Agentic AI Use cases

Only the forecasts that are off reach the planner, with the cause.

Forecast exception and bias review scores every item-location each planning cycle, finds where the statistical line or a planner override is off, explains the cause, lets stable items pass through untouched and routes only the exceptions to the demand planner.

FORECAST EXCEPTION AND BIAS REVIEW, AS DESIGNED · A301Score4,800 item-locationsagainst the versionederror band02Pass through4,737 inside the band: thebaseline is written03Explain63 breaches: an override,an outlier, a moved signal04DecideL. Tan accepts, edits orrejects, with a reasoncode05Releasethe consensus plan locked· requirements to the ERPreason codes and decisions tune the band and the next proposals
The mechanism in five steps: the stable items pass, the breaches are explained, and only those reach the planner. Showcase estate; a designed role, yet to be built.
90%
of planning processes run touchless in AB InBev’s advanced markets, a vendor-published figure (o9 Solutions, 9 June 2025)
dated source
38%
less planning time with a light-touch review, at matched accuracy, across 3.5 million decisions (International Journal of Forecasting, 2025)
dated source
5+ days
to build a demand plan at a typical consumer goods company (McKinsey, 2 March 2022)
dated source
63 of 4,800
item-locations breach the band on the showcase estate’s cycle day; the rest pass through
showcase estate
A moment in the day

Where this use case starts

LT
Thursday 09:00 · Penang · cycle day

It is cycle day, and the scene chapter 3 used for A3. Overnight the forecasting agent, a role built from the blueprint, scored 4,800 item-locations against version 2 of its rules. For 4,737 the refreshed statistical line sat inside the error band versioned with the rule, so it wrote the new baseline to the planning suite and kept each job status. Sixty-three breached the band, and those are L. Tan’s queue at 09:00.

For material 7741-X at plant 1210 the last three overrides raised the statistical line by 18% on average, and actuals came in below the line each time. The reason code on each was “customer upside”, and the Northbay Logistics forecast upload shows flat demand. The agent proposes to withdraw the October override and to mark the family as stable, so that next month it passes through untouched.

L. Tan works the 63: 58 accepted, 3 edited, 2 rejected with a reason. J. Ridley locks the consensus plan at 11:30. Juno writes the key figures to the planning suite, releases the planned independent requirements to SAP S/4HANA, and keeps the job status and the API response against rule version 2.

Why this is a top use case

What the market is doing, with the evidence

Gartner points early agentic efforts at high-volume planning activities with a low cost of error and names touchless forecasting for stable SKUs first; most current agentic capability, it notes, improves query interpretation, recommendations and conversational support (Gartner, 20 May 2026) 1. Forecast review is high in volume, measured every cycle and cheap to get wrong on one item, and a demand plan still takes more than five days to build at a typical consumer goods company (McKinsey, 2 March 2022) 12.

The offers are dense. Logility ships four demand agents, one of which detects forecast winners and laggards and reveals bias patterns, each switchable between autonomous and planner-in-the-loop (Logility, 7 April 2026) 45; the first Kinaxis Agent Studio example analyses forecast quality across business units (Kinaxis, 5 February 2026) 18; o9’s decision replay agent diagnoses overrides, pricing assumptions and safety-stock policy as root causes, and AB InBev runs 90% of planning processes in its advanced markets touchless (o9, 9 June 2025) 39.

Daybreak’s demand agent reviewed 1,200 product-site decisions in one cycle, submitted four in five under policy and held 240 for review, and a consumer goods manufacturer reports $7M a month less inventory; an independent review notes that the figures are vendor-published (Daybreak, product page; Lokad, August 2026) 51 52.

The research on overrides is older and independent. Across 60,000 forecasts at four supply chain companies, upward planner adjustments were much less likely to improve accuracy, and small adjustments often damaged it (International Journal of Forecasting, 2009) 59. Across 3.5 million decisions, a light-touch method that asks the planner only about the items that need judgement matched accuracy with 38% less planning time (International Journal of Forecasting, 2025) 60. That design is the one this use case follows.

By hand today

How it is done today

After the statistical run, the demand planner exports the forecast and the last one to three months of error, accuracy and bias by item-location, to a spreadsheet, sorts by error and volume, and reviews the worst 50 to 100. History is checked by eye for outliers: promotions, one-off orders, stock-outs recorded as low demand.

