What the review examined
The authors investigated how research on multi-agent logistics, reinforcement learning, and autonomous coordination informs the emerging use of agentic AI in quick-commerce last-mile fulfilment.
Q-commerce depends on real-time coordination of orders, local inventory, pickers, couriers, routes, customers, and service recovery. The review asked what autonomous systems can already do in that environment, what benefits the literature reports, and where governance, accountability, labor, fairness, and customer trust remain underdeveloped.
The review design
A Scopus search returned 29 records. After title and full-text screening, 16 studies published between 2016 and 2026 remained. Thirteen were excluded because they addressed adjacent topics without a clear connection to quick-commerce fulfilment.
The researchers used VOSviewer for bibliometric mapping and Evidano for the qualitative synthesis. The combined design let them compare keyword structure with the concepts and arguments contained in the full texts.
Where Evidano fit
Evidano, cited in the article under its former name AILYZE, supported theme coding, pattern identification, and the organization of text segments. It generated preliminary document summaries, possible codes, and provisional thematic groupings.
The authors explicitly state that these outputs were not treated as final findings. They manually reviewed the full texts, compared the generated structure with the bibliometric results, refined and reorganized the codes, and created the final theme labels and interpretations.
What the review found
The literature is technically active but uneven. Earlier work emphasized centralized and decentralized optimization, simulation, and multi-agent logistics. More recent studies increasingly use reinforcement learning, intelligent agents, adaptive routing, and delivery-zone planning.
The final synthesis organized the evidence around five themes: real-time operational optimization and adaptive delivery control; decentralized coordination, simulation, and fleet orchestration; platform-market mechanisms, auctions, and intelligent trading; sustainability, consolidation, and crowdsourced last-mile delivery; and sociotechnical governance, labor behavior, and policy-safe AI.
Most evidence concerned operational outcomes such as response time, routing, utilization, consolidation, and vehicle distance. Live deployments were limited, and governance topics, including transparency, worker effects, accountability, privacy, fairness, and customer appeal mechanisms, lagged behind the technical literature.
Why this case matters
This is a clear example of AI-assisted evidence synthesis with two independent analytic lenses. Bibliometric mapping showed how the field was structured by keywords; Evidano organized concepts and arguments in the documents; researchers reconciled the two and made the final interpretive decisions.
The case is relevant to literature reviews where teams need both a high-level map and a source-sensitive thematic account. It also demonstrates appropriate disclosure: the article names the tool, describes its preliminary role, and states that generated outputs were not accepted as findings without manual review.
Limits and responsible interpretation
The final corpus contained only 16 studies and used one database. The authors did not conduct citation snowballing, and the small corpus makes keyword maps sensitive to individual papers. The review combined conceptual, simulation, modeling, and limited empirical work, so it did not perform a formal quality appraisal.
The conclusions are a map of an emerging research field, not proof that agentic AI is already mature in live Q-commerce operations.
How Evidano fit the workflow
- 1Define the intersection
Search terms targeted agentic or autonomous AI together with rapid fulfilment, on-demand delivery, and last-mile logistics.
- 2Screen the literature
Twenty-nine Scopus records were reduced to a 16-study conceptual corpus.
- 3Map the field bibliometrically
VOSviewer identified co-occurring keywords and the relationship between operational, learning-based, and coordination-oriented research.
- 4Generate a qualitative first pass
Evidano produced summaries, candidate codes, and provisional themes from the full texts.
- 5Compare analytic lenses
Researchers checked the qualitative structure against the bibliometric map and returned to the source papers.
- 6Refine by hand
The final themes, gaps, and theoretical implications were developed by the authors.
Bibliometric keyword maps and Evidano’s thematic first pass were reconciled by the authors.
The paper states AI outputs were reviewed against full texts before any theme was accepted.
Codes were refined, reorganized, and relabelled by the researchers.
The article names the tool and its preliminary role.