Retail AI agent · Dynamic pricing
Dynamic-Pricing & Markdown AI Agent
Propose markdowns from your own sales history and genuinely public data, freeze anything a live emergency declaration reaches, and hold every price for the pricing manager who releases it.
- Typical delivery
- 4–6 weeks
- Deployment
- Your stack
- Pricing manager
- Own data only
- After launch
- Agent Care
What this agent does
Proposes the price, never releases it
In
- 01
Ingest sales, cost, stock and calendar data from the retailer's own systems and genuinely public sources.
- 02
Normalise the fields, and carry every input forward with the feed it came from.
Reason
- 03
Estimate demand and propose a markdown or a price move inside the approved band.
- 04
Apply the pricing rules, bands and channel-parity requirements configured for the banner.
- 05
Check the provenance of each input, and reject anything a competitor could have supplied.
Decide
- 06
Freeze any SKU that a live emergency declaration reaches in a served geography.
- 07
Route every proposed price to the pricing manager who releases it.
Out
- 08
Retain the inputs, the rule, the proposal, the approver and the geographies published to.
- 09
Execute write actions only inside the approval boundaries agreed during implementation.
Product statement
The agent proposes a price; the pricing manager releases it — and algorithmic pricing is now an antitrust question, not only a commercial one.
Example workflow
One price change, input to release
- 1Price trigger receivedA markdown calendar, a stock position, a cost change or a rule review
- 2Inputs gatheredSales history, stock, cost, calendar and published list prices, each with its feed
- 3Price proposedOld price, new price, the rule that produced it, the channels and confidence
- 4Guardrails appliedProvenance checks, band checks, emergency-declaration checks, channel parity and confidence threshold
No human action required
Stages one to four run unaided and no price moves at any of them — the agent is proposing, and the pricing manager's lane opens at the confidence gate.
Goes to the pricing manager to release.
Adds a legal read first.
Pricing-manager approval
The change is held with its inputs, their provenance and the confidence.
Approve · Adjust · Send to legal review
Approved — released to publish
- 6Price systems updatedOnly where write access and approval policy allow it
- 7Change evaluatedMargin and sell-through, gate failures, channel divergence and post-release corrections
Rollbacks
Every pricing-manager adjustment is counted in the evaluation.
What should not run autonomously
Human approval stays in control
Outside the boundary — human approval required
8 items
- Releasing any price to a live channel.
- Approving a move outside the agreed band.
- Onboarding a new price feed or data source.
- Pricing on any input drawn from an identifiable shopper.
Automation boundary
Agent acts unaided
- ✓Estimate demand from the retailer's own sales history for the named owner.
- ✓Propose markdowns inside the approved band and queue.
- ✓Reconcile shelf, till.
- ✓Freeze a SKU when a legal gate fails, and hold it for approval.
Any write happens inside the boundaries agreed at implementation, never ahead of approval.
- Wording and placement of an algorithmic-pricing disclosure.
- Certifying a former price before a strike-through runs.
- Accepting a vendor model or rule-set upgrade.
- Remediation after a wrong price has been charged.
Example output
One price change, annotated
Everything the agent proposes is attached to the inputs it was drawn from.
Pricing output · single SKU
Illustrative example- SKU
- Seasonal apparel line
- Proposed change
- Markdown inside the approved band for the clearance window
- New price
- $34.00
- Input provenance
- Own sales history
- Confidence
- 93%
- Personal data
- None used
As received
Taken from the retailer's own systems and published list prices.
Inputs used
- Own sales history
- Stock and cost position
- Published list price
Why this price
It sits inside the band the pricing manager approved — the move they weigh.
What the score decides
Below the configured threshold the change picks up a legal read before it reaches the pricing.
Action
ApproveAdjustSend to legal reviewValue
Where AI adds value
The same four claims, placed at the point in the workflow where each one applies.
Where the value lands
Value 01 – 04
Every price change
From the retailer's own systems
Price from your own data
Draw on first-party sales, cost and stock, and the bands the pricing manager approved.
Move the price, keep the record
Routine markdowns inside the band arrive already worked out.
Send review to the risky moves
Gate failures and low-confidence proposals are marked, so the pricing manager's read starts where exposure concentrates.
Build an evidence trail
Each price keeps its rule, its inputs and the approver who released it.
Integrations
Typical integrations
Five system groups connect to the same agent. Which of them are in scope is decided in discovery.