Sales and key accounts are e-mailed for market input, and the planner waits. Overrides go in cell by cell in the planning suite or the spreadsheet, often with a habitual uplift. The consensus meeting with sales, marketing and finance locks the plan, which is released to the ERP as planned independent requirements.

A demand plan still takes more than five days to build at a typical consumer goods company (McKinsey, 2 March 2022), across thousands of item-locations per planner, on a monthly cycle with a weekly refresh. The time goes to the items that were fine, because the planner has to look at each one to know.

Day 1
Statistical run and exportThe forecast and three months of error land in a spreadsheet, by item-location.
Day 2
Sort and reviewThe worst 50 to 100 item-locations, checked against history by eye.
Day 3
E-mail sales and waitMarket input from account managers, when the replies come.
Day 4
Override cell by cellA habitual uplift on the items the planner knows.
Day 5 and after
Consensus and releaseThe plan is locked and sent to the ERP as planned independent requirements.

The cycle by hand: more than five days at a typical consumer goods company (McKinsey, 2 March 2022). Illustrative day split.

Stage by stage

How Juno enables it

This is the A3 scene of the ladder, seen from the machinery. The agent decides and executes within versioned rules, the error band being the rule, and L. Tan handles the exceptions only. The estate already holds the sales orders and the planned independent requirements from the first week; the planning suite’s forecast versions and the sell-through feeds would be added.

01
Register
You would register the forecast versions (statistical, consensus, final) by product, location, customer and period from the planning suite’s demand module; sales and order history and planned independent requirements from SAP S/4HANA; the acquired division’s demand data from Oracle Fusion Demand Management; design wins from the CRM; the promotion calendar and lifecycle status; and the distributor sell-through feeds and customer forecast files (EDI 830) that land in the lake.
02
Map
The ontology’s forecast object carries the version, the item-location, the period and the lag; the sales order and the planned independent requirement sit beside it, and forecast error, bias and value added are measures on the same objects. Overrides with their reason codes map to the forecast version they changed, which is what lets the agent see that three “customer upside” overrides ran above the line three times.
03
Ask
L. Tan would ask “Which item-locations at plant 1210 have been overridden upwards three cycles running and come in below the statistical line?” and “What is the forecast value added of our overrides on controller boards this year, by planner?” ChatLima answers from the mapped forecast versions and actuals, with the query on show.
04
Teach
The team would add “FVA” for forecast value added, the desk’s names for the segments (“stable”, “lumpy”, “new”, “end of life”) and the reason codes as a vocabulary with their meanings, so “customer upside” is a term the agent can count. The Thursday review would be saved as a question path.
05
Act
The trigger is the statistical run completing, period close, or a sell-through or customer forecast upload beyond its band. At A3 the agent would refresh the baseline for every item-location inside the error band, write it to the planning suite, and route each breach to L. Tan with a proposed correction; the lock stays with J. Ridley. The suite returns the job status and the ERP the API response.
06
Learn
Forecast accuracy and bias by item-location and lag improve, and the share of item-locations touchless each cycle rises as families earn the stable flag. Each accepted, edited and rejected correction is recorded with its reason code, so the band and the proposals move towards what the planner would have done, and the value added of overrides is reported each quarter against the statistical line.
Exception queue · L. Tan · forecast review · Thursday 09:00 · 63 of 4,800 breached the band · rule version 2
Item-locationBandMissCauseProposed correctionDecision
7741-X · 1210±12% · lag 1+18%, three cyclesoverrides “customer upside” · customer upload flatwithdraw the October override; mark the family stableAccepted 09:20
8802-K · 1210±15% · lag 1−22%, last perioda stock-out recorded as low demandrestore history with the outlier flagAccepted 09:26
5513-B · 1220±10% · lag 1+9%, two cyclesdesign-win pipeline moved out a quarterlower the baseline by 6%Edited to 4% · 09:41
9021-M · 1220±20% · lag 1+31%, last periodlaunch ramp ahead of its curvemove to the new-product ramp curveRejected: keep the ramp · 09:55
59 more rowseach with its causeWorked by 10:40 · 58 · 3 · 2

L. Tan’s screen on cycle day: one row per breach, with the band, the miss, the cause and the proposed correction. Showcase estate; illustrative rows.

