BidMatrix VP Mark Nedzelskii Says AI Agents Optimise Bids While Humans Set Strategy

September 12, 2026 · exchangewire.com

A premium finance app could see its cheapest registrations come from low-quality inventory with suspiciously fast conversion times, the kind of pattern an agentic AI system might read as a win. Mark Nedzelskii, VP of growth at BidMatrix, used that example in a September 8 interview with ExchangeWire to draw the line between where he trusts AI agents and where he does not. Agentic systems can already turn briefs into first-pass plans, check campaign setup, monitor pacing, flag anomalies and recommend routine optimisations, tasks Nedzelskii called measurable, reversible and governed. He measures his trust in an agent by the quality of its feedback loop, and named in-app inventory connected through MMP integrations to post-install events like KYC completion and first deposit as the strongest current environment for autonomous optimisation.

Campaigns run on agent-set objectives inherit whatever value ladder the strategist defines, from registration through KYC, first deposit, D30 activity and positive margin, so a badly specified brief lets the agent optimise the wrong step of that chain. Nedzelskii's advice points toward machine-readable briefs that state both the goal and the limits, such as increasing first deposits in a given market while keeping payback below a threshold and excluding named sources. Planners who write briefs at that level of precision give an agent clear guardrails; those who hand over a KPI alone risk an agent that hits the number while damaging the brand, overexposing an audience or buying inventory the business would reject. BidMatrix's MatrixSDK investment signals where the industry is headed: agencies that want more autonomous execution will need equivalent visibility into their own supply paths and post-install signals.