Autopilot, human in the loop

Brief the need.
Approve the plan.
The platform runs it.

RetailPlus Agentic Solutions is autopilot for retail media: describe your campaign in plain language, agents build and run it, while you stay in control of the rules and approvals. Routine is automated, strategy stays yours.

Stop operating the platform. Start briefing it.

Today, media teams manage platforms manually — selecting audiences, configuring placements, setting pacing, and rebuilding campaigns for each retailer. Agentic Solutions flips that: you define the goal, the platform builds the plan, shows it for approval, and runs it once approved.

Before

You drive

Every decision is a form. Someone has to know which audience, which surface, which frequency cap, which pacing strategy, for every retailer, every time.

Now

You brief

You write the need in plain language, or your own agent submits it. The platform interprets the intent, fills the gaps, and proposes a plan.

Always

You stay in control

Nothing spends without your approval. Nothing runs outside the guardrails you set. Every agent decision is logged and explainable.

The decisioning flow

From brief to outcome, with one human gate.

This is what happens between the sentence you write and the campaign that runs. Six stages of automated decisioning, with a single human-in-the-loop approval gate where you sign off the plan before any budget moves.

  1. 01 Intake

    The need becomes a complete brief

    You state the need in plain language, or your agent submits it as a structured call. The intake layer turns it into a complete brief: objective, KPI, budget, flight, audience intent, hard constraints. If something essential is missing, it asks. If something can be inferred, it infers and flags the inference for you.

  2. 02 Decisioning

    The orchestration layer decides and dispatches

    The brief is decomposed into its parts. The orchestration layer holds the constraints that span all of them, the budget ceiling, the flight, the brand safety floor, weighs the trade-offs between reach, frequency and cost, and dispatches each part to the specialist agent that owns it.

  3. 03 Deployment

    The agent fleet works in parallel

    Specialist agents run simultaneously, not sequentially. The audience agent matches the cohort, inventory builds and prices the mix, creative checks fit and fills gaps, while compliance verifies POPIA routing. What once took a week now happens in parallel, in minutes.

  4. 04 Assembly

    The plan is assembled and costed

    The planner agent pulls every agent's output into one coherent plan: the audience, the surface mix, the creative, the cost, the forecast lift, and any inference made along the way, surfaced clearly rather than buried.

  5. Human in the loop

    You approve, adjust, or reject

    This is the gate. Nothing has been booked, nothing has been spent. You read the plan, usually in a couple of minutes, and approve it, change it, or send it back. For routine campaigns inside the limits you set, your own agent can approve on your behalf. The unusual always comes to you.

  6. 05 Live

    The campaign runs and optimises continuously

    Once approved, it goes live. The agents don't just set and forget — they track performance, reallocate budget, manage creative fatigue, and flag what needs attention. This runs 24/7, across every surface, for the entire flight.

    Runs continuously, not on a weekly review cycle.

  7. 06 Learn

    The outcome feeds back

    At the end of the flight, the result is stitched to actual purchase through the cleanroom and POS feed. The lift goes into platform memory, so the next brief on a similar need starts from what worked last time, not from a blank form.

Specialist agents, each owning one decision.

No single agent runs the campaign. Each owns a domain, with its own rules and data access, and does one job well. Buy-side agents work the brief, sell-side agents represent supply, while neutral agents handle compliance and commercials.

Buy-side

Planner agent

Decomposes the brief, builds the campaign structure, holds it together and presents it for approval.

Decides: the shape of the plan

Buy-side

Audience agent

Translates audience intent into a concrete cohort, queries the catalogue, negotiates rental terms, checks for overlap.

Decides: who to reach

Buy-side

Inventory agent

Selects the surface mix across retailers, checks availability for the flight, prices it, holds inventory pending approval.

Decides: where it runs

Buy-side

Creative agent

Matches approved creative to surface specs, flags gaps, briefs production or a creator for what is missing.