Who does what

The person and the agent

The agent A3 Delegated
  • Scores every item-location each cycle against the error band versioned with the rule
  • Refreshes the baseline for the items inside the band and writes it to the planning suite
  • Explains each breach: an override pattern, an outlier in history, a moved signal, a lifecycle change
  • Proposes the correction and the segment flag, and records what the planner did with each
63 exceptions with the band, the miss and a proposed correction; the decisions come back
The person
  • L. Tan accepts, edits or rejects each proposed correction with a reason code
  • A sales or account manager supplies the market reason when an item is routed for input
  • The demand planning manager approves changes above the threshold; J. Ridley locks the consensus plan
  • J. Harrison checks that the volume plan matches the revenue plan before pre-S&OP

The market’s usual level is inform and draft-and-approve for the exceptions, with execute within rules for stable items: of the vendors cited above, one sets each agent to autonomous or planner-in-the-loop with a switch, one keeps a person in the loop throughout, and two submit decisions under policy. The blueprint designs statistical baseline forecasting and bias analytics at A3, consensus orchestration at A2, and the plan sign-off and lock at A0. The two agree, and the split is the point: the stable items run delegated, the lock stays human.

What you would configure

The settings on this agent

Agent settings · Forecast exception and bias review
ElementSettingSet byChanges
ObjectiveKeep forecast error inside the band for every item-location at plant 1210, and put only the breaches in front of the plannerL. Tan, plannerNew version
Data scopeForecast versions, actuals, overrides and reason codes, sell-through, customer forecasts, lifecycle status; controller boards at plants 1210 and 1220A. Novak, data leadNew version
RulesThe review policy and its error band by segment, version 2: ±12% at lag one for stable items, ±20% for new itemsL. Tan with the Planner AgentNew version, simulated first
TriggerWhen the statistical run completes each cycle, at period close, and on a sell-through or customer forecast upload beyond its bandL. Tan, plannerAt once
Actions permittedWrite the baseline · propose a correction · set the segment flag · request sales input · release the consensus plan on the lockL. Tan, ownerNew version
LimitsA correction above 15% of volume or $50,000 of revenue waits for the manager · confidence floor 0.80 · the lock is a person’sL. Tan and J. Harrison, financeNew version
EscalationBreaches to L. Tan · items flagged as sales bias to the account manager with the evidence · above the threshold to the demand planning managerL. Tan, ownerAt once
LevelA3 Delegated for the baseline inside the band; the lock at A0 with J. RidleyJ. Ridley, directorUp on evidence, down in one step
FeedbackEvery accept, edit and reject, with its reason code, is recorded and used in the next cycleAlways onFixed

Every row is a setting or a versioned rule; the level and the limits are the ones a manager signs.

Built from the blueprint

The roles this use case would draw on

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.

Blueprint · designed, yet to be builtThese roles are designed on the map. Each is configured from the same parts.
Plan 11 roles
Demand signal ingest & cleanse A3
Statistical baseline forecasting A3
Forecast bias & accuracy analytics A3
Consensus forecast orchestration A2
Channel sell-through signal A3
Design-win demand shaping A2
New-product ramp curve model A2
EOL & last-time-buy demand A2
Foundry customer demand intake A2
Demand plan sign-off & lock A0
Plan-attainment scorecard A3

Additional capability in Plan-to-Produce lists the rest of the map for this chain.

How it is measured

The measures that show it working

Measure
Forecast accuracy and bias by item-location and lag (case baseline: SKU-level accuracy up 10 to 12%, McKinsey)
Forecast value added of overrides against the statistical line
Share of item-locations touchless each cycle
Planner hours per cycle (baseline more than five days per plan, McKinsey; 38% less time with light-touch review)
Expedite events caused by forecast error (baseline 49%, APQC via Logility)
Fill rate and finished goods inventory (case: fill rate up 3 to 5%, finished goods down 6 to 8%, McKinsey)

Baseline forecast value added first: the accuracy of the final plan against the statistical line, by item-location and lag, because it shows in one number whether the overrides the agent withdraws were helping. The share of item-locations touchless each cycle comes next, read from the run record, so the 4,737 of 4,800 on cycle day has a history behind it. Planner hours per cycle is the measure the desk feels, and the five days of the manual path is its starting point.

