Data vs. Commerce

Everyone agrees the legacy PIM has hit its ceiling. Almost nobody wants to own the move off it.

ㅤ

In the third and final part of the PIM series, Matt Johnson and Floyd Blaikie get into the part the first two episodes set up. How do you actually move? Jay Roxe of inriver and Willem Van Dijk of Pivotree both argue the move has gotten faster and cheaper in the space of a year, because AI now sits at every stage of the delivery cycle rather than one.

ㅤ

That creates a different problem. If standing up a platform that looks like your old one is now easy, the temptation to lift and shift gets stronger exactly when it's least appropriate. Jay's position is that B2B buyers are already researching and selecting with AI, so a migration that preserves the old model preserves the old invisibility. Floyd pushes on the credibility question from the buyer's chair, and Matt names the valley of pain in the middle of every implementation where the schedule quietly goes.

ㅤ

👤 Guests

Jay Roxe is Chief Marketing Officer at inriver, the product data platform. He came up through software engineering and product before marketing leadership roles at Microsoft, GE, athenahealth, Rapid7, and BitSight, and most recently HYPR. He lives near the Boston Marathon course and has run it.

ㅤ

Willem Van Dijk is Director of Data Platforms & Services at Pivotree. He has spent close to twenty years in master data management and PIM, including multiple leadership roles at Stibo Systems where he opened the retail vertical and later the Australia and New Zealand business.

ㅤ

📌 What We Cover

• Where AI now sits in a PIM delivery cycle: needs gathering, user stories and the grooming of them, build instructions, test scripts written as the build happens, and post-delivery support triage.
• Why Pivotree can quote fixed bid now, and what has to be true about delivery for that to work.
• Why writing user stories with AI surfaces the workflow that actually exists, including the Tuesday afternoon email from Bob to Mary that no system has ever recorded.
• Jay on lift and shift as the wrong goal, and what happens to discoverability if you carry the old model across.
• The line between free content and governed, on-brand, compliant, regulatory approved content, and why only one of them needs infrastructure.
• Floyd on the person who has been burned before, and what has actually changed since their last migration went sideways.
• Jay's customer survey: about 65 customers, more than two-thirds cutting product data maintenance time by more than 30%, and 75% launching faster.
• The valley of pain in the middle of every implementation, requirements bloat when the light bulbs go off, and what a good SI does about both.

ㅤ

🔗 Resources Mentioned

• inriver product data maturity research, covering more than 400 manufacturers and distributors
• Pivotree
• LinkedIn Live event with Jay and Willem, Wednesday September 2nd, linked from the Data vs. Commerce, Pivotree, and inriver LinkedIn pages
• MCPs, referenced by Jay on moving prototypes into production
• Contact: [email protected] and [email protected]

Everybody has an AI pilot. Almost nobody has a written path for getting one into production, and that gap is where this solo episode of Data vs. Commerce sits. Matt Johnson runs this one alone, with Floyd Blaikie out of the studio for the week.

ㅤ

Matt keeps hearing the same sentence from data, commerce, and supply chain leaders: we built this, we proved it works, and we can't figure out how to get it into production. His answer isn't a better tool or a smarter consultant. It's a stage gate process, six gates adapted from the years he spent leading R&D at a manufacturing company, applied to AI experiments in digital commerce.

ㅤ

The friction this week isn't data against commerce. It's experimentation against governance, and Matt argues most companies make the trade backwards. They restrict who gets to experiment, then leave the road to production undefined. He flips it. Train more people, give them permission, and put the checkpoints at the gates instead of at the door. He also prices out the free-for-all: duplicated work, abandoned projects, and tokens burned with nothing to show for them. Matt is a digital commerce leader at Pivotree.

ㅤ

📌 What We Cover

• Why pilots stall in silos, with the data team running one experiment, marketing running a pet project, and operations testing something nobody else has heard about
• The counterintuitive move: empowering more people to experiment, not fewer, because the ideas that matter come from whoever is closest to the friction
• What happens with no guardrails at all, from 50 uncoordinated experiments to burning through tokens like crazy and wondering whether AI is returning anything
• Gate one: a business sponsor, a rough business case, and a working prototype, and why none of it has to be precise yet
• Gate two and the council: the Shark Tank demo, real metrics like hours saved and costs reduced, and a budget approval that covers tokens and internal time
• Gate three and four: running the pilot as a controlled science experiment, 10,000 SKUs through AI and through the traditional process, then letting the data decide, including the decision to kill it
• Gate five as a checklist: documentation, integration with legacy systems, who owns the nine o'clock request, and whether it complies with your data governance policies
• Gate six and the seventh that isn't really a gate: adoption plan, success metrics, feedback loop, and continuous improvement

ㅤ

🔗 Resources Mentioned

• Pivotree
• OpenAI
• Anthropic
• Claude Code

Product data enrichment is one of the most manually intensive jobs in commerce, and in B2B distribution it scales badly: hundreds of supplier feeds, a million-plus part numbers, and a normalization standard that varies category to category and supplier to supplier. Which is why every vendor now claims to fix it with AI.

