AI agents · capability

Agents that keep the plan true, live.

Specialist agents re-forecast, chase POs, stage replenishment and draft markdowns against the same plan — most of it applies on its own, inside the limits you set.

Planning·Agents · Action layerLast sync · just now

The plan, turned into moves

The same ML that runs each surface now proposes real moves — gated by hard guardrails, tiered by risk, logged end to end.

Agents · this week
142118 approved
83%Auto-run
24Awaiting you
100%In limits
31hHours saved
Every action grounded, gated, logged
3 tiers
risk gate
100%
checked vs limits
0
irreversible w/o sign-off
31h
planner time back
Proposed · this week
142 actions ›
Reorder · Yarrow +640uTier 1 · autoOTB ✓
Record delivery · PO-4127Tier 2 · approveheld
The numbers behind it

Why agent leverage gives the week back to planning.

14 pts

Full-price sell-through gap between retail leaders and the industry — most of it down to how reliably the plan stays current.

Source: Incisiv × WRC × Anaplan 2026

~53%

Share of unplanned markdown cost attributed to upstream decisions that didn't get made in time.

Source: Coresight industry research

Always on

Specialist agents re-forecast, chase POs, stage replenishment and draft markdowns inside the limits you set — day and night.

Source: Tightly

The problem today

What breaks before AI agents run to plan.

Planning·Manual planning · this week
Planner hours · manual runs
31h/wk
hand-driving macros
Decisions made
40
of 120 needed
Avg decision lag
6 days
signal → action
Run outside budget
12
actions over OTB
Reruns · fat-finger
9
redone by hand
Planner hours · by task
31h/wk hand-run
hours sunk hand-running each task · this week4h8h9hReforecast6hRebalance7hReorder5hChase4hMarkdowntask run by hand →
Decisions waiting · going stale
80 unmade
Merino Crew · Charcoal · M
cover breached · reorder
5 d late
Bramble Pleated · Cream · S
22wk past EOL · markdown
8 d late
London ↔ DC · Wren Navy
store short · rebalance
4 d late
PO-4021 · Fabric Co.
delivery unrecorded · chase
6 d late
No guardrails · no audit
Over-buy caught after the PO ships
macros run blind — nothing checks limits, nothing logs why
Diagnosis — 120 decisions needed a week, 40 made by hand, none guardrailed or loggedSource: Incisiv × WRC × Anaplan benchmark 2026
How Tightly does it

Three steps from brief to an action you can audit.

Tightly · Agents live· three stages from data to decisionLive
01Propose
LIVE PLAN → PROPOSED ACTIONMerino Crew2.1wPROPOSEDReorderMerino Crew +640uwhy · cover breachedwithin OTB headroom

A move off the live plan

The same ML that runs each planning surface reads the current model and proposes a concrete action — reorder, rebalance, chase, markdown — with the reasoning attached, never a free-text guess.

from the live plan
02Check
GUARDRAIL GATE · FAIL-CLOSEDRISK TIEROTBCOVERTier 0ReforecastAUTO-RUNTier 1Reorder · in OTBAUTO-RUNTier 2Bulk import · unclassifiedHELD

Every guardrail, fail-closed

Each action is scored against three risk tiers and checked against OTB headroom and days-of-cover. Within limits, it auto-runs; over the line — or anything unclassified — it holds for a human.

risk tier · OTB · cover
03Approve
INBOX → AUDIT TRAILRecord delivery · PO-4127Tier 2 · irreversible · awaiting youApproveAUDIT · PO-4127proposedDemand agentguardrails passed3 tiers ✓approvedMark TillPO drafted #4127logged

Approve, then it's logged

Held actions land in the inbox with the guardrail record. Nothing irreversible runs without a name on it — and every step, from proposal to sign-off, is written to the audit trail.

inbox → audit trail
See it run

Your team's leverage, not your replacement.

Specialist agents re-forecast, chase POs, stage replenishment and draft markdowns against the same plan — most of it applies on its own, inside the limits you set.

