Just 37% of US ecommerce decision-makers would trust AI to make purchasing decisions
In the Linnworks Market Survey 2026, conducted by OnePoll, 58% of US respondents would trust AI to generate demand forecasts, while 37% would trust it to make purchasing decisions. Among sellers trying to automate more, poor integration between systems (43%) is cited more often than cost (37%), underscoring how connected data and systems shape automation efforts.
This story was produced through MarketScale. See how Retail teams put it to work with Sales Enablement.
Key takeaways
Trust in AI among US ecommerce sellers drops from 58% for demand forecasting to 37% for purchasing decisions, even though 62% report an AI assistant already acts for them (for example, sending messages or updating records). Separately, only 37% would trust AI to communicate with suppliers.
Forty-three percent cite a lack of integration between systems as a barrier to greater automation, compared with 37% who point to cost. Just 10% cite unclear return on investment as a barrier to greater AI adoption.
With 46% of US respondents expecting AI and automation to be included as standard in operations software, the buying question could move from whether a tool has AI to which actions it can take without human sign-off—and how approvals are set.
Free workspace
Turn your Retail expertise into content.
Record interviews, organize footage, and write with AI on a free trial of the MarketScale platform for qualifying companies. No demo required, no credit card.
Fifty-eight percent of US ecommerce decision-makers in the Linnworks Market Survey 2026 would trust AI to generate demand forecasts. Just 37% would trust it to make purchasing decisions.
That gap sits at the center of the survey, which Linnworks released Nov. 12, 2026. OnePoll ran the online survey for Linnworks from July 28 to August 4, 2026, polling 500 ecommerce decision-makers, with 250 in each of the US and the UK, selected at random from OnePoll's panel. Each respondent has a say in choosing operations and fulfillment software at a business that makes most of its sales through ecommerce and brings in annual revenue of approximately $1.3 million to $130 million.
Every figure in the release comes from the 250 US respondents, and the results are unweighted. Linnworks builds the kind of software these respondents buy. The company says it has been in business for 20 years, serves 4,000+ brands and retailers and connects inventory, orders, warehouses and listings.
For 64% of respondents, AI ranks among the most important factors in picking operations software. Yet fewer than four in ten would let it make purchasing decisions. That leaves open the question of what buyers expect the AI in that software to do.
Where sellers stop handing over the keys
Linnworks says AI adoption has moved past the experimental stage. Some respondents are also becoming more cautious, and concerns about data and security are part of the reason. In the release, Linnworks CMO Hilary Smith described the change: ecommerce businesses have stopped debating whether they should take up AI, and are now deciding which tasks they are willing to let it lead.
AI in US ecommerce operations, 2026
Linnworks Market Survey 2026, conducted by OnePoll
Among US respondents, 62% report that an AI assistant already acts for them, for example by sending messages or updating records. Only 37%, though, would trust AI to communicate with suppliers. So the objection is not to AI acting at all. Sending an internal message is apparently fine; sending one to a supplier is a different matter.
Look at which tasks pass and which don't. Fifty-eight percent would trust AI to generate demand forecasts, but confidence drops when money and external relationships are involved. The pattern could line up with that split more than with technical difficulty.
For an operations leader planning an AI rollout, that offers a practical way to sort workflows. Ask whether each one's output leaves the building, and how easily it can be reversed if it does.
IT teams draw the same line
Ecommerce sellers aren't alone. On Oct. 19, Grafana Labs released findings from its State of Observability, 2026 report, which David Ramel covered for Redmond. Grafana worked with research firm ETR to poll 300 senior IT and engineering decision-makers who oversee spending on observability technology.
ETR ran the survey in June 2026 and limited it to decision-makers at director level, vice president level or above. The United States supplied 150 respondents, and Japan, Latin America and Western Europe supplied 50 apiece. The results suggest comfort with AI falls in a similar place: the moment it moves from advising to acting.
The cliff falls between recommending and doing. In the Grafana Labs survey, 85% of IT leaders are comfortable with AI doing read-only investigation, but 50% are comfortable with it deploying production changes. In the Linnworks survey, 58% would trust AI to generate demand forecasts, while 37% would trust it to make purchasing decisions. One survey asked about comfort and the other about trust, and the respondent groups differ. Still, both suggest people accept AI that informs a decision well before they accept AI that executes one.
The endpoints alone understate how sharp that cliff is. At 84%, comfort with AI recommending a fix is barely lower than comfort with investigation. The figure falls to 56% once AI is making code or configuration changes and to 52% for fully autonomous remediation. In the announcement, Grafana Labs co-founder Anthony Woods put it plainly: having AI diagnose a problem is meaningfully different from allowing it to change something in production. Put a purchase order where production is and his point fits the Linnworks data well.
