Generative engine optimization for physical products: Riikka Söderlund of Katana on why AI search cannot see what you cannot count

Generative engine optimization is usually discussed as a content problem. Riikka Söderlund's argument is that for anyone selling a physical product it is an operations problem, and the deciding field is whether the thing is in stock.

Söderlund is COO of Katana Cloud Inventory, a Finland-based platform running inventory, manufacturing and multi-channel sales for small product businesses. Merchants in more than 70 countries move over $3 billion of goods through it. She is also the first COO I have interviewed who came up through marketing rather than sales, and she thinks that is a missed opportunity for the rest of the field.

The Oprah problem

Her origin story is the whole thesis in miniature.

A decade ago she ran digital and communications for a Finnish beauty brand, pitching products to publications like Vogue. Around 2014 they landed a body butter on Oprah's Favorite Things.

It sold out the next day. And they could not get more shipped from Finland to the US for the rest of the holiday season.

A perfect demand event, converted into almost nothing, because the operation was not ready. That gap between attention and fulfillment is the thing she has spent the rest of her career closing.

What marketers know that operators do not

Her claim is that the old skills did not become useless. They became rare.

Learning to build a brand and tell a story before social media existed forces a particular discipline, and underneath it is a general one: influencing people. Teams, customers, partners. Packaging a thing and making the case for it.

She argues most go-to-market COOs come from sales and therefore under-invest in narrative. And that the payoff is not only external. A COO is constantly positioning the company to its own teams, facilitating the story the company tells itself, getting everyone running toward the same ending.

Then she corrects the obvious misreading. The most important storytelling skill is listening, because you cannot tell a story if you do not know what the other side cares about. Most of the time people are not saying what they are actually asking, and interpreting that is the skill she uses every day.

Her point about founders is sharp and self-aware. Founders often have natural gravitas and can skip the craft. She is an operator, brought in to scale things, so for her it had to be learned.

Training for the job before anyone gives it to you

Asked how she learned to be a COO, her answer is that she started doing it first.

The path she describes is a ladder: brand marketing to growth marketing to revenue marketing to revenue operations. Understanding where the money actually comes from and how to optimize it. Her claim is that if you can run the operation of how money flows in, you can run any operation in the company.

That included deliberately training as a Salesforce specialist while already a VP, in order to genuinely understand the tooling, the funnel and conversion rather than supervising them from a distance.

What she is training for now is product management, on the grounds that building products has fundamentally changed and she wants to understand the fundamentals. What surprised her is how much of it is the same job: listening to the customer's voice and understanding the need. In marketing she was communicating to that need. In product she is building to it.

Six week sprints under quarterly OKRs

Katana runs standard quarterly OKRs, reviewed rather than necessarily reset each quarter, because initiatives do not respect calendars.

The addition is hers. She asks teams to execute in six week increments. Four weeks is too short to see an impact. Eight weeks is too long, because things have already changed.

Her reasoning is that OKRs set direction and contain nothing about execution. So a big, scary key result gets broken into six week attempts: here is what we are going to try to move pipeline, and in six weeks we all see what worked, what did not, and share it.

She traces the habit to marketing, where that is simply how you run a campaign. It is a testing framework laid over a goal framework, and it is the most portable idea in the episode.

Why SAP has been the leader since 1974

Her framing of the market is one I had not heard.

SAP has led inventory management since 1974. She asks what other software category has had the same leader for fifty years, and the answer is none.

Katana is not competing with SAP. It is filling the gap underneath it. Buying SAP means hundreds of thousands of dollars and months of implementation, which puts it out of reach of most small brands. Katana's founders hit exactly this: they were running a physical product business, went looking for software to manage inventory, found nothing suitable, and started the company nine years ago.

What the product does starts with visibility. What do you have, how much, in which warehouse, how much is free to sell across channels, when do you reorder or manufacture, what materials are you missing to make it in time. Optimization comes second.

Her observation about first conversations is the telling one. Customers often only see their own workflow for the first time when someone maps it, because they have never connected the dots between selling a candle here, back-ordering a raw material there, manufacturing in a third place and shipping from a fourth.

Product data is the new SEO

Here is the part every operator should sit with.

Consumers are increasingly searching with AI. Even when they reach Google, the AI results answer the question without a click. Traffic to online stores is falling and everyone running one knows it.

So the only thing legible to an AI chat is your product data. ChatGPT, Claude and Gemini do not care about your photography, your branding or your homepage. They care what the product is, what it costs, what the dimensions are, and whether it is in stock.

And that last field is decisive. These systems are learning, and they will not surface a product that is out of stock. Game over, as she puts it.

Which means a stockout is no longer just a lost sale. It is removal from the surface where discovery now happens.

On timing, she is measured rather than alarmist. She does not think many people are buying inside the LLM yet. But the referral traffic is real: Shopify reported on its last earnings call that AI referral traffic to its stores grew 13-fold quarter over quarter. Her framing is the one that lands. Ten years ago the question was whether your brand exists if it does not show up in Google. Now it is whether it exists if it does not show up in ChatGPT.