ㅤ

Matt Johnson and Floyd Blaikie take opposite ends of that claim in this solo episode of Data vs. Commerce, the show from Pivotree. Floyd takes the data side, worried about AI being confidently incorrect when your ability to sell depends on the data being right. Matt takes the commerce side, having started his career copy and pasting supplier data out of sales sheets into spreadsheets and ultimately into a PIM.

ㅤ

Then Matt tells the story that complicates both of their positions: a large national distributor ran an AI enrichment tool built for industrial supply against Claude, which isn't marketed as a product data enrichment tool at all. Claude killed it, and it wasn't even close. The reason came back to what made product data teams successful 20 years ago.

ㅤ

📌 What We Cover

Why the gap between the MSCs and Amazon B2Bs of the world and mid-market distributors is a data labor gap, not a technology gap

ㅤ

The customer line Matt can't shake: enrichment is table stakes, enrichment is a commodity, PIMs do it, offshore teams do it, even the birds and bees do it

ㅤ

Why enrichment experiments come back as a total cost of ownership wash, with QA and correction time replacing the time saved

ㅤ

The head-to-head test where Claude beat a purpose-built industrial supply enrichment tool, and what the winning team did first

ㅤ

The documentation that made it work: a defined set of attributes and acceptable values, tier one attributes marked required, and a task that fails if the model deviates

ㅤ

Floyd's argument with her seven-year-old about Gemini and the ice cream store, and the magic wand problem it exposes

ㅤ

Why distributors doing hundreds of millions in sales stall out when two internal groups can't agree how to standardize a part type

ㅤ

Why the enrichment conversation starts with your director of sales and head of e-commerce, reverse engineering context from the downstream channels

Everyone is watching the front end. The AI shopping assistants, the chat interfaces, the storefront that talks back. Pivotree's Q2 retail report went looking at what retailers are actually doing instead, and the answer was hiring. Across more than 1,600 North American retailers tracked from April to June, one in five posted a product data role. That beat AI roles, commerce platform roles, and developers.

ㅤ

Matt Johnson hosts this one solo, with Floyd Blaikie out for the week. His guest is Dan Ornstein, head of retail at Pivotree and the person behind the report. Dan takes the data side, hard.

ㅤ

His argument: humans forgive a mislabeled product photo and buy the blue one anyway. An LLM doesn't. If your attribute list says blue and the shopper's agent is looking for gray, you don't have a gray shirt, and you'll never find out you lost the sale. Matt pushes on where the money is actually going, and the answer sits in the back office, not the storefront.

ㅤ

👤 Guest Bio

Dan Ornstein leads the retail industry practice at Pivotree, the company's largest segment, working with major retailers across Europe, North America, and Asia. Before Pivotree he was a partner at KPMG Canada and a director at Publicis Sapient, with years spent on post-merger integration. This is his second appearance on the show. His first was on AI search.

ㅤ

📌 What We Cover

• Why one in five of 1,600 tracked retailers posted a product data role, and what that says about where the real bottleneck sits
• The attributes LLMs now want that never mattered for SEO: fabric, recyclability, sustainability, sourcing, reviews, return policies
• Humans forgive a mislabeled color. An LLM looking for gray finds blue and moves on, and you never learn it happened
• Why AI investment in retail is landing in catalog audits, taxonomy cleanup, and price reconciliation across channels rather than the storefront
• A nice new storefront on a cracked foundation still can't answer price, SKU-level availability, or delivery timing
• The consolidation tax: what DICK'S and Foot Locker, or Ferrero and WK Kellogg, actually have to reconcile in the data layer before any synergy shows up
• Why a loyalty merge that works for 75% of the base is a merge you can't announce
• Why smaller distributors often stop at a financial rollup instead of a true integration
• Dan's starting point for retailers behind the ball: top 100 products, top 10 categories, check every channel a machine reads

ㅤ

🔗 Resources Mentioned

• When the Shopper is a Machine, Pivotree's Q2 retail report
• Pivotree, including the agentic readiness assessment Dan mentions
• dataversuscommerce.com for past episodes
• Kroger's AI Shopping Assistant
• ChatGPT, Perplexity, Gemini, Copilot
• DICK'S Sporting Goods and Foot Locker; Ferrero and WK Kellogg

Product data used to be a back-office chore. Now it decides whether an AI recommends you, whether a customer can find a replacement part, and whether you show up in the consideration set at all. Most teams have no real picture of what is happening down in that foundation.