Price and markdown optimisation
- Revionics
- Blue Yonder Pricing
- Pricefx
- PROS
Price science and lifecycle
- DemandTec / Acoustic
- Zilliant
- Eversight
- Bloomreach
Published-price monitoring
- Competera
- DataWeave
- Public list and ad feeds
Agent
Dynamic pricing & markdown
- Reads your own data
- Proposes the price
- Holds for approval
Channels and workflow
- Price book
- POS
- ESL feed
- Web PDP
- marketplace
- ad feeds
Observability & evaluation
- OpenTelemetry
- Langfuse
- Supported monitoring/evaluation sources
Integration availability depends on the client's existing systems and API access.
Agent controls
Six layers between the model and the shelf
Each control contains the next; the map below names the gaps.
L6 · Outermost — last line of defence
Inward → L1 · closest to the model
- L1 – L2
- Test whether a price may stand
- L3
- Puts the release in a pricing manager's hands
- L4 – L5
- Hold the price history and the rule behind it
- L6
- Freezes prices at the last approved set
How Nestack evaluates it
Evaluate the pricing workflow — not only the final number.
Coverage runs the whole depth of the workflow, and every layer is cut by slice.
Surface — the price the shopper pays
- E1Final-output evaluationDid the published price clear every gate it had to clear?
- E2Step-level evaluationDid the agent use the right inputs, bands and jurisdiction configuration?
- E3Tool evaluationDid it write the correct SKU and the correct channel?
- E4Confidence calibrationDo low-confidence proposals actually attract more adjustments?
- E5Slice evaluationHow does performance change across specific price cohorts?
- E6Business outcomeHow many changes needed an adjustment or a correction after release?
Floor — the price that is charged
Failure modes
Where each failure originates in the agent
Seven modes plotted against the pricing lifecycle.
Agent lifecycle
Direction of processing →
01 · Retrieval
1 mode
- DP-03
Competitor feed ingested
A co-listed seller's promotional calendar is read as market data.
Stage gathers
Sales history, stock, cost, with the source each came from
02 · Reasoning
2 modes
- DP-04
Emergency read as seasonality
Demand under a declared emergency is priced as an ordinary spike.
- DP-06
Former price reconstructed
A one-day reversal is taken as the price regularly offered.
Stage proposes
Old price, new price, the rule and confidence
03 · Tool / write
2 modes
- DP-02
Disclosure not emitted
A personalised markdown publishes without the required label.
- DP-05
Channels left divergent
Shelf and web take the new price while the price book times out.
Stage writes
Only where write access and approval policy allow it
04 · Output
1 mode
- DP-01
Former price unsupported
A strike-through renders against a price too old to support it.
Stage returns
The price the shopper is charged
05 · Change / Version
1 mode
- DP-07
Silent targeting regression
A vendor update re-enables geography below the level configured.
Stage tracks
Model, prompt, rule bands and jurisdiction config
- Sev-1 · a price released outside the boundary
- Sev-2 · a non-compliant price reaches a channel
- Sev-3 · a feed degrades, the SKU freezes
Affected slices
Overall compliance can hide one cohort
An aggregate gate-failure rate can look settled while one cohort of price changes carries almost all of it. Nestack reports that rate by slice as a lift on the all-price-change.
Slice performance — reported separately, not only in aggregate
Illustrative example| Slice | Failure rate | Lift | Lift vs. threshold | Status |
|---|---|---|---|---|
| Emergency geographies, personalised | 5.4% | 2.7× | Review | |
| Strike-through in strict states | 4.2% | 2.1× | Review | |
| Shared-engine concentrated categories | 2.4% | 1.2× | Watch | |
| Single-channel, non-personalised | 1.2% | 0.6× | Normal |
Bar: gate-failure lift vs. the all-price-change baseline · scale 0–4.0× · tick marks the 2.0× review threshold
2 of 4 slices over threshold
Evidence-linked improvement
A cycle is not closed by a fix
What closes a cycle is a case in the suite, not agreement about what went wrong That suite is what the following detection is measured against..
Improvement cycle · five stages
Switchback — the path turns at Improve and returns at Learn
- 01
Detect
Gate failures rise in one price cohort.
- 02
Diagnose
Until the cause narrows to a single feed, rule or prompt, the pricing manager keeps reading the runs behind the failures.
- 03
Improve
Each change is version-stamped and linked to the run that surfaced it.
- 04
Verify
The release waits on the affected cases passing again.
- 05
Learn
It becomes a permanent case and a change to the pricing guardrails.
Learn → Detect
The return edge. Detection next time runs against a longer suite than this one.
Typical build scope
Twelve workstreams across six weeks
The build scope read against the delivery timeline. Week structure follows the six-week plan — discovery, sources, pricing workflow, evaluation, integration, then production validation and handover.