She also thinks the full shift will be slow, noting it took roughly twenty years for e-commerce to overtake brick and mortar. Her advice is to follow the pace rather than going all in, and if you serve a niche, to watch what that audience does rather than the broad market.

Opening the API while SAP closed it

Katana and SAP made opposite bets. SAP closed its API to protect its own copilot. Katana opened it.

She is disarming about how that decision got made, which is that it did not. It was never an executive discussion. The platform has been extendable since day one, because customers know their business better than Katana does, and the same logic applied to letting them point whatever AI they want at their own data.

The part that is a genuine values choice: they believe the customer's data is the customer's. Locking a business out of its own inventory data is not something they were willing to do.

What that enables is what she calls conversational inventory management. Connect Claude to your inventory and ask where the shipment of candles is, how long it will take, how long it took last month.

The golden era of creativity

Her counterintuitive claim is that this is a good moment for branding.

AI generates infinite mediocre content, ads and branding, because anyone can produce it. The volume is tiring people out. Which means the things that actually stand out get attention, and standing out has always been how branding worked.

AI is not good at sparking emotion. That still comes from people. And because so many brands are defaulting to generic output, the space for a genuinely good idea has widened rather than narrowed.

The invisible inventory

My favorite question produced two stories at opposite ends.

The worst: a brand that did not know its cost of goods sold was negative. It was selling at a loss without realizing, and it did not end well.

The best: a brand that started with Katana this spring and found $40,000 of inventory it did not know it had. Simply misplaced. For a small business that is transformative.

Her reaction is the point. She comes from marketing, where everything is visible and measurable down to the individual ad and email. The idea of operating without knowing what you have is, to her, genuinely difficult to comprehend. How can you be a COO if you do not know what you are operating with?

The 5 things I took away from this conversation

1. A stockout now deletes you from search. This reframed the whole category for me. Out of stock used to mean a missed sale that you recover next week. If AI systems will not surface unavailable products, inventory accuracy has quietly become a demand-generation problem owned by operations.

2. Six week execution increments under quarterly goals. OKRs set direction and say nothing about how. Riikka's fix is to make teams commit to what they will try in six weeks, because four is too short to read a result and eight is too long to still be relevant. I am stealing this.

3. The most important storytelling skill is listening. Not presenting, not writing. Knowing what the other side actually cares about, and noticing that people rarely say what they are really asking. That reframing makes internal communication a research problem rather than a performance one.

4. Train for the job before it is offered. Riikka taught herself Salesforce as a VP so she could genuinely understand revenue operations, and is learning product management now. The through line is following where the world is going and building the skill ahead of the title.

5. Your customer's data is your customer's. SAP closed its API to protect its own assistant. Katana opened it and did not even convene a meeting about it. That is a values position with a commercial consequence, and I suspect the open side wins.

FAQ

What is generative engine optimization? The practice of making a business visible inside AI-generated answers rather than traditional search rankings. For product companies, Söderlund's argument is that this comes down to structured product data, because AI systems evaluate what a product is, its price, its dimensions and its availability.

Why does being out of stock hurt AI search visibility? Because these systems are learning what is genuinely available and will not surface a product that cannot be bought. Söderlund's point is that this turns inventory accuracy from a fulfillment concern into a discovery concern, since an out-of-stock item is effectively invisible.

Is AI search actually driving commerce traffic yet? Shopify reported AI referral traffic to its stores growing 13-fold quarter over quarter. Söderlund does not believe many consumers are completing purchases inside an LLM yet, and treats it primarily as a discovery channel whose share is rising as Google click-through declines.

What should a COO learn from marketers? Narrative building, according to Söderlund. The ability to articulate a value proposition and package it compellingly, which applies internally to aligning teams as much as it does externally to customers. She notes most go-to-market COOs come from sales and under-invest in this.

Why has SAP led inventory management since 1974? Söderlund raises this as a genuine anomaly, noting no other software category has had the same leader for fifty years. Her explanation for the gap Katana fills is cost and complexity: SAP implementations run to hundreds of thousands of dollars and months of work, putting them beyond small and medium businesses.

Also mentioned

• Katana Cloud Inventory, and its open API position
• Shopify, and the 13-fold quarter over quarter growth in AI referral traffic
• SAP, market leader in inventory management since 1974
• Oprah's Favorite Things, and the Finnish body butter that sold out before it could be restocked
• Cannes Lions, and the arrival of AI as a category in creative awards
• Linda Tong of Webflow on answer engine optimization, the companion conversation on the citation side of this shift

Listen to the full episode

Riikka Söderlund on Between Two COO's

Between Two COO's is hosted by Michael Koenig. Subscribe on Apple Podcasts, Spotify, or wherever you listen.