village": establishing sources of truth, data accountability — and for a function, owning its data "means existing within the company." Good news: thanks to AI, operations teams have come to understand the value of maintaining data — ten years ago, the reaction was "why would I bother?" What remains is helping them take the next step: eliminating manual entry of ingredient and packaging specifications, via a supplier portal or semi-automation.

AI first to relieve low-value-added tasks — searching for information, summarizing a document, extracting data — since this friction is a daily source of frustration. Then agentic AI: conversing with agents and delegating — identifying why a workflow is stuck and proposing actions, or pre-simulating an ingredient change and its impact on the formula and finished product. And process mining applied to NPDI: analyzing where time is lost and where loops form — a deviation from the standard isn't always an error, sometimes it's a field practice worth adding to the process catalog. The fundamental shift, according to Sarah Stocco: "for a long time, we asked users to adapt to the solution; now, it's the solution that has to adapt to users" — speaking the language of the industry, and now that of the users. With one safeguard shared by all: if upstream data is weak, AI only amplifies that weakness — and creates a dangerous false sense of control.

Data first (where is it, who's responsible for it, what's it used for), then process, then only the tool — with people running throughout, from start to finish. PLM as a differentiation lever, at a time when every company already has an ERP. An opportunity for functions to talk to each other and progress together. And for those just starting out: map where complexity creates value — and where it doesn't. That's where the real pain points lie, a starting point specific to every company.

Watch the full panel, along with the other Tech For Industry Show replays, on our dedicated page.