ㅤ

Floyd Blaikie hosts this one solo while Matt Johnson is on vacation, going underneath the question part one left open: is it the platform holding you back, or the data beneath it?

ㅤ

She's joined by Jay Roxe, who runs marketing at inriver, and Willem Van Dijk, Pivotree's director of data platforms and services. Jay is the guy who thinks about the PIM. Willem is the guy who thinks about everything holding it up.

ㅤ

Both took the data side. The friction surfaced anyway, over who owns product data (nobody agrees) and how far you can trust AI with it (not far, yet).

ㅤ

👤 Guest Bios

Jay Roxe is Chief Marketing Officer at inriver, the PIM software company serving brands, manufacturers, and distributors. He came up through product and marketing leadership at Microsoft, GE, athenahealth, Rapid7, and HYPR. This is his second appearance on the show.

ㅤ

Willem Van Dijk is director of data platforms and services at Pivotree, where he runs data audits, remediations, and MDM and PIM implementations. He spent years at Stibo Systems in roles spanning ANZ managing director and VP of platform strategy for product data syndication.

ㅤ

📌 What We Cover

• A disgruntled employee pushing bad content live because nobody controlled who input, reviewed, and approved publication
• The $250 million revenue hole caused by unpublishable replacement parts data, and the gray market that absorbed the demand
• Why spare parts break first: fitment information, interchanges, and country of manufacture changing over a part's life
• The information supply chain, and the layer teams skip, which is control over upstream people rather than technology
• Ownership versus stewardship, and why IT and marketing both think product data is theirs
• Trust signals in a PIM: taxonomy, fill rates, outlier dashboards, and anomaly detection across correlated attributes
• What running on Excel actually looks like from the inside, and the third of manufacturers and distributors keeping more than 85% of publish-ready SKUs available
• PIM plus an MCP server: distributors sending agents to shop for content, with no aggregator remapping taxonomies in the middle
• Where human in the loop is still non-negotiable, and why launch-day clean is not the same as solid

ㅤ

🔗 Resources Mentioned

• inriver
• Pivotree, and [email protected]
• Microsoft Excel, and Clippy
• Claude
• MCP servers

Swagelok is running 97% of order to cash through automated multi-agent AI, lifting purchase orders straight into SAP. Most manufacturers and distributors are still walking printouts from one desk to another. The distance between those two sentences is the whole episode.

ㅤ

Matt Johnson and Floyd Blaikie host this one solo, ahead of the Pivotree State of the Industry report for industrial manufacturing and distribution through Q2.

ㅤ

Floyd takes the data side. Machines are ingesting your catalog whether you signed off or not, 67% of B2B buyers now want a rep-free experience, and the strongest hiring signal of the entire quarter was product data. Matt presses from the commerce side, where relationship selling still wins deals and the order counter is quietly turning into an API.

ㅤ

They arrive at the same uncomfortable place from opposite directions. Order management is where AI goes first in industrial, product data decides whether it works, and being behind on digital no longer keeps you out of agentic commerce. It just means you lose inside it.

ㅤ

📌 What We Cover

• Why the report's headline is that order management goes first, and what makes quote to cash the most manual process in the building
• The 97% number from Swagelok, and the change management and governance that had to happen before anyone could realize it
• Why 71% of B2B businesses offering e-commerce is not the same as a digital experience from start to finish
• Whether Gartner's $15 trillion by 2028 is real, or number crunchers making things up for slide decks
• The strongest hiring signal of Q2: one in seven companies across more than 700 tracked in North America posting product data roles, PIM roles, and engineers to build AI agents
• Why product data is where back-office AI automation breaks, and what happens when catalog knowledge lives as tribal knowledge inside the inside sales department
• Grainger as the value proposition mid-market distributors have to compete with, and what's usually missing behind the rep
• The downstream problem manufacturers haven't priced in: if you're invisible on distributor websites, AI doesn't consider you a viable option