Terms on this page
Error bandThe forecast error range, by segment and lag, inside which the agent refreshes the baseline itself.
Forecast value addedThe accuracy of the final forecast against the statistical line; positive when overrides helped.
BiasError that runs the same way cycle after cycle, above or below the actuals.
TouchlessAn item-location whose forecast went through the cycle untouched by a planner.
Planned independent requirementThe demand figure the planning run reads from the ERP once the consensus plan is released.
Questions
Why let stable items through untouched at all?

Because the research says small and upward adjustments often damage accuracy, and their time comes from the items that need judgement. The band is versioned: a team that wants it narrower sets it and simulates the last three cycles first.

What stops the agent from smoothing away a real change in demand?

A moved signal is a trigger: a sell-through or customer forecast upload beyond its band brings the item to the planner with the evidence. A launch or an end of life takes the item out of the stable segment.

Who owns the reason codes?

The demand planning manager, as a vocabulary the agent counts. A code that keeps appearing on overrides that ran above the line is what the quarterly value-added review is for.

Next in this chain
The question this leaves

The plan is right more often. Between the plan and the shop floor sit the parameters that turn it into orders: safety stock, reorder points, lead times. Who keeps those current?

Answered next

Safety stock and replenishment parameter tuning

The agent would recompute each parameter against the service target and the measured variability every cycle, write the small changes and hand the large ones to L. Tan.

Sources

Dated sources used on this page

  1. [1] Gartner, Gartner Warns of Agent Washing Risks in Supply Chain Planning Technology Market, 20 May 2026 (checked 23 September 2026) · https://www.gartner.com/en/newsroom/press-releases/2026-05-20-gartner-warns-of-agent-washing-risks-in-supply-chain-planning-technology-market
  2. [12] McKinsey, Autonomous supply chain planning for consumer goods companies, 2 March 2022 (checked 23 September 2026) · https://www.mckinsey.com/capabilities/operations/our-insights/autonomous-supply-chain-planning-for-consumer-goods-companies
  3. [18] Kinaxis, Kinaxis Introduces Maestro Agent Studio (press release), 5 February 2026 (checked 23 September 2026) · https://www.kinaxis.com/en/news/press-releases/2026/kinaxis-introduces-maestro-agent-studio-unlocking-next-level-decision
  4. [39] o9 Solutions, o9 Co-Founder and CEO Charts the Next AI-Driven Frontier in Enterprise Planning and Execution, 9 June 2025 (checked 23 September 2026) · https://o9solutions.com/articles/o9-co-founder-and-ceo-charts-the-next-ai-driven-frontier-in-enterprise-planning-and-execution/
  5. [45] Logility (Aptean), DemandAI+ on Aptean AppCentral (press release), 7 April 2026 (checked 23 September 2026) · https://www.logility.com/press-release/demandai-on-aptean-appcentral/
  6. [51] Daybreak, Daybreak home and product pages (Dawn and Sol agents, customer figures), page carries the current year only (checked 23 September 2026) · https://daybreak.ai/product
  7. [52] Lokad, Review of Daybreak, AI-Native Supply Chain Planning, August 2026 (checked 23 September 2026) · https://www.lokad.com/review-of-daybreak-ai/
  8. [59] International Journal of Forecasting (Fildes, Goodwin, Lawrence, Nikolopoulos), Effective forecasting and judgmental adjustments: an empirical evaluation and strategies for improvement in supply-chain planning, vol. 25(1), pp. 3-23, 2009 (checked 23 September 2026) · https://ideas.repec.org/a/eee/intfor/v25y2009i1p3-23.html
  9. [60] International Journal of Forecasting (van der Staak, Basten, van de Calseyde, Demerouti, de Kok), Light-touch forecasting: A novel method to combine human judgment with statistical algorithms, vol. 41(2), pp. 440-451, 2025 (checked 23 September 2026) · https://ideas.repec.org/a/eee/intfor/v41y2025i2p440-451.html

See Juno answer and act on your own data

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.