Research
Go beyond the AI headline.
Agentic commerce, industrial AI and the work between a promising demonstration and a governed business decision. Company cases, deployments, benchmarks and practical implications—with sources, not borrowed promises.
142 research articles
Source-linked research on what has been built, what the evidence supports and what businesses can use. External examples are not OfferSignal client results. Each article keeps its source dates, evidence limits and practical implications.
Page 7 / 12
142 articles · Showing 73–84 · Page 7/12
Klarna's shopping assistant: useful discovery, unproven checkout authority
Klarna's 2024 shopping assistant put conversational product search and comparison inside its app. The case shows why relevant recommendations still need fresh merchant data—and why a shopping assistant is not evidence of autonomous purchasing.
Read research
L’Oréal Beauty Genius: personalised advice before the retailer hand-off
Beauty Genius shows how visual inputs, a curated first-party catalogue and conversational guidance can support product discovery. Its reported usage does not establish clinical accuracy, causal sales lift or autonomous buying authority.
Read research
Lazada AI Lazzie: catalogue-grounded shopping help is not autonomous buying
Lazada’s AI Lazzie shows how conversational product discovery, deal assistance and seller content generation can operate at marketplace scale while checkout, payment and publication authority remain unproven.
Read research
Lio: procurement execution needs a governed stop rule
Lio’s enterprise procurement case shows how multi-agent workflow automation can connect messy buying requests to existing systems, while its reported savings still require customer-level validation before they become a business case.
Read research
Livestream shopping’s AI assistant: orders rose, but authority stayed with the buyer
A randomized field experiment on an unnamed Chinese livestream-commerce platform tested an assistant that supplied product, policy and discount information. It found higher order counts and normalized spending, while showing why shopping assistance is not autonomous buying or proof of ROI.
Read research
Maersk and Keelvar: sourcing automation still needs buyer-controlled authority
Maersk’s named Keelvar deployment shows how optimisation and rules-configured RFx workflows can scale sourcing activity without establishing unconstrained machine buying. The practical lesson is to govern data, thresholds and hand-offs before expanding automation.
Read research
Magentic: procurement agents need governed data before delegated action
Magentic’s manufacturing-procurement case shows how agents can connect contracts, transactions and supplier evidence to find leakage, while leaving negotiation, purchase-order and payment authority explicitly controlled.
Read research
Makersite: product intelligence before the binding offer
Makersite shows how manufacturers can connect BOM, supplier and lifecycle data to design and sourcing decisions, while its customer cases also show why data quality and human approval remain essential.
Read research
Mango Stylist: conversational discovery is not autonomous buying
Mango’s nine-market fashion-assistant launch shows how a retailer can add conversational product discovery while leaving purchase, payment and post-sale authority unproven. The useful lesson is to govern product facts, permissions and evaluation before expanding the chat surface.
Read research
Marqo: product discovery can improve relevance without becoming procurement authority
Marqo has evolved from an open-source multimodal search engine into a managed ecommerce discovery suite. Its reported customer gains are promising but provider-authored, so industrial teams should test retrieval and relevance while keeping compatibility, price, stock and quotation decisions elsewhere.
Read research
Mastercard Agent Pay: authenticated checkout is not autonomous procurement
Mastercard’s Australian pilot shows how a payment network can identify an agent and bind a consented card transaction. It does not yet show reliable product selection, broad customer access or measurable commercial performance.
Read research
Mercanis: procurement agents can prepare the RFQ, but buyers still approve the award
Mercanis combines intake, sourcing, supplier management, contracts, risk and analytics in a commercial procurement suite. Its named cases report meaningful sourcing and time outcomes, while the evidence still leaves industrial requirements, pricing and autonomous authority to validate.
Read researchFrom evidence to work
Bring a real business question.
OfferSignal connects product facts, discovery and governed enquiry work for established manufacturers and B2B/B2B2C businesses. A focused workshop can help your team choose what to test—and what not to automate.
OfferSignalBusiness & AI Workshop