Workstream
Week 1Week 2Week 3Week 4Week 5Week 6
- 01Pricing workflow discovery and boundary definition.Week 1
- 02Price-engine and margin-source assessment.Week 1, Week 2
- 03Rule, band and jurisdiction mapping and rule mapping.Week 1, Week 2
- 04First-party sales, stock and cost ingestion.Week 2
- 05Proposal logic and input-provenance.Week 2, Week 3
- 06Confidence scoring and gate-failure routing.Week 3
- 07Pricing-manager approval.Week 3, Week 4
- 08Price-book, shelf and channel integration.Week 5
- 09Pricing regression cases.Week 4, Week 5
- 10Guardrails and release controls.Week 4
- 11Price-history instrumentation.Week 5, Week 6
- 12Deployment, documentation and Agent Care handover.Week 6
12 workstreams · 6 weeks · bar shows the weeks a workstream is active — several run in parallel
Final scope and sequence confirmed in discovery
Engagement tiers
What each tier includes
Rows are the capabilities named in each tier's scope. Higher tiers include everything below them.
| Capability✓ in scope · — not at this tier | PilotOne banner, one channel | ProductionProduction price systems | AdvancedMultiple banners / markets |
|---|---|---|---|
| Introduced at Pilot | |||
| Proposals to your rules and bands | ✓ | ✓ | ✓ |
| Pricing-manager approval | ✓ | ✓ | ✓ |
| Price-integrity baseline | ✓ | ✓ | ✓ |
| Introduced at Production | |||
| Reporting by category | — | ✓ | ✓ |
| Approval workflow in your systems | — | ✓ | ✓ |
| Approved write-back to price systems | — | ✓ | ✓ |
| Price-engine integration | — | ✓ | ✓ |
| Introduced at Advanced | |||
| Multi-banner rule sets | — | — | ✓ |
| Multi-stage pricing approvals | — | — | ✓ |
| High SKU and channel volume | — | — | ✓ |
| Enterprise pricing controls | — | — | ✓ |
| Build price | From $5,000 | From $8,000 | Custom quote |
- Final build price
- Confirmed after discovery based on integrations, workflow complexity, transaction volume, approval controls and deployment requirements.
- Separate from build
- Build pricing is separate from recurring Agent Care, which covers managed monitoring, evaluations, incidents and verified improvements after launch.
What we need from you
What you bring, and what we build with it
Each input maps to a piece of build scope and a week in the delivery timeline.
You bring
We build with it
01Your price rules, bands and markdown calendar
→Rule, band and jurisdiction mapping
Week 102Representative past price changes
→Proposal baseline, elasticity and provenance binding
Week 203The feeds you price from, and where each one comes from
→Rule and band mapping, and the automation boundary
Week 104Access to relevant APIs, feeds or exports
→Price-engine, POS and channel assessment, then integration setup
Week 205Prices that should not have gone live
→Pricing cases and failure-mode testing
Week 406Which prices may never move unattended
→Guardrail bands, release routing and controls
Week 307A named pricing manager to release prices
→Pricing-manager approval workflow, then pilot and production validation
Weeks 5–6
Nothing else is required
Deployment, documentation and Agent Care handover are ours.
Delivery timeline
Four phases across six weeks
Phases occupy real weeks, and evaluation overlaps launch in the fifth rather than being stretched.
Phase
W1
W2
W3
W4
W5
W6
- DiscoveryW1
- BuildW2 – W3
- EvaluateW4 – W5
- Pilot & LaunchW5 – W6
Week focus
W1Pricing workflow discovery, rule mapping and the automation boundary
W2Cost and competitor-public feeds in place
W3Proposal workflow, confidence logic and release controls
W4Pricing cases and release guardrails
W5Channel integration, pilot price changes and targeted corrections
W6A live pricing cycle released by the pricing manager, then handover
- Reading the band
- Each bar spans only its named weeks; the overlap in week 5 is real work.
- At the end of W6
- A verified cycle in production, then monitoring sits with Agent Care.
- Duration
- Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.
Next step · Retail AI agent
Build a pricing agent that can show where every input came from.
Show us the feeds you price from, the bands you work inside and who releases a change. A price that has already been charged cannot be unwound by a refund alone, so we set the provenance, emergency and disclosure gates before anything moves.
- Typical delivery
- 4–6 weeks
- Deployment
- Your price systems
- Pricing manager
- Own data only
- After launch
- Agent Care available
Nestack Agents · Dynamic pricing and markdown
AGT-RT-08 · Agent Care available after launch