Analysts group a product like Lima Juno into published market categories: agentic analytics, conversational analytics inside analytics and business intelligence platforms, decision intelligence platforms, and AI agent development platforms.
Two of the four categories cover the conversation; two cover the agents. Juno carries both in one product.
Gartner opened this market for software that joins AI agents with analytics for data exploration, insight generation and workflow orchestration. ChatLima and the agent layer belong here.
Gartner Peer Insights · 2026Gartner lists conversational analytics, semantic modelling, agentic insights, analytics governance and data preparation as the mandatory features of an analytics and business intelligence platform.
Gartner · 29 June 2026Gartner published its first comparison report for this market on 26 January 2026. Its mandatory features include automated decision execution with human oversight, decision governance and outcome assessment.
Gartner · 26 January 2026Gartner covers agent platforms through an Emerging Market Quadrant of 8 June 2026 and its comparison report for AI application development platforms of 17 November 2025, whose mandatory features include grounding, orchestration and guardrails.
Gartner · 2025 and 2026These four categories describe where Juno fits, in the words each market definition uses. Recognition inside a category is the firms’ own decision: a Peer Insights listing follows a market alignment request, a Voice of the Customer position follows twenty eligible customer reviews inside eighteen months, and a briefing is open to any vendor.
Forrester reads the same ground through its Wave for AI Platforms, Q3 2026 (6 August 2026), which scores data modelling, agent development, governance controls and runtime architecture; its 2026 evaluation of the agent control plane market, which names agent inventory and identity, policies and guardrails, monitoring, control and audit; and its AEGIS framework of 22 August 2025, whose principle of least agency gives an agent the fewest capabilities, tools and permissions its job needs. dated source
Twenty-three features across four definitions; each row names a chapter or a capability page and its evidence label.
| Category | Mandatory feature | Where Juno demonstrates it | Evidence |
|---|---|---|---|
| Agentic analytics | Data source connectivity | Chapter 6, Register and Sources and connectors: 40 and more connectors for ERP, planning, warehouse and lake sources, read in place. | production |
| Agentic analytics | Natural language query | ChatLima and Question to query: a typed question, the intent and the objects recognised, the query written, run and shown. | production |
| Agentic analytics | Automated insights with narratives and confidence | Insights and forecasts: causes ranked with their rows and a confidence, trends, forecasts as a range. Recommendations carry a basis. | product |
| Agentic analytics | Agent workflow orchestration with self-correction | Chapter 5 and Outcome checks: the plan runs at the level set, the outcome is checked against the objective, each decision feeds the next cycle. | production |
| Agentic analytics | Data preparation that recognises semantic relationships | Chapter 6, Map and Linking: columns proposed against the ontology with a confidence and a reason; the same supplier in two systems recognised as one. | showcase estate |
| Analytics and BI | Data preparation | Data registration and Source sync: tables registered where they run; a sync scan lists what is new, changed or gone. | production |
| Analytics and BI | Analytics governance | Secure answers and Versioning: the asking person’s row and column rights apply to every answer; every published object is versioned. | product |
| Analytics and BI | Agentic insights | Monitoring and alerts watches a stated rule and raises a finding; Insights and forecasts explains the movement behind a number. | product |
| Analytics and BI | Semantic modelling | Chapter 6, the ontology and Ontology model: objects, attributes, relationships and measures, extended with your own; the API and the Power BI connector read the same meaning. | product |
| Analytics and BI | Conversational analytics | ChatLima, Answers and Follow-ups: plain language in, a table, a chart or a narrative out, the thread kept for the next question. | production |
| Decision intelligence | Decision execution by humans | Chapter 4 and Approvals: a named approver sees the draft, its basis and the limits checked, then releases or edits it. | production |
| Decision intelligence | Decision governance | Chapter 7 and Audit trail: signal, reasoning, recommendation, limits checked, approver, action, acknowledgement and outcome, per run. | product |
| Decision intelligence | Automated decision execution with oversight | Chapter 3 and Write-back: five levels set per agent; write-back through the ERP’s supported interfaces, with the acknowledgement kept. | production |
| Decision intelligence | Decision modelling | Chapter 5 and Planner Agent: a stated intent becomes a planning table of task, procedure, data and rule, simulated before release. | product |
| Decision intelligence | Decision service composition | Chapter 8: one rules engine, one rule store, one ontology and one audit trail serve every agent; the 960 designed roles of the blueprint would all be built on them. | blueprint |
| Decision intelligence | Monitoring and outcome assessment | Outcome checks, Observability and Learning loop: results checked against the objective; approvals, edits, overrides and exceptions counted per 1,000 decisions. | product |
| Agent platforms | Performance evaluation | Accuracy testing: fifty of the client’s own questions with expected answers, scored for correctness, speed and access, re-run after every change. | method |
| Agent platforms | Grounding in organisational knowledge | Chapter 6, Training and Grounding: the ontology, the client’s vocabulary, and the named policies and plans Juno relies on. | product |
| Agent platforms | A model catalogue | Model choice: Gemini, Claude, OpenAI, Llama or another cloud model at design time; a small model inside the estate at run time. | product |
| Agent platforms | Orchestration for agent collaboration | The rules engine reads every published plan; Arbitration ranks two agents that want opposite changes to one field and hands the rest to a named person. | product |
| Agent platforms | Framework support for assistants, agents and applications | ChatLima answers, the PO cancellation agent acts, and SmartApps put the approval queues and clocks in front of people. | production |
| Agent platforms | Guardrails | Chapter 4 and Token manager: value thresholds, daily action budgets, confidence floors and two-person approval, checked before every action; every model call metered. | product |
| Agent platforms | Cloud and hybrid deployment | Chapter 7: an open-source core in microservices and containers, on GCP, Azure, AWS or a private cloud; row data stays inside the estate. | product |
Mandatory features as read on Gartner Peer Insights in 2026 for agentic analytics, analytics and business intelligence platforms, decision intelligence platforms and AI application development platforms (sources 1, 2, 4 and 6 below). Chapters are the story; the other links open the reference tier.