Planning·Agents · InboxLive · every action checked before it runs
Proposed
142
actions · this week
Auto-run
118
Tier 0 / 1 · in limits
Awaiting you
24
Tier 2 · needs a name
Rejected
3
sent back by planner
Hours saved
31h
planner time back
Proposed actions · this week
Approve 24 held
ActionReasoningGuardrailsTierStatus
Reorder · Merino Crew +640ucover breached, within OTBOTB ✓ Cover ✓ Budget ✓Tier 1Auto-run
Rebalance · DTC→Retail 1,200uLondon short, DC longCover ✓Tier 1Auto-run
Markdown · Bramble −20%22wk past EOLFloor ✓Tier 1Awaiting you
Record delivery · PO-4127GRN from ERPirreversibleTier 2Awaiting you
Bulk cost import312 rowsunclassified → heldTier 2Held
Audit · PO-4127proposedguardrails passedapproved by Mark TillPO drafted #4127
Nothing irreversible runs without a name on it — within-limits actions auto-run, the rest hold for you31h planner time back
Why teams change

What changes once agents keep the plan true.

Eight roles, one team

Planning, Buying, Merchandising, Demand, Supply, Markdown, Finance and Analytics agents specialise in the decisions a real team makes.

They propose, you approve

Every move comes with a one-line rationale, the numbers it touches, and an Apply button. The audit log captures who applied what and why.

Inside the limits you set

Routines define what an agent may do unattended (rebalance within ±5% of OTB, replenish below MOQ thresholds) and what needs a human.

Learn from your team's calls

Approved and rejected proposals tune the next batch. The team's judgment compounds; the agents get sharper.

Your agents

Always on, always in policy

Every move an agent makes is within your limits and recorded in the audit log — you stay in charge of what ships.

Meet the agents
Tightly agent
just now · within your limits
Live

Re-forecast ready — 3 categories have drifted from plan this week. Want me to stage the moves for your review?

Drifted vs plan · this weekΔ wmape
Tailored Trousers+9%8%
Woolly Layers−12%11%
Activewear+5%9%
Rebalance 240u DC → SFRe-baseline OTB Q3Hold buy on OCN-072
Stage movesReview firstLogged · audit ready
What this replaces

The copilots you have today, and the agents Tightly delivers.

Macros · rules · generic copilots
REORDER.XLSM · RUN MACROno limits · no log
Sub AutoReorder()  ' fires blind
  For each SKU in Range("B2:B900")
    order = max - onhand  ' no OTB check
  Next  ' no audit, no reason
no limits, no logfire blindcan't reasonnot grounded in your plancan't act on the plan
Excel macros fire without a single guardrail and leave no trail; rules-only tools can’t reason; generic copilots chat but aren’t grounded in your plan and can’t act on it. Every real move still lands on a planner.
Tightly · Agents · the action layer
ACTION · GROUNDED · GUARDRAILEDfail-closed
Reorder · Merino +640uTier 1 · auto
GuardrailsOTB ✓ Cover ✓
Record delivery · PO-4127Tier 2 · held
Auditevery step logged
grounded in the live planguardrailed + risk-tieredfail-closedapproved + audited
The same ML that runs each surface proposes real moves off the live plan, checks each against three risk tiers and your hard limits, auto-runs what’s safe and holds the rest for a named human — every step written to the audit trail.
FAQ

Questions buyers ask, answered straight.

Something not covered here? Talk to the team.

Will agents commit a PO without me?

No. Every agent proposes; nothing commits without a human approval. The only exception is policy-locked moves (e.g. within-cover replenishment) and only if you opt in.

How do I set the policy bounds?

Per category × channel: margin floor, OTB envelope, reserve %, sign-off thresholds, pack constraints. Most teams configure once and revisit at cycle.

What's the audit log capture?

Every proposal, every override, every approval. Who, when, why, what changed. Exportable for CFO and compliance review.

Can different agents have different authority?

Yes. Demand agent might be approve-only; Replenishment agent might be auto-stage-then-approve. Configurable per agent and per organisation.

Your team's leverage. Specialist agents inside the limits you set.

There's nothing to rip out. Tightly runs on your existing ERP, EDI, e-commerce and POS. Give us 30 minutes and we'll show it on your own categories.