The Grafana data also hints at what could narrow the gap. Reliability is still the top barrier to adoption, named by 26% of respondents. Next come regulatory or compliance constraints at 23%, then customer trust at 21%. That third item suggests IT leaders, like the ecommerce sellers, think about how AI's actions affect people outside the organization.
When respondents were asked what would make them trust AI more, two answers tied for the top spot at 45% each: explainability and a proven track record of accuracy. Neither is a feature you can switch on. A system earns both over time, by showing its reasoning and then being right.
Disconnected systems outrank price
The Linnworks data has a second finding that matters more to anyone with a software budget. Among sellers trying to automate more, 43% name poor integration between systems as a barrier, against 37% who name cost. On wider AI adoption the gap is bigger: integration complexity is the most common barrier at 21%, while only 10% cite unclear return on investment.
The US responses turned up other results as well. Over the past 12 to 18 months, 47% have moved into new sales channels or marketplaces. Some 41% name inventory accuracy and stock synchronization as a top area where added automation would help. And 57% now use EDI with their trading partners.
Read together, those figures sketch a seller with more channels, more trading-partner connections and stock numbers they want automation to keep in sync. That would matter for the trust gap. A purchasing decision rests on an accurate stock count, so if inventory doesn't reconcile across channels, a cautious buyer has a concrete reason to keep a person on the order.
In the release, Smith said AI can help teams make smarter calls about demand, stock levels and daily operations. She added that companies still have to be able to trust the data that those decisions rest on.
The next stage of AI adoption isn't about handing over control. It's about giving teams the right intelligence to act faster and make better decisions., Hilary Smith, CMO, Linnworks
The findings come with limits. Data and security concerns have made 34% of US respondents more cautious about AI. Every figure rests on 250 US respondents from businesses with approximately $1.3 million to $130 million in annual revenue, and the results are unweighted.
Built-in AI makes permissions the purchase
Among US respondents, 46% want AI and automation included as standard in their operations software instead of paying extra for an add-on. That expectation may line up with a broader trend. In research published Nov. 18, 2025, MIT Sloan Management Review drew on a global survey of more than 2,000 respondents plus interviews with executives. It reported that agentic AI reached 35% adoption in just two years, with another 44% of organizations planning to deploy it soon.
For comparison, the same research put traditional AI adoption at 72% after eight years and generative AI at 70% in three. According to the researchers, vendors were adding agentic capabilities to their offerings as features, so organizations put agentic AI in place before building a strategic management framework for it.
The MIT Sloan Management Review figures come from a global survey of more than 2,000 respondents across industries, not from ecommerce sellers. They describe how agentic AI is spreading through organizations broadly, so the link to the Linnworks findings is a parallel in how features arrive, not a measure of ecommerce adoption.
The MIT Sloan authors, Sam Ransbotham, David Kiron, Shervin Khodabandeh, Sesh Iyer and Amartya Das, name "supervision versus autonomy" as one of four tensions in adopting agentic AI. Traditional oversight models, they write, assume either full human control or complete automation, not systems that need some human control and differing degrees of automation. The Linnworks split between forecasting and purchasing reads like that middle ground in practice.
Put those together and the procurement question could change. If AI arrives inside the platform, buyers may find the key decisions are about setup: what the AI may do alone, what needs a person's sign-off, and who that person is. The researchers argue that competitive benefits will come not from early access to agentic AI but from superior organizational design around it.
For buyers, the findings suggest three questions worth putting to any vendor:
- Can approval thresholds be set separately for forecasting, reordering and supplier messages?
- Will the system show the stock data and logic behind a purchase recommendation, so a buyer can check it? IT leaders in the Grafana Labs survey named explainability as one of the things most likely to build trust.
- Does inventory reconcile across every channel and trading partner before the forecast runs?
The gap between the 58% who would trust AI to generate demand forecasts and the 37% who would trust it to make purchasing decisions is the number to watch in any follow-up research. It could show how far sellers will let a forecast run before a person signs the order.
Sources
Featured companies
Your experts belong here
Every story in MarketScale Retail starts with a company putting its merchandising leads, store operations teams, and category managers on the record. Buyers are already reading this topic. The only question is whose experts they find.
Category buyers trust operators, so your merchandising leads shorten the distance between first search and first call.
About the author
The MarketScale Newsroom reports on the companies, technologies, and trends shaping 16 B2B industries. It turns primary sources and expert commentary into clear, useful coverage for the people doing the work.