Five end-to-end chains run a manufacturer: from the order to the cash, the requisition to the payment, the forecast to the finished goods, the entry to the report, and the idea to the launch.
A role on the map is a design; configured from the eight parts, simulated and released, it becomes an agent.
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.
The map was drawn with a high-tech manufacturer’s processes as the model, on APQC’s Process Classification Framework and the SCOR Digital Standard: five chains, 25 phases, 114 process steps and 960 roles, each with a designed starting level. The five roles of the gamut in chapter 3 sit on it, and PO change and cancellation is the one that has been built.
This chapter lights twenty of them: in each chain, the three or four use cases the market is deploying now, chosen on evidence. Each has its own page, from the moment in the day to the settings and the roles it would draw on. Everything else on the map appears on the chain pages as additional capability, grouped by phase.
The twenty were chosen on evidence, and the pages say where that evidence is thin.
Each chain has a research note that ranks 13 to 15 candidate use cases on the same four criteria, in this order of weight, from sources checked 23 September 2026. A candidate entered the top four only where the blueprint holds a role for it, so every page can name the roles it would draw on. Everything below the line became additional capability.
The evidence is deep for finance and purchasing and thinner for product development. Each chain page says how its four beat the rest of its candidate table, and each use-case page marks the facts that rest on one source or on a vendor’s own claim.
| Chain | Candidates ranked | Dated sources | Where the evidence is deep, and where it is thin |
|---|---|---|---|
| Order-to-Cash | 14 | 53 | Deep. Cash application has two analyst benchmarks of its own and named results above 95% straight-through; order intake ranks fourth on thinner adoption evidence. |
| Procure-to-Pay | 13 | 68 | Deep for invoices and sourcing. The surveys behind purchase order exceptions speak of managing purchase orders in general; Lima’s own production agent is the strongest proof point. |
| Plan-to-Produce | 14 | 64 | Deep for planning, with the highest adoption figures of any chain. Customer numbers for scheduling agents are the thinnest of the four, and the expedite-cost figure rests on a single source. |
| Record-to-Report | 14 | 60 | Deep. The densest field of named agents in any chain; the customer numbers for anomaly detection are vendor claims. |
| Idea-to-Market | 15 | 59 | Thin. Analyst surveys have yet to rank agentic use cases inside product development, so the ranking leans on vendor offers, named customers and measured pain, and the pages say so. |
Candidates and sources are counted from the five research notes, checked 23 September 2026; the notes hold the full ranking tables.
Four worked use cases; 161 more roles on the map.
Order-to-Cash turns revenue into cash, and most of its cost sits in people handling documents and exceptions: PDF orders, orders held behind a credit decision, remittances that arrive apart from the payment, short-pays that need a reason code. Companies that redesign the process around AI run it at 52% to 59% lower cost (The Hackett Group, 15 July 2026)1.
The evidence is deep and favours the finance end. The median company auto-matches 70% of payments, leaving up to $15 million unapplied at a large enterprise (The Hackett Group, 11 August 2025)6. One adopter recovers $124 million of invalid deductions a year (HighRadius, 10 September 2026)7, and SAP, Oracle, BlackLine and Esker all advertise named collections agents8, 9, 10, 11. Order intake ranks fourth on thinner adoption evidence.
Match every incoming payment to its invoices, post it the day it arrives, and code each short-pay.
Read the use case →Rank overdue accounts by cash impact and payment likelihood, send the right reminder, and follow every promise to pay.
Read the use case →Turn every short-pay, debit memo and distributor claim into a case with its evidence and a validity call.
Read the use case →Read orders from every channel, create them in the ERP, and clear the credit, delivery and data blocks.
Read the use case →Additional capability in Order-to-Cash: 161 more roles on the map, designed and yet to be built, grouped by phase. blueprint
Four worked use cases, one of them in production; 116 more roles on the map.
Procure-to-Pay is the most transactional of the five chains, and the market measures it in cost per document: $9.40 for the average supplier invoice against $2.78 at the top performers, 9.2 days against 3.1, and 22% of invoices in exception against 9% (Ardent Partners 2025, via WEX, 20 July 2026)2. About 28% of procurement staff time goes on bids and purchase orders (Gartner, in Supply Chain Management Review, 18 February 2026)12.