Decides: what gets shown

Buy-side

Measurement agent

Sets up attribution before the campaign runs, monitors pacing and lift mid-flight, feeds the result into memory.

Decides: what worked

Neutral

Compliance agent

Runs continuously, not as a step. Verifies cleanroom routing, consent coverage, cross-border rules and brand safety.

Holds: a veto, not a vote

Neutral

Settlement agent

Models the three-way split, generates the commercial terms, produces the single contract and the single invoice.

Decides: how the money flows

Sell-side

Retailer agents

Represent each retailer's pricing, availability and rules about who may run on their estate.

Represent: the supply side

Sell-side

Audience-provider agent

Represent each bank, bureau or telco: their rental terms, segment definitions and use-case restrictions.

Represent: the data side

Autopilot is not autonomy.

Agentic Solutions runs at autopilot level: the agents execute the routine campaign end to end. But the human keeps the decisions that are strategic, creative, relational or irreversible. The split is deliberate, and the boundary is yours to move.

The human keeps

The strategic and the irreversible.

  • Setting guardrails: budget limits, brand safety, approvals
  • Approving the plan before any budget moves
  • The brand strategy and the big creative idea
  • Handling exceptions and the unusual campaign
  • Relationship and trust decisions with retailers and partners
Two colleagues reviewing a campaign plan on a tablet
The agents handle

The routine and the reversible.

  • Decomposing the brief and assembling the campaign
  • Matching audiences and pricing the surface mix
  • Routine negotiation on price and terms
  • Activation, pacing and continuous optimisation
  • Settlement, attribution and feeding the result back

Every agent decision is logged and explainable.

You can always ask why this audience, why this price, why this surface, and get a real answer with the reasoning attached. The audit trail is continuous, not assembled after the fact. Autopilot means the agents do the work. It does not mean you lose sight of it.

Time back. Optimisation that never sleeps.

The first gain is time: the campaign build collapses from days of team work to a brief and an approval. The second gain is bigger and easier to miss. Agents optimise continuously, across every campaign at once, which is something a human team cannot sustain no matter how good it is.

The manual way

  • Campaign build

    Two days of team work across multiple retailers

  • Optimisation cadence

    A weekly check-in, if the team has time

  • Overnight

    Nothing happens. The campaign waits for Monday

  • Campaigns per person

    A handful, run carefully, at capacity

  • Where the time goes

    Assembly, configuration, reconciliation

With Agentic Solutions

  • Campaign build

    A brief, and a plan back in minutes

  • Optimisation cadence

    Continuous, around the clock, every campaign

  • Overnight

    Agents pace, reallocate and cap fatigue while you sleep

  • Campaigns per person

    Many times more, each one better optimised

  • Where the time goes

    Strategy, the creative idea, the approvals that matter

2 days →
minutes

From multi-retailer campaign build to a briefed plan

24 / 7

Continuous optimisation across every live campaign

1 brief

One sentence touches every relevant surface and retailer

The guardrails are structural, not promised.

An autopilot that moves real money needs its limits built into the architecture, not left to the agents to honour. The unsafe action is not something the agents are trusted to avoid. It is something the platform does not make available.

See the full POPIA mechanics

Budget locks and approval gates

Per-campaign and per-period ceilings. An agent cannot spend what the guardrail will not release. Anything above your threshold comes to you.

The cleanroom is the only source of truth

An agent cannot invent an audience. It can only query, match and book what actually exists, inside the cleanroom, never touching raw personal information.

The compliance agent holds a veto

POPIA routing, consent coverage and brand safety are checked on every plan. The compliance agent can stop a plan. It cannot be overruled by the buy-side.

Every decision is logged and recoverable

Explainability is a requirement, not a feature. Every plan comes with its reasoning. Every action an agent took can be retraced and queried.

See autopilot run on your brief.

Bring a real campaign need. We'll brief it into the platform live, walk you through the plan the agents assemble, and show you exactly where you stay in the loop.