ㅤ

🔗 Resources Mentioned

• Pivotree State of the Industry report, industrial manufacturing and distribution through Q2
• Swagelok, on 97% of order to cash through automated multi-agent AI into SAP
• Graybar, on AI ordering and quote assembly
• Gartner and Daryl Plummer, on AI agents mediating more than $15 trillion in B2B spending by 2028, and on custobots
• Grainger and Sonepar
• Episode 6 with Lauren McCullough, Co-Founder and CEO of Tromml, on using AI in field sales
• datavscommerce.com, where the report will be posted alongside show notes and transcript

Everyone feels the ceiling before they can name it. The channels keep multiplying, the attributes per product keep growing, and the whole operation is held together by workarounds and one very large spreadsheet. On this episode of Data vs. Commerce, hosts Matt Johnson and Floyd Blaikie from Pivotree sit down with Jay Roxe, CMO of inriver, who takes the data side.

ㅤ

The friction: buyers are now building their shortlists from structured product data, not marketing copy, and most product orgs can't see the sales they're losing because the digital tells are gone. Jay argues PIM is shifting from a system of record into a system of work. Matt pushes on the commerce reality downstream, where an incomplete catalog forces the buyer back to calling a rep.

ㅤ

👤 Guest Bio

Jay Roxe is Chief Marketing Officer at inriver, a product information management platform. An engineer turned marketing leader, he's spent his career defining new software categories, and he's focused on the shifts pulling product data to the center of how people buy. On this episode he takes the data side: get the product information right, or lose the sale before it starts.

ㅤ

📌 What We Cover

• Why "hitting a ceiling" shows up first as workarounds and spreadsheets, not as a clear tooling decision
• What changed post-COVID: channels more than doubled, and so did the attributes each product carries
• How AEO and GEO moved product data into the buyer's journey, with Claude, ChatGPT, and Gemini building shortlists from structured data
• The aftermarket parts problem, and why a broken fridge clip is the whole thesis in miniature
• PIM as a system of work, not just a system of record, and where orchestration and governance actually live
• The risk when product knowledge is "consolidated in somebody's head" and that person wins the lottery
• Two exercises leaders can run this week: staple yourself to an order, and get an AEO read on how you show up
• AI as a multiplier for merchandising, not a replacement for the team

ㅤ

🔗 Resources Mentioned

• Inriver (inriver.com), including its AEO assessments
• Excel
• ChatGPT, Claude, Gemini
• ERP, PIM, MCP (referenced throughout)

Every pitch for agentic commerce opens with the model. Which frontier lab it runs on, how smart it is, how many PhDs sit behind it. This episode argues the model is the part that matters least.

ㅤ

Hosts Matt Johnson and Floyd Blaikie of Pivotree's Data vs. Commerce sit down with Anatolii Iakimets, Director of Product Marketing at Kibo. Anatolii takes the data side: an agent is a model plus a harness, the model is becoming a commodity, and the real work sits in the data and integration layer underneath.

ㅤ

The friction is clean. Floyd keeps reframing agentic commerce as a familiar platform decision, agents in a trench coat. Anatolii keeps pulling it back to the data. Commerce is deterministic. The price and the tax and the T-shirt size have to be exact, and an agent is only as good as the structured data it can reach and read.

ㅤ

👤 Guest Bio

Anatolii Iakimets is Director of Product Marketing at Kibo Commerce, where he focuses on B2C commerce. He has spent more than a decade in commerce and telecommunications, with earlier roles at Bold Commerce, Elastic Path, and Netcracker, and he has worked hands-on with AI and machine learning since around 2015. On this episode he takes the data side, arguing that agentic commerce succeeds or fails on data quality and integration, not on the model you pick.

ㅤ

📌 What We Cover

• The two parts of any agent: the LLM and the harness, the code that tells the model what to do
• Why the model is becoming a commodity, and why the switching cost for users stays low
• Why coding and text agents tolerate a slightly different answer every time, and commerce does not
• Why the price, the tax, and the T-shirt size have to be accurate to the point, not "good enough"
• Why an agent has to be integrated like an application, talking to your systems through MCP or an API
• How burning tokens turns into a real budget problem, with Uber's four-month AI burn as the warning
• The three things to weigh before you buy: composability, simplicity, and the ability to start small
• Why the "explain" function is the one customers reach for first, and the risk of a three or five year lock-in

ㅤ

🔗 Resources Mentioned

• Kibo (the guest's company)
• Anthropic (Claude) and OpenAI (ChatGPT) as frontier model providers
• Google as a frontier lab
• DeepSeek and Kimi K2 as open-weight models
• Model Context Protocol (MCP), APIs, and agent-to-agent protocols
• Microsoft Copilot's shift from subscription tiers toward usage-based pricing
• Uber's AI budget story
• Sam Altman on models becoming "intelligence on a tap"

Every AI conversation in distribution right now assumes the data underneath is ready to be acted on. Most of the time, it isn't. Matt Johnson hosts this one solo, with Floyd Blaikie out of the studio, joined by Ryan Smith, an account executive at Pivotree, fresh off the Applied AI for Distributors event in Chicago run by Distribution Strategy Group.