Invoice matching is the clearest case in the whole set: analyst adoption numbers, published benchmarks and one manufacturer at 87% touchless invoices (IT Brief UK, 16 April 2026)13. Sourcing has the best named results in procurement14. Purchase order change and cancellation is the use case Lima runs in production production · anonymised; the surveys behind it speak of managing purchase orders in general. Supplier risk takes fourth place on three analyst surveys (The Hackett Group, 17 March 2026)15.
Capture supplier invoices, match them to order, receipt and contract price, and clear the blocked ones by cause.
Read the use case →Build the request from the last award, invite suppliers, normalise the bids, negotiate inside a playbook, draft the award.
Read the use case →Runs in productionCompare open purchase order lines with demand each morning; draft cancellations, reschedules and chasers by policy.
Read the use case →Screen suppliers and parts against financial, sanctions, geopolitical and product-change signals; draft the response for the owner.
Read the use case →Additional capability in Procure-to-Pay: 116 more roles on the map, designed and yet to be built, grouped by phase. blueprint
Four worked use cases; 179 more roles on the map, the largest chain.
Plan-to-Produce turns demand into a supply plan and finished goods, and it is where analysts find the most AI in use: 74% of enterprises report AI in S&OP and 72% in advanced planning and scheduling (The Hackett Group, 14 May 2026)16. The pain sits with planners, who spend more than half their time sorting dashboards3, and on the floor, where unplanned downtime costs the Fortune Global 500 about $1.4 trillion a year (Siemens, via AEMT, 1 August 2024)17.
Shortage triage has the widest vendor coverage in the chain; expedite spend runs at 3% of logistics cost for top performers and 10% for the rest (APQC, via Logility, 30 April 2026, a single source)18. Forecast exceptions and replenishment parameters are the two activities Gartner names as the place to start, high in volume and low in cost of error (Gartner, 20 May 2026)19. Schedule repair carries the thinnest customer evidence of the four.
Find the components that will stop a build, price each shortage by the orders at risk, and resolve it.
Read the use case →Find the item-locations where the forecast or an override is off, explain the cause, and route only the exceptions.
Read the use case →Keep safety stock, reorder points and planned lead times current against the service target and the measured variability.
Read the use case →When a machine stops or a lot is held, propose the re-sequence that protects committed dates, and release ready orders.
Read the use case →Additional capability in Plan-to-Produce: 179 more roles on the map, designed and yet to be built, grouped by phase. blueprint
Four worked use cases; 133 more roles on the map.
Record-to-Report turns the period’s transactions into a closed ledger, and the close has barely moved in a decade: 59% of companies close within six business days, against 60% in 2019 (ISG, 12 December 2023)4. 94% of teams close in Excel, with reconciliations the most time-consuming close activity and accruals the second (Ledge, 2025)20, and 18% of accountants make a financial error every day (Gartner, via Accounting Today, 21 February 2024)21.
Reconciliations win on every criterion: 55% of accounting teams name them their top automation opportunity while 5% call them highly automated (FloQast, 11 August 2026)22. Anomaly detection has the clearest survey signal, used by 34% of finance functions (Gartner, 18 November 2025)23; its customer numbers are vendor claims. Accruals and flux commentary follow on 2026 agent releases and a 40% shorter reporting cycle at one adopter (CFO Dive, 12 February 2026)24.
Tie every balance-sheet account to its source each day, draft the adjusting entry, and run the close calendar.
Read the use case →Score every journal as it posts against the company’s own patterns, flag the unusual ones, and hold the material ones.
Read the use case →Calculate period-end accruals from open orders and receipts awaiting invoices, and propose each entry with its basis.
Read the use case →Compare every account with the prior period, the prior year and the budget, drill the movements, and draft the explanation.
Read the use case →Additional capability in Record-to-Report: 133 more roles on the map, designed and yet to be built, grouped by phase. blueprint
Four worked use cases; 163 more roles on the map; the thinnest evidence, and the pages say so.