Five labels run across the whole site; this table says exactly what each one covers, and what stands behind it.
| Label | What it covers | What carries it on this site | What a reviewer can check |
|---|---|---|---|
| production | Work running at a client today, described with the client anonymised. | The PO cancellation agent at a high-tech manufacturer: planners author the policy in plain language, the rules engine executes it on federated ERP data, write-back after a named approval, 100% of actions traced to a rule, zero token cost at run time. ChatLima on a client’s supply chain hub since 2023; 40 and more source systems; 200 and more business users. | A live session on client data under an agreement, and an anonymised run trace for one cancellation. |
| anonymised engagement | Results measured during four client engagements, published anonymised with the method shown. | Warehouse space and capacity, about €300K a year · stock re-routing at port, more than 2,000 recommendations and cash release above $500K · dynamic network selection, $3M of lost sales avoided and lead time down 11% · container utilisation up 23%. | The calculation behind each figure, the period it covers, and the client name under an agreement. |
| showcase estate | Invented values in one consistent example estate, so the mechanics can be followed end to end. | L. Tan’s Monday: 1,240 messages, 38 lines worth $412,000, 29 drafts, 38 minutes against 2.5 days; the first week: 23 tables, 412 columns, 371 proposed, 338 confirmed, a 4.2 s answer, a policy released at A2 and moved to A3 after 212 approvals. | A live walkthrough of the showcase estate, where every value on screen is the value quoted here. |
| blueprint | A design of what Juno can be configured to do; the roles are yet to be built. | 960 agent roles across five chains, 25 phases and 114 process steps, on APQC’s Process Classification Framework and the SCOR Digital Standard, each with a designed starting level. Twenty worked use cases show what would be configured; the rest is additional capability. | The use-case pages: the settings, the roles drawn on and the level, all in the conditional. |
| dated source | Market facts from Gartner, Forrester, IDC, McKinsey and others. | Every market number carries the publisher and the date in the sentence. | The numbered source list at the foot of this page, and the same treatment on every other page. |
Lima’s agent blueprint is a design of 960 agent roles across five end-to-end chains. It is the map of what Juno can be configured to do, and the first agents are built from it one at a time, the way the PO cancellation agent was.
So the count an analyst should carry away is one: one supply chain agent in production, with ERP write-back under a named approval. The blueprint is the map of what the next ones would be, and every page that draws on it uses the conditional and the label above.
Everything an analyst asks for in a first session, prepared and owned before the session begins.
| Line on the sheet | What it states | Status |
|---|---|---|
| Company and history | Founding year, offices in Palo Alto and Hyderabad, and the products in the platform. | ready |
| Team | Headcount by function on the date of the briefing. | confirmed before each briefing |
| Revenue band and growth | Held in a confidential sheet for the analyst, because these thresholds stay private. | shared in the session |
| Customers in production | Count by industry and by region, with the systems connected at each. | confirmed before each briefing |
| Reference customers | The clients who agreed to speak with an analyst, and what each one runs. | placeholder · analyst relations to fill |
| Production evidence | The PO cancellation agent; ChatLima since 2023; 40 and more source systems; 200 and more users. | ready |
| Engagement results | The four logistics results with the calculation behind each one. | method sheet in preparation |
| Agent blueprint edition | The high-tech and semiconductor edition: 960 roles designed, 114 process steps, five chains, yet to be built. | ready · labelled blueprint |
| Certifications | Each certification with the body that issued it and the date it was granted. | placeholder · security team to fill |
| Roadmap | The next two quarters, with the dates already committed to clients. | placeholder · product to fill |
| Pricing principles | How usage is metered, how budgets are capped, and the cost per decision where it can be stated. | in preparation |
| Demonstration | A 15-minute product tour that follows the mandatory-feature lists above. | ready |
Analyst relations keeps this sheet; each line carries an owner and the date it was last confirmed.
Forty minutes of reading covers the claims a briefing rests on, in the order they build.
Fourteen sources carry the facts on this page; each page on this site carries the sources for its own numbers.
Every analyst fact on this site is a press release, a public article or a public market definition, quoted with its firm and date. Gartner asks that every external use of its research go through its Content Compliance team, with the title, the authors and the date, and that its research be read as the opinion of its research organisation. Forrester asks for written permission for each external use and keeps citations inside eighteen months of publication. Approval is pending for the pages that carry these facts.
Agentic analytics covers ChatLima and the automated insights. Decision intelligence covers the agent layer, because its vocabulary of governance, oversight and outcome assessment matches the autonomy ladder directly.
One in production: the PO cancellation agent, at a high-tech manufacturer. The 960 roles of the blueprint are designed and yet to be built; each would be configured from the same parts, the way that agent was, and the use-case pages say so on every row.
The mapping screen, one ChatLima answer with its query, one policy in plain language with its planning table, one approval by a named person, one ERP acknowledgement and the audit record of all of it, on the showcase estate or on a client’s data under an agreement.
The Lima material on this page is free to quote with attribution. The analyst facts stay under each firm’s citation policy, so Gartner items go through Content Compliance and Forrester items need written permission.
We walk the six stages on live screens, show one production agent with its approval and its ERP acknowledgement, say plainly which roles are designed and yet to be built, and leave the facts sheet and the source list with you.
