Episode 153

Orium: Why Most Brands Should Buy, Not Build, Agentic Commerce

Jason Cottrell
Jason Cottrell
Founder and CEO

In this episode we talked about:

  • How Jason’s consultancy, Orium, and the MACH Alliance are positioning brands for agentic commerce
  • Why AI does not make composable commerce obsolete, and the almost six-times difference in AI success the MACH Alliance sees with composable foundations
  • How brands on Shopify, Salesforce, or BigCommerce can get agent-ready without replatforming
  • Why a big custom agentic roadmap is the approach Jason worries about most
  • What is real in agentic shopping today: product discovery in answer engines, with transactions a matter of when, not if
  • Three questions to ask every vendor about answer engines, owned experiences, and operations
  • Where agents work best in operations: product data, search, order fallout, and fraud
  • Why super agents that are too broad fail, and why five or ten narrow use cases beat one big blob
  • What Jason expects to take off next year in consumer shopping and in secure agent-to-system connections

🎧 Listen now on Apple Podcasts, Spotify, or YouTube

Episode highlights:

02:22 Does Agentic AI Make Composable Commerce Obsolete?

05:05 Getting Agent-Ready Without Replatforming

07:53 What’s Real In Agentic Commerce Versus Theoretical

13:04 Where Agents Work Best In Ecommerce Operations

16:26 Why Agentic Projects Fail: Going Too Broad

17:49 What Takes Off In Commerce Next Year

Jason’s bottom line: Most brands do not need to build their agentic commerce stack. They need to buy it well. Jason Cottrell, Founder and CEO of Orium and President of the MACH Alliance, says the vendors are already spending millions to productize what most brands are about to commission as a custom roadmap, and that is the approach he worries about most. His advice: ask every commerce partner what they are doing to get you listed and purchasable in answer engines, how they improve your owned site and app, and how they help you run operations with agents, and check whether their agents only work inside their own four walls. On the operations side, agents do best where work is repeatable but variable, like product data, search, order fallout, and fraud. The common failure is going too broad, with super agents nobody trusts or uses. Jason’s rule: pick five or ten narrow use cases, prove the ROI, and let someone else evolve and own the capability.

FAQ

Jason Cottrell is the Founder and CEO of Orium, a systems integrator and consultancy that has been around for about 15 years and focuses on agentic commerce and the infrastructure that will run autonomous commerce and operations. Orium spent over a decade building composable commerce, and Jason says it is the largest pure-play consultancy in the Americas doing that work. He is also the President of the MACH Alliance, an industry association of about 125 members, 250,000 practitioners, and 4,000 brands that engage in its programming and events each year.
No. Jason says that for brands doing over $10 million, and especially over $100 million, the question is not whether to go composable, because they already run more than one system for experience, commerce, data, and fulfillment. The question is how composable to be. The MACH Alliance’s research found almost a six-times difference in AI success for companies with proper composable foundations compared with those without. Systems that can be extended, vendor onboarding muscle, better data, and teams already through change management all help.
Jason says most probably do not need to replatform yet, especially in this first phase. Modern commerce platforms are investing heavily in productized capability, and a number of upstart add-on vendors can help bridge how you publish your catalog and accept orders. His advice is to ask your commerce partner what they offer, explore about three of the upstarts, and then decide who adds the capabilities you need. Vendors that work well with plug-ins and other add-ons have an advantage.
He says it is the approach he worries about most right now. Firms that commission a big, complicated custom roadmap are spending a lot on capabilities that are very quickly being productized, and they are unlikely to get the return. His view is that you can probably buy the capability, work with a partner to put it in use, prove the ROI, and let someone else evolve and own it for you.
Jason says there is not yet much completed checkout happening inside assistants like Gemini, ChatGPT, or Claude, but product discovery and choosing between options is already something the majority of Americans do there, and the trend is rising. He sees transactions as a matter of when, not if. The core job for brands today is getting their products to appear in answer engines, both from training and on demand, and platforms and specialty vendors are forming one-to-one relationships to help with catalog distribution. Over the next couple of years, brands should also think about payments, loyalty, and dynamic experiences in those channels.
First, what are you doing to help me get listed and be purchasable within answer engines, and where is your roadmap going? Second, how are you helping me on my owned experiences, like my website and app, and raising the bar for what customers can do there? Third, how do you help me run my business smarter with agents and automation? Jason adds a fourth check: whether a vendor’s agents only work inside its own four walls, or work across the six or so products a brand actually uses to run campaigns.
Jason says agents do well where work has some repeatability but some variability. Examples include search and product discovery for complex or B2B catalogs, product data management and tailored product descriptions, personalization, conversational clienteling and customer service, order fallout and the customer communication around it, and fraud detection. He notes that fraud cuts both ways, since agents also give fraudsters new tools, and that these use cases are largely productized, with repeatable ROI available from vendors.
Going too broad. Jason has seen brands spend millions or tens of millions on internal super agents for employees that are too broad to train, specialize, or get right, and consumers only use small parts of the experience. He also cautions against building very intricate consumer-facing agentic experiences before shoppers are used to them. His lesson for 2026 is to pick five or ten narrower use cases rather than one big blob of capability, like your own mini ChatGPT.
On the consumer side, a more substantial group of shoppers doing their shopping inside answer engines and wanting to transact in conversational experiences, helped by new personal agent launches from Meta, OpenAI, and Google. He says much of the technology and standards to make those channels transactable are already in place. On the operations side, he expects more secure, direct connections between agents and the data and systems they act on, replacing today’s copy-and-paste handoffs, because that is where much bigger process gains are unlocked.
Toolbox listeners get an exclusive free trial of Noibu’s AI agents for CRO, A/B testing, performance, accessibility, ROAS, and bug resolution. The agents start by auditing your store and ranking the biggest opportunities, from slow pages to broken journeys. Then they build each change, wait for your approval, ship it, and measure what it did. The trial runs until you have at least two changes live and measured. Get started at noibu.com/podcast-audit.

Want to be featured on The Ecommerce Toolbox?

We feature senior ecommerce leaders with real things to say. If you’ve led something worth talking about, we want to hear from you.

Apply to be a guest →

Other episodes

See all episodes

Audit your site today

Supervised AI agents find what's broken, slow, or leaking revenue, do the work, and hand it back for your approval.