Idea-to-Market runs from the roadmap decision to production release. Verification takes 60% to 70% of a chip project’s effort (Siemens, 3 September 2025)25; first-silicon success fell to 14% and 75% of projects run late (Semiconductor Engineering, 27 March 2025)5. The median engineering change takes 7.0 days (APQC, checked 23 September 2026)26, and over 620,000 electronic parts were discontinued in 2025 (Z2Data, 13 March 2026)27.
Here the evidence is thinnest: analyst surveys have yet to rank agentic use cases inside product development, so the ranking leans on vendor offers and measured pain. Regression triage leads by a distance, on named agents from Synopsys, Cadence and Siemens28, 29; they do the root cause inside the design tools, and Juno’s roles would run the flow around them. The other three touch PLM, two ERPs and the planning suite at once, and the map across those systems is the enabling layer.
Run the nightly regressions, group failures by signature, assign owners, and keep coverage and bug convergence on one page.
Read the use case →Find every assembly, plant, stock position, open order and supplier a design change touches, and draft the change record.
Read the use case →Catch every end-of-life notice on active parts, size the last-time buy, and propose qualified alternates for approval.
Read the use case →Check every readiness item before the first production order, fix what rules allow, and route the rest to owners.
Read the use case →Additional capability in Idea-to-Market: 163 more roles on the map, designed and yet to be built, grouped by phase. blueprint
The role you sit in decides the first use case; the first week runs the same way for all of them.
| Role | First use case | Why start there |
|---|---|---|
| Supply planner (L. Tan) | Purchase order change, cancellation and confirmation chasing | The one agent that runs. Her policy sentence is the starting point, and her Monday shrinks from 2.5 days to 38 minutes on the showcase estate. |
| Buyer (D. Okafor) | Supplier risk, sanctions and end-of-life monitoring | Supply continuity is the top procurement priority for 2026 (The Hackett Group, 17 March 2026)15. The roles start at A1 and A2, so the buyer keeps every decision and gains the screening. |
| Finance lead (J. Harrison) | Cash application and remittance matching | The median company auto-matches 70% of payments (The Hackett Group, 11 August 2025)6. It is the A4 scene of the gamut: the agent posts, and she audits a sample each week. |
| Data lead (A. Novak) | Engineering change impact across BOM, stock and orders | One change touches PLM, two ERPs, the planning suite and the supplier portal. The map across them is the enabling layer, and it is hers to build in the first week. |
| IT architect (S. Iyer) | Invoice matching and exception clearing | The best-measured cost per document in any chain ($9.40 against $2.78)2, and write-back through the ERP’s released interfaces, switched on only after a confirmed test write. |
Whichever row is yours, the first week is the same: two source systems, one weekly decision, three people for two hours a day. The use case sets which tables are registered on the Monday, which question is asked on the Wednesday and which policy is written on the Friday. The next chapter walks that week day by day.
One agent runs in production: the PO cancellation agent, at a high-tech manufacturer. The other roles are designed on the map, and each is built from the same eight parts when a client chooses it, simulated first and released at a level a manager signs.
Each chain’s research note ranks 13 to 15 candidates on adoption evidence, vendor offers, named results and measured pain, from sources checked 23 September 2026. The top three or four won on that evidence, and each chain page says how they beat the rest.
Yes. Every chain page lists its additional capability by phase. A role there is configured from the same parts as a worked use case, and the first week takes one weekly decision of your choosing.
They are the market evidence: how many serious vendors offer a named agent for a use case is one of the four criteria. Each name carries a dated source, and the use-case pages keep to the same rule.
What does the first week actually look like?
Registered on Monday, answering on Wednesday, acting on Friday, on two of your own systems.
Each one builds on the one before it.
Reads your rows in place.
Connect and move ↗One record per real thing.
Master and govern ↗Apps for every decision.
Operate and decide ↗Asks and acts, at your level.
Ask and act ↗Bring two systems and one decision your planners make every week. In the first week we register the data, confirm the mapping with your team, answer your questions in ChatLima and run the first agent at the level you choose.