ㅤ

Ryan took the data side. Before you buy the next AI tool, you finish step zero and step one: knowing where you're actually headed, then cleaning and governing the data those tools depend on. He and Matt get into why last year's shiny purchases went sideways, why distributors keep walking up to the table saying they're behind, and what happens now that customers show up with prototypes they built themselves. It's a distribution and manufacturing conversation, but retail and B2B operators will recognize the pattern.

ㅤ

👤 Guest Bio

Ryan Smith is an account executive at Pivotree, where he works with distributors and manufacturers on their AI and data transformation. He joined the company recently and spends his time helping companies figure out where they actually stand before they buy the next tool. He took the data side of this episode, arguing the foundation has to come before the technology.

ㅤ

📌 What We Cover

• Why most of last year's AI tool buyers hit a wall, and how much of it traced back to data that was never cleaned or governed
• Step zero, step one, step two: mapping where a distributor actually is before recommending anything
• Distributors building their own tools in-house, with AI engineers vibe coding bespoke solutions
• The security and governance risk when two departments build the same agent and spend the budget twice
• Customers arriving with ready-made prototypes built in Claude Code, then handing off the integration
• Why a working prototype is not a multi-tenant, secure, production-ready system
• The generational shift from tribal knowledge walking out the door to buyers who expect the Amazon experience
• AI fatigue, the 50/50 reliability problem, and why the human element still decides the outcome

ㅤ

🔗 Resources Mentioned

• Distribution Strategy Group (DSG) and the Applied AI for Distributors event
• Claude Code
• The earlier Data vs. Commerce episode with Bill Di Nardo and Joel Farquhar on RI plus AI
• Matt's breakout session on AI catalog management, on Pivotree's YouTube channel

Field sales is the part of commerce that data never quite reaches. The rep has the relationship, the context, and forty years of counter knowledge, and most of it dies in a notebook nobody reads. This episode is about closing that gap. Matt Johnson calls in from the AI for Distributors event in Chicago while Floyd Blaikie hosts two guests from Pivotree's automotive world: Lauren McCullough, co-founder and CEO of Tromml, and Pivotree's Sam Russo. Lauren takes the side most software founders won't: the human relationship is the asset, and AI exists to make it sharper, not replace it. The friction that comes out of it is where you point AI in a high-trust business, and where you keep it out.

ㅤ

👤 Guest Bio

Lauren McCullough is co-founder and CEO of Tromml, a vertically focused software company for the automotive aftermarket and industrial distribution. Tromml started on the analytics side, giving distributors visibility into what products were actually making money, and now builds field-sales tooling that captures rep conversations and turns them into next best actions. Lauren argues for keeping relationship-driven selling human while using data and AI to support it.

ㅤ

📌 What We Cover

• Why reps in distribution aren't just selling parts, they're selling relationships and trust, and why that changes how AI fits
• The difference between a system of record (put information in, pull a report out) and a system of action (surfaces the next best action and gets smarter over time)
• A rep's actual Monday workflow: prioritized accounts, optimized routes, a human-digestible briefing before the visit, and voice-note capture after
• The field signal that never reaches the boardroom, like a shipping delay or a launch that isn't landing, because nobody reports bad news up the chain
• Hiring implications: when the data lives in the system, you hire for emotional intelligence instead of category memory
• Where to point AI in a high-trust industry, and Lauren's blunt take on automating the customer relationship away

ㅤ

🔗 Resources Mentioned

• Tromml (Lauren's company; mobile conversation-capture app for field reps)
• Pivotree
• AI for Distributors event, Chicago (Matt's reference; the event's formal name is Applied AI for Distributors, run by Distribution Strategy Group)
• Claude (AI note-taking tool referenced by Sam)
• Salesforce (referenced as a CRM)
• Genuine Parts Company (referenced by Sam; see flag below)