The nine questions
The same nine questions, answered twice
Numbers in brackets point to the numbered sources at the foot of the page. The line under each question says what a complete answer contains.
1 · Where must the data live for questions and agents to work?
A complete answer names the systems the data has to be copied into, and the systems that are read where they stand.
AWSBedrock AgentCore runs agents inside the customer's own AWS account, working with several outside agent frameworks, the Claude Agent SDK, OpenAI's Agents SDK or custom code [1]. Amazon Quick's chat agent works over Quick Sight datasets and Quick Index connections to more than fifty sources, including cloud storage, Snowflake, Databricks and SharePoint [6].
Lima JunoYour rows stay in the systems that hold them. Juno registers each table through the Lima data fabric and reads it in place through the federated query engine, joining several sources in one answer. 2 · How far does it reach into SAP, Oracle and planning systems?
A complete answer names the interface used for each system and says which of them are live today.
AWSAgentCore's Gateway component turns a company's existing APIs into governed tools an agent can call, and Quick Index connects to more than fifty sources [6]. Specific connectors into SAP or Oracle transactional data stay unnamed on the pages reviewed, leaving that reach for a planning or ERP system as a task a client or partner carries out on top of the gateway [1].
Lima JunoMore than 40 connectors cover SAP ECC and S/4HANA, Oracle Fusion and E-Business Suite, planning suites, warehouse systems, lakes, databases, files and REST services. One question can join several of them, and the one agent in production, the PO cancellation agent, runs on federated ERP data the same way. 3 · How is the shared meaning, the semantic model or ontology, built, and by whom?
A complete answer says what arrives ready, what your team builds, and how long that modelling work takes.
AWSInside Amazon Quick, an author builds a Topic across as many as twelve datasets, adding descriptions, synonyms and custom instructions so the chat agent can traverse the data with shared meaning attached [5]. One enterprise reviewer reported the setup taking several months to configure well [7]. Bedrock AgentCore leaves ontology and business meaning to the customer's own data and knowledge bases [1].
Lima JunoLima’s supply chain ontology arrives filled in: materials, plants, suppliers, orders, stock and the relationships between them. Juno proposes a mapping for each registered column, and your team confirms it. 4 · How do business users ask questions, and is the working shown?
A complete answer shows the question, the query that ran, the sources read and the access rules applied.
AWSAmazon Quick's chat agent answers a question typed in plain language over Quick Sight dashboards and documents indexed in Quick Index, drawing on the Topic an author has enriched [5]. Whether the generated query or the reasoning behind an answer is shown alongside it stays undocumented on the pages reviewed.
Lima JunoPeople ask ChatLima in plain language. Each answer shows its working: the ontology terms used, the systems read, the seconds taken and the query itself, under the asking person’s row and column access. 5 · How are agents and rules defined, and by whom?
A complete answer says who writes a rule, in which language, and how it is reviewed before it runs.
AWSAgentCore is code-first: a developer brings an outside framework, the Claude Agent SDK, OpenAI's Agents SDK or custom code, and AgentCore hosts and runs it [1]. Amazon Quick's Topics are defined by an author who enriches each dataset with descriptions, synonyms and custom instructions through Quick's own configuration screens [5].
Lima JunoOne agent runs in production, the PO cancellation agent, built from a policy a planner stated in plain language. Lima’s agent blueprint designs 960 further roles across five chains, yet to be built the same way: a planner states the intent and the Planner Agent drafts the rules for review. 6 · How is autonomy set, and how do approvals work?
A complete answer gives the levels available, who approves what at each one, and how a level is raised and lowered.
AWSAWS's Policy component turns plain-language rules into enforcement at the gateway, separate from the agent's own code, so a policy change stays independent of changes to the agent itself [3]. AWS's own security guidance describes four scopes running from prescribed agency with approval on every action to full agency that starts itself, for a team designing its approval steps [3].
Lima JunoEach agent carries a level you set, A0 to A4. At A2 a named person approves every action; at A3 the agent acts inside value, budget and confidence limits and sends exceptions to people. The level is one of eight settings on an agent, each set by a named person. 7 · How do actions reach the ERP, and is each write confirmed?
A complete answer names the interface for each ERP and says where the acknowledgement of a write is kept.
AWSActions an agent takes travel through tools registered behind AgentCore's Gateway, which is where governance is enforced before a call reaches the underlying system [1]. A specific confirmation step for a write into an ERP such as SAP or Oracle stays undocumented on the pages reviewed, left as a detail for the tool a client builds behind the gateway.
Lima JunoWrite-back uses each ERP’s own supported interfaces: released APIs and BAPIs for SAP, REST resources and interface tables for Oracle. It starts per application after a confirmed test write, and every write keeps the acknowledgement. The PO cancellation agent in production writes back this way, under human approval. 8 · How does it learn from people’s decisions, and what is audited?
A complete answer says what is recorded for each decision and how that record shapes the next recommendation.
AWSAgentCore's memory component keeps long-term context across runs, and its Evaluations tool tests a new agent version against live traffic before it takes over fully [1]. Observability traces each decision for review, and Amazon Quick's Topic carries the synonyms and custom instructions an author has taught it over time [5].
Lima JunoEvery accept, edit and override is stored with the rule version, the facts at that moment and the outcome, so the next recommendation improves. The audit trail holds signal, reasoning, limits, approval, action and acknowledgement for each run; in production, 100% of the cancellation agent’s actions trace to a rule. 9 · What is the pricing unit, is there a trial, and where do analysts place it?
A complete answer names the unit of charge, the route to try it, and the analyst reports that cover the category.
AWSAgentCore bills by consumption: small charges per vCPU-hour, per gigabyte-hour and per call, with up to $200 of free-tier credit [2]. Amazon Quick charges $20 or $40 per user a month plus $250 per account, with a 30-day trial for 25 users [4]. Forrester names AWS a Leader in its Q3 2026 AI Platforms Wave; ISG lists it among the leaders in its 2026 AI Analytics guide [8][9].
Lima JunoJuno is licensed with the Lima platform and priced per engagement. The token manager meters every model call, and routine language help runs inside your estate at zero token cost. Twenty use cases are worked through on this site, and the analyst guide covers the markets.