Moving deliberately rather than fast: How brands and retailers are achieving AI success
Celigo and MIT research reveals 90% of firms with AI workflows in production rely on integration platforms — meaning brands attempting standalone AI deployment are statistically likely to fail at scale. This sponsored piece from Modern Retail (March 2026) signals that the AI-in-retail conversation has shifted from 'should we adopt' to 'how do we operationalize without breaking our P&L.' The 'least agency' framework — using deterministic rules-based automation first and reserving AI for genuinely complex steps — is emerging as the de facto standard for brands that are actually shipping results. For marketplace operators, this directly impacts order exception management, inventory workflows, and cross-channel catalog orchestration costs.
The non-obvious play here is that the 90% integration-platform dependency stat is a margin compression signal disguised as a technology insight — every brand that tries to bolt AI onto fragmented system architecture (separate Shopify, Amazon, Walmart, eBay connectors with no unified data layer) will pay 2-3x in implementation costs and operational errors before seeing ROI.
The competitive moat erosion risk is real: brands that nail integration infrastructure NOW will automate order exceptions, catalog updates, and ad feed management at a fraction of the labor cost their competitors carry into Q3 2026.
A $10M/year seller should immediately audit whether their tech stack has a single integration layer or a spaghetti of point-to-point API connections — because the latter makes every AI tool they buy essentially a expensive toy.
The decentralized-initiative-with-centralized-control model described here is exactly what separates 8-figure operators from 7-figure ones in 2026.
This article is a leading indicator that the AI-in-commerce shakeout of 2026 will split operators into two camps: those with clean, integrated data pipelines who can deploy AI incrementally, and those spending budget on AI tools that can't talk to each other.
It fits squarely into the platform consolidation trend — Amazon, Walmart, and Shopify are all building AI-native seller surfaces that reward operators with structured, real-time data feeds.
The 'move deliberately, not fast' framework is also a regulatory hedge: as AI governance scrutiny increases in the EU and cautiously in the US, brands with auditable, human-in-the-loop workflows will face fewer compliance surprises than those who automated aggressively without guardrails.
Audit your integration layer THIS WEEK: Log into your ERP or OMS and count how many point-to-point API connections exist between your sales channels (Amazon Seller Central, Walmart Seller Center, Shopify, eBay). If you have more than 5 direct integrations with no middleware hub, you are in the 10% that will fail at AI deployment — get a Celigo, Pipe17, or Boomi demo scheduled before end of month.
Identify your top 3 manual workflows by headcount hours — specifically order exception management, inventory reconciliation, and listing updates across channels. Map which steps are rule-based (automate now with deterministic logic) versus judgment-based (AI candidates). Build a 2-column doc and bring it to your ops meeting this week with a 90-day automation target per workflow.
In the next 30-60 days, Amazon and Walmart will continue pushing AI-native seller tools (Amazon's generative listing creation, Walmart's AI-assisted content) — brands without a unified product data layer will be unable to feed these tools consistently across SKUs, creating catalog quality gaps that suppress organic rank. Start building or buying a PIM (Product Information Management) system now if your product data lives in spreadsheets.
Bottom Line
90% of AI winners run integration platforms first — your spaghetti stack isn't an IT problem, it's a margin problem.
Source Lens
Industry Context
Useful background context, but lower-priority than direct platform, community, or operator intelligence.
Impact Level
medium
90% of AI winners run integration platforms first — your spaghetti stack isn't an IT problem, it's a margin problem.
Key Stat / Trigger
90% of firms with AI workflows in production rely on integration platforms
Focus on the operational implication, not just the headline.
Full Coverage
Sponsored // March 18, 2026 Moving deliberately rather than fast: How brands and retailers are achieving AI success By Celigo It’s easy enough to talk about how AI can benefit brand and retail operations, but the extent to which brand and retail organizations are ready to successfully utilize the technology is a different conversation entirely.
And AI readiness looks different for every brand and retailer, depending on what data they have access to, what kind of shape that data is in and whether they’re in a position to be truly successful in integrating AI across their operations on their own or if they need help to do so — especially considering that recent Celigo and MIT research found that 90% of firms with AI workflows in production rely on integration platforms.
Ahead of this year’s ShopTalk event, Modern Retail sat down with Ronen Vengosh, Chief Strategy Officer at Celigo, the intelligent automation platform that unifies the predictable and the fully agentic, to talk about what cross-functional brand and retail success looks like in the era of AI, how brand and retail organizations are successfully scaling the technology and the kind of framework that sets brands and retailers up for integrating AI.
As commerce evolves in the era of AI, what does it look like to safely apply the technology where it matters most for brands and retailers? Ronen Vengosh: The brands and retailers getting the most value from AI are not the ones moving fastest — they are the ones moving most deliberately.
The right starting point is identifying processes that carry a high manual workload and a clear, measurable outcome, then committing to automating that entire process end-to-end rather than just sprinkling AI on top of individual tasks.
Critically, this is also an opportunity to rethink the process itself, not simply replicate what already exists at higher speed.
Once you have defined the new process, it is worth asking how much of it can actually be handled by deterministic automation — rules-based logic, workflow orchestration, system integrations — and reserving AI for the steps that genuinely require it. That “least agency” mindset reduces error surface and makes the system auditable.
Finally, keep humans in the loop, especially early on and in any area where errors are costly. That is not a sign of distrust in the technology, it is how you build the confidence to scale it responsibly. Why is it key to move AI out of IT and into every department for brands and retailers to truly achieve cross-functional success in the era of AI?
Ronen Vengosh: AI will ultimately reshape every function in a brand or retail organization — merchandising, supply chain, finance, marketing, customer service — and treating it as an IT project is one of the most reliable ways to slow that down.
The teams closest to the business problems have the context to identify where AI will actually move the needle, which is a very different skill from knowing how to deploy infrastructure.
That said, democratizing AI deployment across departments does not mean abandoning governance — the risk of fragmented, ungoverned AI proliferating across a brand or retail enterprise is real.
The better model is one where IT brings the AI tools and technical knowledge, the “Art of the Possible,” while establishing the guardrails, integration standards and data pipelines, which allows business units to drive the use cases and own the outcomes.
That balance — decentralized initiative with centralized control — is what allows AI adoption to scale without creating a compliance and security problem in the process. Can you talk about what AI for process automation looks like in real workflows? How does this apply to the idea of a connectivity-first approach to AI?
Ronen Vengosh: A useful example is order exception management — a process that in most brand and retail operations involves people manually triaging failed orders, cross-referencing inventory systems, checking with suppliers and updating the customer. No company’s change management is perfect and everyone is trying to move faster in today’s world.
An AI-augmented version of order exception management connects the relevant systems first, so the AI has reliable, real-time data to work with, then uses that data to triage automatically, propose resolutions and escalate only the cases that require human judgment.
The sequence matters: AI sitting on top of disconnected, siloed data does not produce reliable outcomes; it produces confident-sounding errors.
A connectivity-first approach means building the integration layer before the intelligence layer — establishing clean, orchestrated data flows between ERP, order management systems, warehouse management systems, CRM and commerce platforms — and then deploying AI into a workflow where it can act on information it can actually trust.
That foundation is what separates AI that performs in
Original Source
This briefing is based on reporting from Modern Retail. Use the original post for full primary-source context.
Style
Audience